0 INFO-VAX	Thu, 25 Jan 2001	Volume 2001 : Issue 50      Contents: Re: $setimr problem ' 36GB Disk drives in an RA310 Raid Array + RE: 36GB Disk drives in an RA310 Raid Array + Re: 36GB Disk drives in an RA310 Raid Array  Re: About X-Windows x ICA/RDP  Re: About X-Windows x ICA/RDP # Re: All microsoft web sites offline # Re: All microsoft web sites offline # Re: All microsoft web sites offline # Re: All microsoft web sites offline  Re: Another missed opportunity Re: Another missed opportunity Re: Another missed opportunity Backup Problem...  Re: Backup Problem...  Re: Backup Problem...  Re: Compaq press release4 Re: Compaq: A simple, affordable clustering solution4 Re: Compaq: A simple, affordable clustering solution4 Re: Compaq: A simple, affordable clustering solution4 Re: Compaq: A simple, affordable clustering solution4 Re: Compaq: A simple, affordable clustering solution4 Re: Compaq: A simple, affordable clustering solution4 Re: Compaq: A simple, affordable clustering solution Damned TCP/IP Services rant  Re: dec 3000 300 crashing  fast_io_copy.c questions# For Sale: 512 Mb memory DS20 (new!) # For Sale: Vaxstation 4000 model 90A ' RE: For Sale: Vaxstation 4000 model 90A   ftp'ing savesets from W2K to VMS$ RE: ftp'ing savesets from W2K to VMS Gnu-C for VMS? Re: Gnu-C for VMS?6 Re: Is Compaq dropping support for vms on vax platform$ Re: Linux worm and RedHat 7.0 broken$ Re: Linux worm and RedHat 7.0 broken$ Re: Linux worm and RedHat 7.0 broken$ Re: Linux worm and RedHat 7.0 broken$ Re: Linux worm and RedHat 7.0 broken$ Re: Linux worm and RedHat 7.0 broken$ Re: Linux worm and RedHat 7.0 broken* Midrange I/O with VMS WAS: DS20 - Slow I/O. Re: Midrange I/O with VMS WAS: DS20 - Slow I/O. Re: Midrange I/O with VMS WAS: DS20 - Slow I/O. Re: Midrange I/O with VMS WAS: DS20 - Slow I/O. Re: Midrange I/O with VMS WAS: DS20 - Slow I/O. Re: Midrange I/O with VMS WAS: DS20 - Slow I/O- Re: Moving from VMS to Tru64 Unix, Any Ideas?  MPACK/MUNPACK images Re: MPACK/MUNPACK images Re: Multiple path devices... Re: Multiple path devices...	 Netbackup  NETBIOS Aborting (PW 6.0C) Re: NETBIOS Aborting (PW 6.0C) Re: Newbie question  OpenVMS Consortium Re: OpenVMS Consortium OpenVMS problem  OpenVMS problem  Re: RW-CD drives for VMS' Re: Source for used StorageWorks disks? ' Re: Source for used StorageWorks disks? ' Re: Source for used StorageWorks disks?  Telnet, yes; HTTP, no  Re: Telnet, yes; HTTP, no  Re: Telnet, yes; HTTP, no  Re: Telnet, yes; HTTP, no  UCX Route problem - Re: Urgent:Error recovering files from backup - Re: Urgent:Error recovering files from backup 7 Re: VMS vs UNIX (was: Linux worm and RedHat 7.0 broken) 7 Re: VMS vs UNIX (was: Linux worm and RedHat 7.0 broken) 7 Re: VMS vs UNIX (was: Linux worm and RedHat 7.0 broken) 7 Re: VMS vs UNIX (was: Linux worm and RedHat 7.0 broken) 7 Re: VMS vs UNIX (was: Linux worm and RedHat 7.0 broken) 7 Re: VMS vs UNIX (was: Linux worm and RedHat 7.0 broken) 7 Re: VMS vs UNIX (was: Linux worm and RedHat 7.0 broken) 7 Re: VMS vs UNIX (was: Linux worm and RedHat 7.0 broken) 7 Re: VMS vs UNIX (was: Linux worm and RedHat 7.0 broken)  Re: VMS writable CD on Win2K?  Re: VMS writable CD on Win2K? I Warning: No type converter registered for 'String' to 'Pixmap'conversion. ! Where is the latest MIME for AXP? ! Re: Why is the tape write locked? I Why still VAX? Was: Re: Is Compaq dropping support for vms on vaxplatform ) Window for Alpha success still seems open  Re: Wizard Disappears!!!' Re: Writes flushed to Disk on Dismount? ' Re: Writes flushed to Disk on Dismount?   F ----------------------------------------------------------------------  # Date: Thu, 25 Jan 2001 12:49:27 GMT 0 From: John Santos <john.santos@post.harvard.edu> Subject: Re: $setimr problem> Message-ID: <MPG.14d9eb3bab7cbb149896ba@news.bellatlantic.net>  H In article <94ngun$s3a$1@nnrp1.deja.com>, cstranslations@msn.com says...7 > In article <Z8lb6.58$cu.523@gazette.loc1.tandem.com>, ( >   hoffman@xdelta.zko.dec.nospam wrote:D > > In article <94hsps$u35$1@nnrp1.deja.com>, cstranslations@msn.com	 > writes: $ > > :OpenVMS 7.1-1H2, BASIC V1.2-000 > > B > >   I'd encourage an upgrade to OpenVMS Alpha V7.1-2 with ECO... > > G > > :Q: Is there any reason as to why (or how) $setimr would decided to  > > :return ss$_bufferovr? > > I > >   Because some condition handler somewhere up the stack frame told it  > to?  >  > Ask a stupid question... > :-)  > H > >   I would strongly encourage explicit specification of an event flagI > >   or the don't-care event flag EFN$C_ENF.  (EFN$C_ENF requires V7.0.)  > E > Well - now that I've gotten beaten silly over event flags . . . the H > basic$starlet library that comes with V1.2-000 doesn't seem to provideC > the symbol (not that it likely to change from 128 or whatever the  > actual value is)  F Actually, BASIC$STARLET gets built when you install BASIC (or you can D rebuild it without re-installing BASIC) and I think it gets its infoG from STARTLET.MLB (or .OLB?), not from the BASIC kit itself.  At least, F I've always rebuilt it when upgrading VMS to be sure I have the latest definitions.  B EFN$C_ENF is there on VAX VMS V7.2, BASIC V3.9, in $EFNDEF module.  D Though I agree its value is unlikely to change, I've learned throughB bitter experience to parameterize everything and always used named
 constants.  E BASIC is pretty good about setting unused arguments on function calls H to 0 (not sure if it is 0 by value or a pointer to a longword containing4 a 0), but it is probably a good idea to be explicit.  H On Alpha, always use /nooptimize when debugging.  (I think you mentionedC in another followup that the return code actually came from another D system service that was re-ordered by the compiler.)  BASIC supports; /nooptimize and if you use /debug without it, it warns you.   C Someone else mentioned quad word support in BASIC.  This is only in C the Alpha version, and you need V1.4.  Not in VAX BASIC at least as , of V3.9 which just came out a few weeks ago.   --   John Santos    ------------------------------  # Date: Thu, 25 Jan 2001 11:38:26 GMT % From: Andrew.Rycroft@intrinsitech.com 0 Subject: 36GB Disk drives in an RA310 Raid Array) Message-ID: <94p37i$5c7$1@nnrp1.deja.com>    Hi,   C I have an RA310 Raid Array, which "officially" does not support the E 36GB disk drive. Was reading an early thread where it suggests the 3- ; port RAID controller would work with these disks within its C limitations, i.e. 32GB logical unit size. Would this be true of the  RA310 as well ?    Thanks Andrew     Sent via Deja.com  http://www.deja.com/   ------------------------------  % Date: Thu, 25 Jan 2001 13:49:39 +0000 5 From: "Steeples, Oliver" <Oliver.Steeples@compaq.com> 4 Subject: RE: 36GB Disk drives in an RA310 Raid ArrayN Message-ID: <F498D199EDB12D468CD2C66680D3080116C3F5@reoexc04.emea.cpqcorp.net>   The supported disks are: RZ1BB-VW LYG0 A01 V2.7 1,3  - RZ1BB-VW LYJO/0656 A01 V3.2 1,4,5,6,7,8,9,10   RZ1CB-VW LYG0 A01 V2.7 1,3  - RZ1CB-VW LYJO/0656 A01 V3.2 1,4,5,6,7,8,9,10   RZ1DA-VW 3B05 A01 V2.7 1,3  " RZ1DA-VW 3B05/B012/BDC4   V3.2 11  RZ1DB-VW LYG0 A01 V2.7 1,3  - RZ1DB-VW LYJO/0307 A01 V3.2 1,4,5,6,7,8,9,10  " RZ1DD-VW 3B07/B012/BDC4   V3.2 11 ' RZ1ED-VW 3B05/3B07/B012/BDC4   V3.2 11  % RZ1FC-VW 3B02/B012/BDC4   V3.2 11,12   RZ2CD-LS 0306 A01 V3.2 1,2   RZ2DD-LS 0306 A01 V3.2 1,2   RZ2ED-LS 0306 A01 V3.2 1,2    H The RZ1FC is a 36Gb disk so looks like it is supported but needs V3.2 of	 firmware.    	Oliver    -----Original Message-----% From: Andrew.Rycroft@intrinsitech.com ( [mailto:Andrew.Rycroft@intrinsitech.com]) Sent: Thursday, January 25, 2001 11:38 AM  To: Info-VAX@Mvb.Saic.Com 0 Subject: 36GB Disk drives in an RA310 Raid Array     Hi,   C I have an RA310 Raid Array, which "officially" does not support the E 36GB disk drive. Was reading an early thread where it suggests the 3- ; port RAID controller would work with these disks within its C limitations, i.e. 32GB logical unit size. Would this be true of the  RA310 as well ?    Thanks Andrew     Sent via Deja.com  http://www.deja.com/   ------------------------------    Date: 25 Jan 2001 05:54:23 -07001 From: nothome@spammers.are.scum (Malcolm Dunnett) 4 Subject: Re: 36GB Disk drives in an RA310 Raid Array, Message-ID: <t23Je9xa8lNu@malvm1.mala.bc.ca>  * In article <94p37i$5c7$1@nnrp1.deja.com>, *    Andrew.Rycroft@intrinsitech.com writes:   > Hi,  > E > I have an RA310 Raid Array, which "officially" does not support the G > 36GB disk drive. Was reading an early thread where it suggests the 3- = > port RAID controller would work with these disks within its E > limitations, i.e. 32GB logical unit size. Would this be true of the  > RA310 as well ?  > A    I don't believe so. The RA310 is an entirely different product A from the RA230(+). I'm not currently using any drives that big in = a RA310, but I'm pretty sure I tested some 45GB drives in one + previously and didn't observe any problems.    ------------------------------  % Date: Thu, 25 Jan 2001 07:40:45 -0300 ) From: fabio_compaq@ep-bc.petrobras.com.br & Subject: Re: About X-Windows x ICA/RDPL Message-ID: <OFCA35175C.12DEACC8-ON032569DF.0039A51F@ep-bc.petrobras.com.br>   Brass wrote:  ? "I personally don't like the idea of these porting environments < because they prevent the programmer to take advantage of the< typical strengths of each OS. Especially with VMS which is a> rich and very powerful OS that method is a no-brainer. What we; need are applications which emphasise the strengths of VMS. < This means the apps must be implemented that they use all of> VMS features where appropriate and they must be very reliable.< What we don't need is a host of apps of mediocre quality and= competing against each other. We need one outstanding app for ? each major task. We can compare this to the OS/2 and MacOS case ; where is only necessary to have at least one brilliant text ? processing program, one powerful DB, one modern browser a.s.o."     H So, why OpenVMS is OPEN.. following POSIX( Unix, Unix, Unix) standards?=    E If the idea of porting Unix code is accepted for OpenVMS sysmans, why  not port MS applications too ??    Regards    FC          : Brass Christof <brass@infopuls.com> em 24/01/2001 22:07:30      .       "Vajhoej, Arne" <Arne.Vajhoej@GTECH.COM>          michael.jackson@bcs.org.uk)       fabio_compaq@ep-bc.petrobras.com.br )       Christof Brass <brass@infopuls.com> 3       Christoph Gartmann <gartmann@immunbio.mpg.de> 3       Christopher Smith <chriss@Mufasa.pubserv.com> %       JF Mezei <jfmezei@videotron.ca> +       Jimmi Aakjaer <aakjaer@post7.tele.dk> H       "Ken Fairfield; SLAC: 650-926-2924; FAX: 926-3515" <Fairfield@SLA= C.Stanford.EDU>   & Assunto: Re: About X-Windows x ICA/RDP    > On Wed, Jan 24, 2001 at 06:56:24AM -0500, Vajhoej, Arne wrote:H > >> >                 We have tools like Bristol=B4s Wind/U, with enab= les 1 > >> > to port MFC code to OpenVMS, Unix, etc ...  > >>? > >> AFAIK then MS destroyed that company by simply refusing to - > >> license them for newer Windows versions.  > >>9 > >> I do not think many companies will want to follow in  > >> their foot-steps. > > > >  The idea of porting MS Windoze apps to VMS in entirety or7 > >  some limited 'core' subset, is an exercise in wish  > >  fulfilment. > 5 > Porting Windows environment to VMS is a great idea.  > 9 > But after what MS did to Bristol, then who will want to  > spend money on it ?  >  > Arne  > I personally don't like the idea of these porting environments< because they prevent the programmer to take advantage of the< typical strengths of each OS. Especially with VMS which is a> rich and very powerful OS that method is a no-brainer. What we; need are applications which emphasise the strengths of VMS. < This means the apps must be implemented that they use all of> VMS features where appropriate and they must be very reliable.< What we don't need is a host of apps of mediocre quality and= competing against each other. We need one outstanding app for ? each major task. We can compare this to the OS/2 and MacOS case ; where is only necessary to have at least one brilliant text > processing program, one powerful DB, one modern browser a.s.o.         =    ------------------------------  % Date: Thu, 25 Jan 2001 11:44:48 +0000 ) From: Christof Brass <brass@infopuls.com> & Subject: Re: About X-Windows x ICA/RDP, Message-ID: <3A7011B0.1D180DFD@infopuls.com>  * fabio_compaq@ep-bc.petrobras.com.br wrote: >  > Brass wrote: > A > "I personally don't like the idea of these porting environments > > because they prevent the programmer to take advantage of the> > typical strengths of each OS. Especially with VMS which is a@ > rich and very powerful OS that method is a no-brainer. What we= > need are applications which emphasise the strengths of VMS. > > This means the apps must be implemented that they use all of@ > VMS features where appropriate and they must be very reliable.> > What we don't need is a host of apps of mediocre quality and? > competing against each other. We need one outstanding app for A > each major task. We can compare this to the OS/2 and MacOS case = > where is only necessary to have at least one brilliant text A > processing program, one powerful DB, one modern browser a.s.o."  > I > So, why OpenVMS is OPEN.. following POSIX( Unix, Unix, Unix) standards?  > G > If the idea of porting Unix code is accepted for OpenVMS sysmans, why ! > not port MS applications too ??  > 	 > Regards  >  > FC  > Sorry that I wasn't that clear. Porting Micro$oft applications< to VMS might be as helpful as porting them to MacOS or OS/2.@ There is nothing wrong except that they might not live up to the? standards of VMS, i.e. reliability and good system integration. = And if there were other programs around like StarOffice which = are able to read and write the Micro$oft propriertary formats 7 (M$-Word, M$-Excel a.s.o.) this porting effort could be  superfluous.? But what is important: if a program is ported it has to be done > well. If you use some of these multi-platform environments you7 lose the ability to use all the features of your target  environment in full details.   ------------------------------    Date: 25 Jan 2001 10:06:17 +0100* From: eplan@kapsch.net (Peter LANGSTOEGER), Subject: Re: All microsoft web sites offline* Message-ID: <3a6fec89$1@news.kapsch.co.at>  \ In article <3A6F9B05.8B037ABD@videotron.ca>, JF Mezei <jfmezei.spamnot@videotron.ca> writes: >Keith Brown wrote: . >> I'm sure it will be billed as a feature ;-) > G >CNN mentioned the web site outages, but said that Microsoft blamed DNS K >problems on the internet and none of its systems were incapacitated or had I >been attacked. (i.e. its not the fault of mircosoft or of our software).   D Yes, the problem was in all (4) of M$'s own nameservers (*.MSFT.NET)
 Coincidence ?    --  < Peter "EPLAN" LANGSTOEGER           Tel.    +43 1 81111-2651; Network and OpenVMS system manager  Fax.    +43 1 81111-888 < <<< KAPSCH AG  Wagenseilgasse 1     E-mail  eplan@kapsch.netH A-1121 VIENNA  AUSTRIA              "I'm not a pessimist, I'm a realist"   ------------------------------  % Date: Thu, 25 Jan 2001 09:43:47 +0000 % From: Alan Greig <agreig@my-deja.com> , Subject: Re: All microsoft web sites offline8 Message-ID: <s7tv6t46v91mqrsgva0j3bd0jp9pk01p3n@4ax.com>  , On Wed, 24 Jan 2001 22:18:38 -0500, JF Mezei% <jfmezei.spamnot@videotron.ca> wrote:    >Keith Brown wrote: . >> I'm sure it will be billed as a feature ;-) > G >CNN mentioned the web site outages, but said that Microsoft blamed DNS K >problems on the internet and none of its systems were incapacitated or had I >been attacked. (i.e. its not the fault of mircosoft or of our software).   C But it appears it was the fault of Microsoft. The Register reports:   B Another reader pointed that, according to the whois database, bothC microsoft.com and msn.net rely on the name servers at xxx.msft.net    C He said that Microsoft has all it's name servers in one subnet, and 1 that this subnet was either down or unreachable.    D Cotse.com, a specialist Web site for computer technicians, said thatA use of only one subnet broke one of the 'golden rules' of network E engineering and meant Microsoft was effectively putting "all its eggsN in one basket".n  01 http://www.theregister.co.uk/content/6/16340.htmlm   --
 Alan Greig   ------------------------------  % Date: Thu, 25 Jan 2001 10:00:32 +0000r% From: Alan Greig <agreig@my-deja.com>R, Subject: Re: All microsoft web sites offline8 Message-ID: <46uv6tome213mk6jkjj7h6dj3fkp8jsdk1@4ax.com>  , On Wed, 24 Jan 2001 22:18:38 -0500, JF Mezei% <jfmezei.spamnot@videotron.ca> wrote:e   >Keith Brown wrote:e. >> I'm sure it will be billed as a feature ;-) >tG >CNN mentioned the web site outages, but said that Microsoft blamed DNS K >problems on the internet and none of its systems were incapacitated or hadeI >been attacked. (i.e. its not the fault of mircosoft or of our software).f  B Update: Microsoft admits it was their fault From www.microsoft.com  % Microsoft Explains Site Access Issuesd  > On Tuesday evening and Wednesday, many Microsoft customers had@ difficulty accessing the company's Web sites. The cause has been' determined, and the issue is resolved. r  9 At 6:30 p.m. Tuesday (PST), a Microsoft technician made acE configuration change to the routers on the edge of Microsoft's Domain E Name Server network. The DNS servers are used to connect domain namesoF with numeric IP addresses (e.g. 207.46.230.219) of the various servers4 and networks that make up Microsoft's Web presence.   C The mistaken configuration change limited communication between DNSaA servers on the Internet and Microsoft's DNS servers. This limiteda@ communication caused many of Microsoft's sites to be unreachableD (although they were actually still operational) to a large number of+ customers throughout last night and today. e  C This was an operational error, and not the result of any issue with C Microsoft or third-party products nor the security of our networks.nC Microsoft regrets any inconvenience caused to customers due to thisD issue. w  F At approximately 5 p.m. Wednesday (PST), Microsoft removed the changesE to the router configuration and immediately saw a massive improvementg in the DNS network.   @ All sites are currently available to customers. Again, Microsoft" apologizes for the inconvenience.  --
 Alan Greig   ------------------------------  # Date: Thu, 25 Jan 2001 11:20:42 GMTe= From: system@SendSpamHere.ORG (Brian Schenkenberger, VAXman-)u, Subject: Re: All microsoft web sites offline0 Message-ID: <009F6A1F.68AC7C84@SendSpamHere.ORG>  \ In article <3A6F9B05.8B037ABD@videotron.ca>, JF Mezei <jfmezei.spamnot@videotron.ca> writes: >Keith Brown wrote:n. >> I'm sure it will be billed as a feature ;-) > G >CNN mentioned the web site outages, but said that Microsoft blamed DNSMK >problems on the internet and none of its systems were incapacitated or haduI >been attacked. (i.e. its not the fault of mircosoft or of our software).   L Truth be know, Bill was just having a temper tantrum over the DoJ lawsuit.    I "Nah, nah, nah, nah, nah nah!  If I can't keep MS together to conquer the I world and subsume everything into my empire", screamed Bill while kicking.H his feet on the floor in a childish fit of rage, "I'm just taking every-* thing MS and going home... So there! Nah!"   --O VAXman- OpenVMS APE certification number: AAA-0001     VAXman(at)TMESIS(dot)COMy            rO city, n., 1. a place where trees are cut down and streets are named after them.    ------------------------------  % Date: Thu, 25 Jan 2001 02:42:36 -0500p- From: JF Mezei <jfmezei.spamnot@videotron.ca>r' Subject: Re: Another missed opportunitya, Message-ID: <3A6FD8EA.DC5253A8@videotron.ca>   jlsue wrote:F > lower than Unix - so what?  Unix still has quite a long way to go to> > reach the level that OpenVMS and VMSclusters currently have.  N Not according to Compaq's sales staff and according to Inform. Both claim thatS UNIX has the best clustering and that it has absolutely all the stuff that VMS had.a  > VMS's technical superiority is of no use if it is kept secret.   ------------------------------    Date: 25 Jan 2001 09:58:27 +0100G From: Jan Vorbrueggen <jan@mailhost.neuroinformatik.ruhr-uni-bochum.de>.' Subject: Re: Another missed opportunity H Message-ID: <y4elxs80uk.fsf@mailhost.neuroinformatik.ruhr-uni-bochum.de>  / JF Mezei <jfmezei.spamnot@videotron.ca> writes:   I >> If more than a sound bite was required, I think he risked scaring the  > >> herd-followers who were an important audience in this case.O >Those analyst conference calls are just an opportunity to collect sound bytes.   J Methinks not. AFAIK, if an officer of corporation were to lie public aboutL anything related to the company's business, the SEC and the DA are likely toJ strongly disapprove if they found out, and a class-action suit by the poor8 investors who lost their money would also likely follow.   	Jan   ------------------------------  % Date: Thu, 25 Jan 2001 09:34:30 +0000 % From: Alan Greig <agreig@my-deja.com>T' Subject: Re: Another missed opportunityo8 Message-ID: <7trv6tgullfcehfaik55j5ta3bkgaks65e@4ax.com>  E On Wed, 24 Jan 2001 22:13:29 -0500, rdeininger@mindspring.com (Robertk Deininger) wrote:t  ` >In article <g7tt6toisp02dr4o9tesdpverujfv7e3dc@4ax.com>, Alan Greig <agreig@my-deja.com> wrote: >s >eI >> Come on if he had actually *said* *something nasty all hell would havelE >> broken loose. He's managed to actually be nasty but with plausibler- >> deniability. I don't believe he is stupid.s >> .G >> Capellas is *well* aware of the sensitivity of VMS customers and the G >> issue of mentioning everything other than VMS in end of year reportsgC >> has explicitly been raised before. There is no excuse this time.l >s >If he's so diabolically clever, why didn't he dishonestly add in the sentence or two you claim is needed.  He could have made a bundle for Compaq by bamboozling you and the rest of us.   E How about because he doesn't want to boost VMS sales - especially not > to new customers. The more customers VMS has when the 'freeze'7 announcement comes the more bad feeling Compaq creates.A  D Just imagine if some company switched form Solaris to VMS because it@ was a better platform for their particular need and they had theF confidence to do so because everyone from Capellas down was signallingF that this was a viable thing to do. Then 18 months later VMS goes into maintenance mode.  --
 Alan Greig   ------------------------------  % Date: Thu, 25 Jan 2001 11:26:45 -0500s# From: Jim Agnew <agnew@hsc.vcu.edu>g Subject: Backup Problem...+ Message-ID: <3A7053C5.AADF1166@hsc.vcu.edu>    Dear all netbeings,    Hello and warm fuzzies and whatnot.  I've been seeing this error off and on for a long time, and finally am getting tired of not knowing what it's about exactly.  Since this is in a daily job, and our files turn over slowly, I've not really worried about this.j   Can anyone enlighten me as to what backup-w-fidnotfnd really means? it's not mentioned in the system error message util except forP "contact support", basically, and we don't have support.  oh, yeah, vms 5.5-2...  	 Jim Agnew,    w %BACKUP-E-OPENIN, error opening ANACIN$DUA2:[000000]PD_FREE_FILE.DAT;1 as input           -SYSTEM-W-NOSUCHFILE, no such0` file                                                %BACKUP-I-STARTVERIFY, starting verification) pass                                     E^ %BACKUP-E-VERIFYERR, verification error for block 3 of ANACIN$DUA2:[000000]000000.DIR;1       S %BACKUP-W-FIDNOTFND, internal error. File to be processed not found on process lists  ^ %BACKUP-I-STARTRECORD, starting backup date recording pass                                     $       SETUP ZIPr   ------------------------------    Date: 25 Jan 2001 18:23:24 +0100G From: Jan Vorbrueggen <jan@mailhost.neuroinformatik.ruhr-uni-bochum.de>i Subject: Re: Backup Problem...H Message-ID: <y4ofwvim0j.fsf@mailhost.neuroinformatik.ruhr-uni-bochum.de>  N This is (was) a bug in BACKUP. It might be fixed, or maybe not, but BACKUP has* in the mean time been totally re-written.   M There also was a bug in that time frame where the file system's idea of how aeL file id looked like and BACKUP's idea of the same were different...so BACKUPK used some bits of the file id that were actually valid as indicators that a5* /RECORD or a /VERIFY had been requested...   	Jan   ------------------------------   Date: 25 Jan 2001 17:05:12 GMT3 From: gartmann@immunbio.mpg.de (Christoph Gartmann)c Subject: Re: Backup Problem...0 Message-ID: <94pmc8$6gm$1@n.ruf.uni-freiburg.de>  Q In article <3A7053C5.AADF1166@hsc.vcu.edu>, Jim Agnew <agnew@hsc.vcu.edu> writes:u >Dear all netbeings,   >v >Hello and warm fuzzies and whatnot.  I've been seeing this error off and on for a long time, and finally am getting tired of noti >knowing what it's about exactly.  Since this is in a daily job, and our files turn over slowly, I've not really worried about this. >f >Can anyone enlighten me as to what backup-w-fidnotfnd really means? it's not mentioned in the system error message util except forrQ >"contact support", basically, and we don't have support.  oh, yeah, vms 5.5-2...- >-
 >Jim Agnew >h >2x >%BACKUP-E-OPENIN, error opening ANACIN$DUA2:[000000]PD_FREE_FILE.DAT;1 as input           -SYSTEM-W-NOSUCHFILE, no sucha >file                                                %BACKUP-I-STARTVERIFY, starting verification * >pass                                     _ >%BACKUP-E-VERIFYERR, verification error for block 3 of ANACIN$DUA2:[000000]000000.DIR;1       ST >%BACKUP-W-FIDNOTFND, internal error. File to be processed not found on process list >W_ >%BACKUP-I-STARTRECORD, starting backup date recording pass                                      >$       SETUP ZIP  N Quite simple: during the backup process Backup creates a list of all the filesO in order to know what to compare for the verification pass afterwards. It mightoN happen that a file has been backuped but has then be deleted afterwards and soM it is just not present during the verification pass. Wether you have to worrytG about or not depends on the file. If it is SYSUAF.DAT I would be really A concerned, if it is POP3_SERVER.LOG I am not astonished at all...a   Regards,    Christoph Gartmanng  H -- --------------------------------------------------------------------+H | Max-Planck-Institut fuer      Phone   : +49-761-5108-464   Fax: -452 |H | Immunbiologie                                                        |H | Postfach 1169                 Internet: gartmann@immunbio.mpg.de     |H | D-79011  Freiburg, FRG                                               |H +--------- http://www.immunbio.mpg.de/home/english/menue.html ---------+   ------------------------------  % Date: Thu, 25 Jan 2001 23:23:48 +1030k% From: Jeremy Begg <jeremy@vsm.com.au>e! Subject: Re: Compaq press release1* Message-ID: <3A7021DC.A34EA0E5@vsm.com.au>   Antony Wardle wrote: > 2 > You don't supposed that this is behind the delay/ > on my Alpha server ds20e with san disk array?  > . > My configuration seems to be the same as the2 > sydney stock exchange, and ours has been delayed > for a varity of reasons.  D In the past six months two of my customers have ordered DS20Es via aH long-established Australian reseller.  Both have taken about 10 weeks toE arrive from date of order.  It seems that Compaq is having productioniL difficulties on this model of system.  IIRC in both cases the long delay was3 due to the 4-slot card cage which you have to have.   K It's likely that I'll be involved in the purchase of a third and possibly ahJ fourth DS20E in the near future .... it will be interesting to see if they arrive any quicker.h   Regards,   	Jeremy Begg 	VSM Software Services Pty Ltd.    ------------------------------  % Date: Thu, 25 Jan 2001 02:40:15 -0500 - From: JF Mezei <jfmezei.spamnot@videotron.ca>o= Subject: Re: Compaq: A simple, affordable clustering solutions, Message-ID: <3A6FD85E.8B87DEA4@videotron.ca>   jlsue wrote:E > Nobody ever said that VMSclusters are always problem free, so this,bB > again, doesn't prove anything.  Otherwise, by your reconing, SunG > 10000s are nothing but crap, heck we can point to lots of comments in  > here that say so.   N Just because Sun is inferior to VMS doesn't mean that Sun is nothing but crap.1  Sun hardware and software does run the internet.o  K System availability has more to do with your staff's expertise than it does-L the technology. If you know the limitations of your platform, you can designM something robust enough around those limitations. If you are a serious systemDM manager with a 7*24 mentality, you'll know not to make changes during workingtN hours, and that any change that is made must be documented and have a rollback mechanism if a problem come up.   L A local GSM provider here in Canada decided to switch their web interface toL send SMSs and to subscribe to all the various options. They sent an email toN the phones announcing that the email to phones as well as their web site would, be down for 3 days. (yes, you read right !).  L These guys upgraded their site from a mickey mouse NT - IIS system to a realJ Sun Solaris with Apache. The software may have been upgraded, but I have a' feeling that the windows geeks haven't. G (so much for the reputation that VMS runs most mobile phone companies).     N While VMS may have great clustering features, I would say that most go unsusedM by most sites. And the stability of a cluster is more dependant on the peoplen who run and configure it.s    M Where VMS is different from the rest of the crowd is that the people who have G remained in VMS are those with a certain mentality in which quality andrL robustness are important. It isn't so much the platform as it is the people.  L With True64 having "the world'd best clustering" (according to Compaq), whatI is left in VMS is the one important word: QUALITY.  This is the one pointyH where VMS can shine and others can't attach because VMS is significantlyA different in that area (documentation, solidity of software etc).m   ------------------------------    Date: 25 Jan 2001 10:37:23 +0100G From: Jan Vorbrueggen <jan@mailhost.neuroinformatik.ruhr-uni-bochum.de> = Subject: Re: Compaq: A simple, affordable clustering solution H Message-ID: <y48zo07z1o.fsf@mailhost.neuroinformatik.ruhr-uni-bochum.de>  / JF Mezei <jfmezei.spamnot@videotron.ca> writes:a  3 >  Sun hardware and software does run the internet.(  N It does? I'd have thought Cisco and Nortel and ... hard- and software trun the
 internet.    	Jan   ------------------------------  % Date: Thu, 25 Jan 2001 10:38:10 +0100a= From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>t= Subject: Re: Compaq: A simple, affordable clustering solutiont) Message-ID: <3A6FF401.4DD4B062@gtech.com>N   JF Mezei wrote:eP > Just because Sun is inferior to VMS doesn't mean that Sun is nothing but crap.3 >  Sun hardware and software does run the internet.   	 Do they ?p  B The last time statistics I saw showed Compaq as most used hardware platform for web-servers.  G (unfortunatetly it is not high-end Tru64 and VMS servers, but all thosem* cheap NT/2000 and Linux x86 low-end boxes)     Arne   ------------------------------    Date: 25 Jan 2001 08:48:04 -0500* From: young_r@eisner.decus.org (Rob Young)= Subject: Re: Compaq: A simple, affordable clustering solution:+ Message-ID: <qWtVssjixP0c@eisner.decus.org>a  [ In article <3vbv6t0tf07630g59g5lf9av57qfdjidvk@4ax.com>, jlsue <jlsuexxxz@home.com> writes:k5 > On Wed, 24 Jan 2001 12:45:08 +0000, andrew harrisonF# > <andrew.nospam@uk.sun.com> wrote:>   >>B >>> No, now you are twisting things again.  You have no ability toI >>> comprehend that all apps and all user populations don't behave as youaE >>> think.  I never said that *all* will, but I did say that the vastcA >>> majority, and imho this is over 90% of the applications, willtB >>> definitely operate as I have described.  And I don't make thisC >>> statement lightly, either.  I have seen many, many applicationstH >>> (probably at least 1500) and have only seen rare instances of things9 >>> that don't work as well as we'd like in a VMScluster.' >>>  >>; >>Strange again that you could not get that "vast majority" ? >>concensus from other admins who also manage OpenVMS clusters.e >   9 	Andrew , we don't enter in to debating each and everyonel; 	of your tangential FUDness.  The applications I am workingf- 	with are cluster aware and are in a cluster."   			Rob   ------------------------------    Date: 25 Jan 2001 08:50:33 -0500* From: young_r@eisner.decus.org (Rob Young)= Subject: Re: Compaq: A simple, affordable clustering solutionc+ Message-ID: <aFtAlNBBVvJM@eisner.decus.org>f  \ In article <3A6FD85E.8B87DEA4@videotron.ca>, JF Mezei <jfmezei.spamnot@videotron.ca> writes: > jlsue wrote:  I > (so much for the reputation that VMS runs most mobile phone companies).t   	Billing.  Think about it.   				RobN   ------------------------------    Date: 25 Jan 2001 09:11:33 -0500, From: koehler@eisner.decus.org (Bob Koehler)= Subject: Re: Compaq: A simple, affordable clustering solutionh+ Message-ID: <IO2xVNqDqiss@eisner.decus.org>   i In article <3A6FF401.4DD4B062@gtech.com>, Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com> writes:  > JF Mezei wrote:gQ >> Just because Sun is inferior to VMS doesn't mean that Sun is nothing but crap.b4 >>  Sun hardware and software does run the internet. >  > Do they ?     ? They must be.  Otherwize why would we have a local Sun resellertC advertizing that Sun is "the . in .com for the federal government".r  ? Personally I'm glad my federal government systems are in .gov .h  D > The last time statistics I saw showed Compaq as most used hardware
 > platform > for web-servers. >     F ----------------------------------------------------------------------? Bob Koehler                     | Computer Sciences Corporationi= NASA GSFC Flight Software       | Federal Sector, Civil GroupeE                                 | please remove ".aspm" when replyinge   ------------------------------    Date: 26 Jan 2001 00:53:08 +0800, From: Paul Repacholi <prep@prep.synonet.com>= Subject: Re: Compaq: A simple, affordable clustering solutiony- Message-ID: <8766j360az.fsf@prep.synonet.com>v  / JF Mezei <jfmezei.spamnot@videotron.ca> writes:   M > System availability has more to do with your staff's expertise than it doesrN > the technology. If you know the limitations of your platform, you can designO > something robust enough around those limitations. If you are a serious systemzO > manager with a 7*24 mentality, you'll know not to make changes during workingaP > hours, and that any change that is made must be documented and have a rollback! > mechanism if a problem come up.m  F So when is 'after hours' in a 24x7 shop? If it IS 24x7 and you are notE willing to due stuff during the day, when vendor support starf are ataI work, offices are open etc, then when do you do it? If you are unwilling,l9 then it seems to me, that YOU consider the system flawed.e  N > A local GSM provider here in Canada decided to switch their web interface toN > send SMSs and to subscribe to all the various options. They sent an email toP > the phones announcing that the email to phones as well as their web site would. > be down for 3 days. (yes, you read right !). > N > These guys upgraded their site from a mickey mouse NT - IIS system to a realL > Sun Solaris with Apache. The software may have been upgraded, but I have a) > feeling that the windows geeks haven't.rI > (so much for the reputation that VMS runs most mobile phone companies).e  ! It was the billing systems there.a  D I can add another bit. I know of a SMS co. They have 40 odd NT boxesC held together by a ton of sweat. Why not VMS? Because they can hire-B PC for $$/month with gateshit. If they scale back, send them back.; They don't have to worry about next month, only this month.p  F If they could, VMS and RDB or orible would be in there faster than youK could blink. But, it's all about money, how much money can you spend today.c! Tomorrows millions are of no use.s  O > Where VMS is different from the rest of the crowd is that the people who haverI > remained in VMS are those with a certain mentality in which quality and N > robustness are important. It isn't so much the platform as it is the people.    N > With True64 having "the world'd best clustering" (according to Compaq), whatK > is left in VMS is the one important word: QUALITY.  This is the one pointuJ > where VMS can shine and others can't attach because VMS is significantlyC > different in that area (documentation, solidity of software etc).e   ROTFL   . The biggest plus of VMS is that it is so EASY.   -- t< Paul Repacholi                               1 Crescent Rd.,7 +61 (08) 9257-1001                           Kalamunda. @                                              West Australia 6076. Raw, Cooked or Well-done, it's all half baked.   ------------------------------  % Date: Thu, 25 Jan 2001 18:18:22 +0000 - From: Tim Llewellyn <tim.llewellyn@bbc.co.uk>e$ Subject: Damned TCP/IP Services rant) Message-ID: <3A706DEE.F85F61A8@bbc.co.uk>w  G Damned TCP/IP Services, been trying to update my DNS config for quite a  while B with no sucess. The TCPIP SET NAME/noserver=xxx.yyy.zzz.aaa/SYSTEM commandsB have not been throwing errors but also the modifications have been ignored. Just now B I went back to DCL and "set proc/priv=all" and now it all works as	 expected.n  B Now, just why does TCPIP Services not generate a insufficient priv messageGG like all good VMS friendly apps should? I'm sure I've been in this loops. before, just don't reconfigure DNS very often.  ? Also, just what is wrong with TCPIP SET NAME/NOSERVER=*/SYSTEM?  HELP advises not to use it.v  ? DIGITAL TCP/IP Services for OpenVMS Alpha Version V5.0A - ECO 1i0   on a AlphaStation 255/233 running OpenVMS V7.1   Sorry for rant.    --6 Tim Llewellyn, OpenVMS Infrastructure, Remarcs Project0 MedAS at the BBC, Whiteladies Road, Bristol, UK.A Email tim.llewellyn@bbc.co.uk. Home tim.llewellyn@cableinet.co.uke  A I speak for myself only and my views in no way represent those ofo MedAS or the BBC.    ------------------------------  % Date: Thu, 25 Jan 2001 11:16:43 -0500s2 From: rdeininger@mindspring.com (Robert Deininger)" Subject: Re: dec 3000 300 crashingL Message-ID: <rdeininger-2501011116440001@user-2iveal5.dialup.mindspring.com>   In article <y4r91tj5ko.fsf@mailhost.neuroinformatik.ruhr-uni-bochum.de>, Jan Vorbrueggen <jan@mailhost.neuroinformatik.ruhr-uni-bochum.de> wrote:r  6 > rdeininger@mindspring.com (Robert Deininger) writes: > O > > Now I'm more confused than ever.  Perhaps we have two machines here sick ineO > > exactly the same way, but they never show symptoms, except this T MEM LLSC.r > L > That wouldn't be too surprising: these instructions are only required on aM > multi-processor system to make sure spinlocks et al. actually work. I doubtdP > that a normally-configured uniprocessor ever actually uses these instructions.   You may be right.  I don't have the details in my head.  That is certainly the main use.  Perhaps some smart DMA devices use the hardware locking as well.  E Let's take a poll.  How many folks with DEC 3000 300-ish systems passn >>> T MEM LLSC and how many fail the test?    -- f Robert Deininger rdeininger@mindspring.comt   ------------------------------   Date: 25 Jan 2001 18:24:44 GMT2 From: mathog@seqaxp.bio.caltech.edu (David Mathog)! Subject: fast_io_copy.c questions-, Message-ID: <94pr1c$o8o@gap.cco.caltech.edu>  J I'm trying to get up to speed (pardon the pun) with fast_io_copy.c which I9 obtained from a post to this group.  I compiled it with: R  = $ cc/arch=host fast_io_copy+sys$library:sys$lib_c.tlb/libraryI  & Question 1: I could not build it with:  Z   $ cc/arch=host/standard=ansi89/prefix=all fast_io_copy+sys$library:sys$lib_c.tlb/library  H apparently because it wasn't picking up iosadef.h.  Is it just me or is D there some trick to it?  (I want eventually to modify this slightly,F and prefer to work with code that builds cleanly in ANSI mode with all warnings enabled.) c  > Question 2:  in order to use the IO_PERFORM routines a processA must hold the indentifer VMS$BUFFER_OBJECT_USER, else the programe generates a:  J    %SYSTEM-E-NOBUFOBJID, requires rights identifier VMS$BUFFER_OBJECT_USER  J IO_PERFORM looks like a modern, general purpose, replacement for QIO.  If I so why do processes need special dispensation to use buffer objects?  No d/ such requirement exists for QIO or RMS buffers.d  C Question 3: in order to get fast_io_copy.c to run I had to crank upJI _both_ MAXBOBMEM and MAXBOBS0S1 to 5000 (well, something bigger than 1600mJ anyway) in order to avoid EXBUFOBJLM.  (Which wouldn't print anything whenE followed by HELP/MESSAGE).  Apparently this is because the program isAE hardwired to use buffers=16 and blocks=127.  Do the IO_* system callsDH interact with SET RMS/block/buffer or must these values be hardwired forF each IO_* based program?  Because sometimes you want a lot of buffers,I and other times not, and it would be nice not to have to add command line, switches for every program.l  J Question 4: If many programs run through IO_* system calls, and many usersC use those programs there are going to be an awful lot of EXBUFOBJLMrK messages unless the two MAXBOB parameters are set very much higher than theeF defaults. What's the downside of doing so?   For instance, if we allowL up to 50Mbytes to be used for this purpose, do those pages of memory become M inaccessible for other purposes when they are NOT actively in use as buffers?   K Question 5: (Looking ahead).  What sort of memory interactions are expected H between IO_* programs (and for that matter, QIO and RMS)  and XFC?  ThatP is, imagine a program that just writes data to a file.  To optimize IO now usingJ IO_* routines (or RMS or QIO) the amount of buffer memory (of one type or H another) must be increased, but that's memory that is coming OUT of thatF available for XFC.  I don't see a way to tell the OS to share buffers J between these subsystems in any useful manner, only ways to copy from one E memory buffer to another.  This questin is not theoretical in nature tH because I've already observed  a (negative) interaction between IO_* andE DECRam.  Example, copying a 16262 block text file (the same one as myh, benchmark programs use) gave these results:   6                  [---U2W disk-------]   [  RAMDISK   ]7                  seconds      Dev ops   seconds Dev ops-: fast_io_copy     .131         279       .017      420     5 regular copy     .410         517       .008      545-  H Which shows that fast_io_copy kicks butt on a physical disk, moving dataL roughly 3 times faster and requiring only 3/5 as many device operations, butJ going to RAMdisk it was 2X slower than copy, and required (roughly) 4/5 asJ many operations.  While the number of device operations was very close forI regular copy in the two cases (increasing by 5% going to RAMdisk) it went:K up by 50% for fast_io_copy.   Not that .017 seconds is bad for this sort ofsI work, but one might have expected .003 seconds, or at worst .008 seconds,tI and I simply cannot explain .017.  My point being that it would certainlyeH be nice to be able to write a program that runs as fast as possible bothI when it's written straight to disk, through a file cache, or to RAMdisk. iK But the observed speeds indicate that it is not going to be easy to do so. 0H COPY is the archetypical case - it should always run as fast as possibleI no matter what the output device as it is an extremely common program to t run.   Thanks,s   David Mathog mathog@seqaxp.bio.caltech.eduh? Manager, sequence analysis facility, biology division, Caltech a   ------------------------------  # Date: Thu, 25 Jan 2001 18:03:24 GMTp From: "Cor" <jc.smit@home.nl> , Subject: For Sale: 512 Mb memory DS20 (new!)3 Message-ID: <MTZb6.56126$Wa6.964076@zwoll1.home.nl>   ( For Sale: 512 Mb memory  for DS20 (NEW!)
 (4x 128Mb)   Give me your price.e   ------------------------------  # Date: Thu, 25 Jan 2001 17:58:15 GMTt From: "Cor" <jc.smit@home.nl>A, Subject: For Sale: Vaxstation 4000 model 90A3 Message-ID: <XOZb6.56116$Wa6.963935@zwoll1.home.nl>t  # For Sale: Vaxstation 4000 model 90A  64 Mb Memory 2x SCSI 1 Gb harddisk   	 Fl. 500,=t   ------------------------------  % Date: Thu, 25 Jan 2001 10:27:14 -0800t! From: Tom Linden <tom@kednos.com> 0 Subject: RE: For Sale: Vaxstation 4000 model 90A9 Message-ID: <EJEDILJIPPJOIEFOLDHLEENGCEAA.tom@kednos.com>t  : Well, if you were in California I would take you up on it.   > -----Original Message-----$ > From: Cor [mailto:jc.smit@home.nl]* > Sent: Thursday, January 25, 2001 9:58 AM > To: Info-VAX@Mvb.Saic.Com . > Subject: For Sale: Vaxstation 4000 model 90A >  > % > For Sale: Vaxstation 4000 model 90Aa > 64 Mb Memory > 2x SCSI 1 Gb harddiskt >  > Fl. 500,=d >  >    ------------------------------  % Date: Thu, 25 Jan 2001 09:28:32 -0800 ! From: Tom Linden <tom@kednos.com>c) Subject: ftp'ing savesets from W2K to VMSn9 Message-ID: <EJEDILJIPPJOIEFOLDHLIENDCEAA.tom@kednos.com>   H I think one needs to change block size.  Can someone give me the command: to access the saveset.  It was transferred in binary mode. tiad   ------------------------------  % Date: Thu, 25 Jan 2001 17:40:13 +0000 5 From: "Steeples, Oliver" <Oliver.Steeples@Compaq.com> - Subject: RE: ftp'ing savesets from W2K to VMS N Message-ID: <F498D199EDB12D468CD2C66680D3080116C3FD@reoexc04.emea.cpqcorp.net>   backup/list <saveset>/sav   # this should give correct block size   < set file/attributes=(lrl:<block size>, rfm:fix) filename.sav   	Olivery   -----Original Message-----( From: Tom Linden [mailto:tom@kednos.com]( Sent: Thursday, January 25, 2001 5:29 PM To: Info-VAX@Mvb.Saic.Com ) Subject: ftp'ing savesets from W2K to VMSv    H I think one needs to change block size.  Can someone give me the command: to access the saveset.  It was transferred in binary mode. tiat   ------------------------------  % Date: Thu, 25 Jan 2001 08:23:02 -0700o  From: Jon <jsmyth69@hotmail.com> Subject: Gnu-C for VMS?U2 Message-ID: <T0RwOhNbf+TnOTwlcGouYJrDdHpG@4ax.com>   Hi, C     Can anyone give me a pointer to Gnu-C for VMS?  (Alpha that is)(     Thanks!,   ------------------------------  % Date: Thu, 25 Jan 2001 11:10:44 -0500J" From: Dan Sugalski <dan@sidhe.org> Subject: Re: Gnu-C for VMS?i: Message-ID: <5.0.2.1.0.20010125110911.01b284d0@24.8.96.48>  % At 08:23 AM 1/25/01 -0700, Jon wrote:g >Hi,E >     Can anyone give me a pointer to Gnu-C for VMS?  (Alpha that is)i
 >     Thanks!o  H If you qualify for the hobbyist program, you'd be better off going with J Compaq C instead. It generates faster code than GCC, and the last version L of GCC I know of for the Alphas (2.8.0) has at least one (if not more) code J generation bugs. (You can't build perl on VMS with gcc 2.8.0, for example)   					Dan  I --------------------------------------"it's like this"--------------------2 Dan Sugalski                          even samurai? dan@sidhe.org                         have teddy bears and evens;                                       teddy bears get drunk    ------------------------------   Date: 25 Jan 2001 07:16:25 GMT- From: djweath@attglobal.net (Dave Weatherall)e? Subject: Re: Is Compaq dropping support for vms on vax platforms5 Message-ID: <DTiotGxQ0bj6-pn2-U1RT12jNt1kE@localhost>o  F On Tue, 23 Jan 2001 12:46:26, David B Sneddon <dbsneddon@bigpond.com>  wrote:  > > At 11:26 AM 1/23/01 +0000, steven.reece@quintiles.com wrote: >A > >Fabio wrote:l@ > >>>>So, if Compaq is dropping VMS on VAX, it is time to have a: > >VAX on a chip in  a PCI card running under Windows 2000 > >and CHARON-VAX.....<<<m > < > And just out of curiosity, how many PDPs are still in use?   We're down to one!   -- > Cheers - Dave.   ------------------------------  % Date: Wed, 24 Jan 2001 23:54:43 -0800p) From: Wayne Holland <wholland@tscnet.com>n- Subject: Re: Linux worm and RedHat 7.0 brokendO Message-ID: <50548AF250040795.4E551A979BE438D3.2DE2AD83450FC3F0@lp.airnews.net>e   sabolich@my-deja.com wrote:t > ; > > I guess linux is better than plain unixes in some ways.h > J > Well I cannot have (won't spend the money) for a SGI to try IRIX, rs6000% > for AIX, etc.  What's a plain Unix?e  ? Well, you can try Solaris 8 x86 ($75). But check their Hardwaree Compatibility ListD first.  I've got it, but I'm having a hell of time getting a locally attached printerF to work with it.  The printer works fine in text mode, but getting the filters to workbH correctly to get ghostscript to print postscript files is a hair yanking experience. E It does look a lot better than any Linux distros tho.  But I'd rather- have VMS if I " could afford it for personal uses.  H The topic was interesting to me comparing the two O/Ses.  I suspect that preferences E are a very large part of it tho.  I've had the formal training in VMSn and some of BSD.G But in a production environment I had to move fast to get a new program 	 developedtH to translate Genrads simulators' binary patterns to their own Test Basic	 language.BD I found it a little easier to just look up in the vms docs to see if there was already C a service available that I didn't have to write.  A software designcF facility wanted 6 months to do the job... I did it in three weeks with  two weeks testing and debugging.H But I had a very hothead of a boss breathing down my back waiting for it to get done.G He didn't understand why it should take so long.  Anyway, I found a lotoH of useful services available that I didn't have to write... just plug em in and use.n  H For others, they probably could have done fine in Unix using a different	 approach.l9 I think its a matter of mindset.  I've enjoyed the topic!t   >  > >gJ > > The VMS process/task model is somewhat different than unix's.  NeitherG > maps perfectly into the other, and trying to force them to has causeda > much gnashing of teeth.e >  > >  > > -- > > Robert Deininger > > rdeininger@mindspring.com- > >- >  > Sent via Deja.com  > http://www.deja.com/   ------------------------------    Date: 25 Jan 2001 10:29:51 +0100G From: Jan Vorbrueggen <jan@mailhost.neuroinformatik.ruhr-uni-bochum.de>m- Subject: Re: Linux worm and RedHat 7.0 broken-H Message-ID: <y4bssw7ze8.fsf@mailhost.neuroinformatik.ruhr-uni-bochum.de>   sabolich@my-deja.com writes:  D > Please see my example in reply to Christof Brass comparing fork(),4 > execl() unix style of process creation with VMS.    G That I considered a particularly bad example - for instance, because iteN spreads the actions of one operation in an obscure way over a number of systemA calls, enveloped in an if..then..else that depends on knowing andtJ understanding the semantics of fork(). You mighe be used to it and find itM clear and easy - when I first read about it, I found it absurd. Hey, first wesM go to the effort of updating all those page tables for the new process with amK copy of the previous one, copy-on-write and all, and then throw it all awayi: immediately when performing the execl()? Sure makes sense.  1 > By the way, does VMS have anything like fork()?u  / It had, doesn't have now, will have soon again.,  0 > > -- Simple structureless files like unix has.E > One could argue that this is enough.  If you try to do more you are J > trying to anticipate the needs of the programmer and may include a bunchD > of features he does not need and miss the feature he really needs.  L Well, I'd say in typical Unix fashion that 90% of the programmer's needs canJ be handled by a small set of features - such as those provided by RMS. TheN point is that they are available to everyone using VMS, so they don't all haveM to re-invent the wheel every time they want something more sophisticated thantC those streams-of-bytes. Have you ever checked how many similar but  N incompatible file formats and on-disk database structures exist in a "typical" Unix system?  L And, contrary to popular opinion, RMS is _not_ part of the operating or fileK system. It will gladly get out of the way if you want it to. It's just thatl8 most people don't bother because the see its davantages.  J > Yeah but then if you do use the 'big hammer' all these features sit idle > and just take up space.y  H Space where - memory, disk? Irrelevant today, was irrelevant two decadesI ago. If RMS doesn't provide it, you will have multiple versions provided  B by different applications you're using, "wasting" even more space.  G > There is some data acquisition software running there that makes data6H > available to PCs via FTP server.  Because of FTP the data is stored inI > simple files -- obviously the wrong way to do it on VMS.  The number ofhB > files in the directory climbs into the thousands (maybee tens ofG > thousands).  When it is time to purge the data (erase the files) this E > takes a very long time (hours).  Also it wastes a lot of disk spacelJ > (since the files are small) and causes fragmentation.  Obviously this isJ > the wrong way to do it on VMS.  They should rewrite the data acquisitionJ > to store using variable length records (an Indexed file under RMS).  Ok,F > now they cannot use the generic FTP server and would need to write a
 > custom one.   N These problems are now solved. Yes, it took way too long, but that's (at least& in part) what DEC's bad days gave you.  G > OK, but what can you accomplish with them that you cannot do in Unix?h  M Logical names have an adaptable name-space hierarchy, and they have structuresG (e.g., in search lists). That means you can do a lot of things that areo; impossible with either environment variables or soft links.d  J > Well these faults you mentioned are a problem with that particular Unix,E > not the device=file model.  In addition to open() close(), etc. youu > forgot ioctl().e  N Once you use ioctl(), you're at the same level as VMS driving its devices. TheO only convenience is the consolidated name space Unix has, which makes searching L for things easier - that deficiency has been in part corrected. And tell me,M are all implementations of ioctl() consistent and portable? Don't think so...,  G > You asserted that you need to change kernel source code and recompiled7 > the kernel for "simple" changes in OS configuration. l  K You can install Linux without having exactly the right version of the righteI compiler? Not just boot off the CD in a minimal mode, but have the systeme0 configured for your particular system and needs?  H It took SunOS/Solaris about 15 years or so to get there. It was a designJ mandate for VMS from day 1, because people were fed up with SYSGENs on the various PDP-11 OSes.   	Jan   ------------------------------  % Date: Thu, 25 Jan 2001 10:39:28 +0100n= From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>s- Subject: Re: Linux worm and RedHat 7.0 brokeno) Message-ID: <3A6FF450.484B9C12@gtech.com>h   Wayne Holland wrote:G > It does look a lot better than any Linux distros tho.  But I'd ratheri > have VMS if I $ > could afford it for personal uses.  7 Considering that a VMS Hobbyist license is 0 $ then ...    Arne   ------------------------------    Date: 25 Jan 2001 19:52:43 +0800, From: Paul Repacholi <prep@prep.synonet.com>- Subject: Re: Linux worm and RedHat 7.0 brokent- Message-ID: <877l3jx304.fsf@prep.synonet.com>y   sabolich@my-deja.com writes:   > > or Galaxy,  I > Virtual machines.  Some systems have this in hardware - for exmaple IBM G > s/390 which can run over 1000 instances of Linux on a single machine. I > Also VMware for x86 which allow multiple OSes (NT, Linux, FreeBSD, etc)-& > to run simultaneously.  Also Plex86.  G Galaxy is NOT a virtual machine. It is a partitionable OS archetecture.c  F > > Is it good design that programs compiled and linked many years andI > versions ago will run just fine, unchanged, on current hardware and OS?d  H Misleading. It is not nessisarily a good idea. It is nice if you HAVE to# do it, but it is not a 'good thing'r  I > Yes, as long as you are on the same CPU type.  Can you run VAX binariesi > on Alpha?  > 6 >   Is there a unix hiding somewhere that can do this? >  > Yes, Linux, FreeBSD, etc.a  : Yes, and PDP-1... nova, IBM 14xx, PDP-11, and soon, PDP-10H Called emulation, what's the point. IBM has been at it for 1/2 a century nearly.n   -- N< Paul Repacholi                               1 Crescent Rd.,7 +61 (08) 9257-1001                           Kalamunda. @                                              West Australia 6076. Raw, Cooked or Well-done, it's all half baked.   ------------------------------    Date: 25 Jan 2001 20:01:27 +0800, From: Paul Repacholi <prep@prep.synonet.com>- Subject: Re: Linux worm and RedHat 7.0 brokenb- Message-ID: <873de7x2lk.fsf@prep.synonet.com>o  3 bill@triangle.cs.uofs.edu (Bill Gunshannon) writes:a  N > In article <rdeininger-2401010135070001@user-2iveaup.dialup.mindspring.com>,7 >  rdeininger@mindspring.com (Robert Deininger) writes:i > |> O > |> (C > |> Another one:  Why are null-terminated strings a good design?  e > B > Who said Unix invented them??  Have you ever looked at .ASCIZ inI > Macro-11??  Or are you saying that DEC stole this from Bell Labs??  :-) < Ah they are in the 6 Macro Bill, and that predates and unix.   ...   G > |> And if null-terminated strings are so great, why does so much unixnK > |> code seem so reluctant to validate string arguments before using them?n > J > Can't address this as I don't see what the one has to do with the other. > G > |> Was this due to "planning, debate, and design" or just sloppiness?   E It is a form of 'in band signalling' the phreakers taught AT&T why its is a bad idea.   > F > My guess would be sloppiness.  And we all now there has never been a, > sloppy programmer working on a VMS system.    :)f  G > |> Is it good design that programs compiled and linked many years andnL > |> versions ago will run just fine, unchanged, on current hardware and OS? > G > And what OS might this be??  Surely your not talking about VMS??  The H > hardware changes every few years making this impossible even for that.G > Depending ont he triviality of the program, the same is true for mostnI > Unix versions as well.  Add into that the cross-Unix compatability thatiH > has been built into most of them today.  FreeBSD running SCO binaries,G > Linux running Solaris binaries.  I could probably test it to be sure,oD > but I would bet that an Ultrix-32 3.0 binary (without any builtin F > kernel dependancies) would run just fine on Ultrix-32 4.3.  I know ID > never had to re-compile any of my locally generated programs after > upgrading the OS.u  # Does that include shared libraries?   7 > |> Is there a unix hiding somewhere that can do this?  > 6 > Just about any of them that I have ever worked with.  7 Try none, unless you restrict it to static linked only..  F Not, that it is really *unix*. To the kernel, it just a buch of bytes,G if it want's to do anything, that's its problem. Great for prototyping,i0 not good for production and long term stability.   -- i< Paul Repacholi                               1 Crescent Rd.,7 +61 (08) 9257-1001                           Kalamunda.a@                                              West Australia 6076. Raw, Cooked or Well-done, it's all half baked.   ------------------------------  % Date: Thu, 25 Jan 2001 10:52:05 -0500h2 From: rdeininger@mindspring.com (Robert Deininger)- Subject: Re: Linux worm and RedHat 7.0 brokennL Message-ID: <rdeininger-2501011052060001@user-2iveal5.dialup.mindspring.com>  E In article <94ocf5$k4l$1@nnrp1.deja.com>, sabolich@my-deja.com wrote:   D > Please see my example in reply to Christof Brass comparing fork(),H > execl() unix style of process creation with VMS.  By the way, does VMS > have anything like fork()?   I posted a VMS example, but it may have been eaten by my news server, which was getting sick at that moment.  I'll repost it if it doesn't appear on the server soon.n  sI > Unix way is often to have a complex job be a set of simple jobs.  So if G > unix approach makes it easier for simple jobs it also makes it easieruB > for a collection of simple jobs which do something more complex.  } Sometimes.  Some complex jobs aren't easily expressed as a combination of simple ones, and extra layers of effort are needed.-    J > > It is a file system with a very rich set of features.  While my littleH > list isn't pure "RMS" (the boundaries are a little blury), I'll rattle > off a few features:m0 > > -- Simple structureless files like unix has. > E > One could argue that this is enough.  If you try to do more you are-J > trying to anticipate the needs of the programmer and may include a bunchD > of features he does not need and miss the feature he really needs.   Well, indexed files are a pretty standard concept, and I have been told that there are various add-ons to support them in many unixes.  Certainly programs commonly need to know how big a chunk of data is, an getting that info in advance can be nice.     1 > > -- A variety of record-oriented file formats. J > > -- Indexed files, with a rich set of index and key features, and toolsM > >    to manage the files even when they are quite large and complex.  Often,L > >    used as a poor man's database, removing the need for the "big hammer" > >    databases.W > J > Yeah but then if you do use the 'big hammer' all these features sit idle > and just take up space.   They aren't idle in VMS.  Most seem to have been added first when they were useful for writing VMS itself, or for supporting stuff from older platforms. The record system is extensible and new stuff has been added from time to time.  In practice, most of RMS seems to be used.  Of course, one can never predict what will turn out to be the best features, and some stuff in RMS might be marginal in practice.i   Another point of comparison might be the resource fork in Mac files.  Much more elaborate than RMS, but another example of providing commonly-need parts and making them easy for programmers to use.0    " > > -- Multiple versions of files. > B > Agree here - I like this.  Doh, didn't mean to delete that file!  c I actually kind of surprised none of the newer file systems have added this to unix.  Or have they?m    L > > -- Comprehensive support for stuff like file name parsing and searching,K > >    built into the OS, available to programmers at many levels, and used  > I > This is kind of general and I don't know exactly the scope of what thisr/ > encompasses.  Give some examples if you wish.,   Well, file specs in VMS are richer than in unix, and I realize that it is fairly common for unix programmers to do it by hand.  A full VMS file spec looks something like this:1G   node"username password"::device:[dir1.dir2.dir3]filename.type;version = (Any of this can be omitted to take the defaults, of course.) This is more than trivial syntax changes from unix.  There's a lot of semantics.  There are 3 kinds of wildcards, which are a little more complex than in unix.  Several parts can be logical names, which sometimes act like wildcards.  The standard parser allows up to two levels of component-by-component defaulting.  And there are little backward compatibility things like allowing <> instead of [], and period instead of semicolon.  And I'm omitting many details.   Virtually everyone who tries to do this parsing by hand gets at least some of it wrong, and ends up with an application that doesn't work for some perfectly valid file specs.  If you let the system do the work, it behaves "right" for anything the user throws at it.  You will say this is a terribly complex file naming system, and it is if you use it all.  But it was all built up over the years to provide real-world functionality.  As features were added, the tools got smarter without breaking compatibility,   E > Let me give an example here.  In a place I work at often there is a J > Alpha cluster running VMS.  I talk with the VMS guy often (he has a good, > amount of experience) sometimes about VMS. > G > There is some data acquisition software running there that makes data H > available to PCs via FTP server.  Because of FTP the data is stored in: > simple files -- obviously the wrong way to do it on VMS.   I wouldn't say so.  If it's simple data, simple files are fine, particularly for cross-platform work.  I would probably do it the same way.m   >  The number ofB > files in the directory climbs into the thousands (maybee tens ofG > thousands).  When it is time to purge the data (erase the files) this0E > takes a very long time (hours).  Also it wastes a lot of disk space 7 > (since the files are small) and causes fragmentation.t  FBoth of these are abuse of VMS.  If he's storing many small files, he should set up his disk to accomodate them.  Or use an up-to-date version of VMS, which allows tiny file chunks even on very large disks.  The old way _does_ become a problem on new, big disks.  The Mac file system was recently upgraded for the same reason.   VMS directories are optimized for fast lookup and reliability, and wholesale deletes or renames don't play happy when there are many files in one directory.VMS makes the directory self-consistent after each file is deleted, and the overhead is high with lots of files.  I've seen the same problem on other systems.  A system can be written with other goals in mind, and pay the cost in time, space or reliability.n  VMS has improved directory caching lately to help with such problems.  More improvements are in the pipeline, IIRC.  In the meantime, VMS folks usually distribute their files among more directories. Logical names make this transparent in many cases, and the intelligent parsing helps too.  Soemone here can probably suggest improvements, if you make the problem a little more specific.o   For existing huge directories, suggest to your guy that he delete the files in reverse order, or use one of the freeware tools that delete many files without promising moment-by-moment integrity of the directory.   > Obviously this is.J > the wrong way to do it on VMS.  They should rewrite the data acquisitionE > to store using variable length records (an Indexed file under RMS).,  ) Not based on anything you've said so far.u   > Ok,nF > now they cannot use the generic FTP server and would need to write a
 > custom one.   K No, there's nothing wrong with his files, so the existing ftp will suffice.-  -C > So in essence we don't need anything more than simple files.  Thei0 > filesystem itself is our 'poor man's database'  } I agree in this case.  But you are doing unix-think in your directory scheme, and making things difficult for no good reason.-  I > > And a lot more.  Symbols are a close match for environment variables.iF >  Soft links are rather different beasts than logical names.  LogicalG > names have recently gone cluster-wide as well.  Maybe you should looke= > them up in the docs, and also look at some examples of use.r > G > OK, but what can you accomplish with them that you cannot do in Unix?t   Probably nothing that's impossible in unix.  But they combine nicely with other parts of the OS, and the standard tools that everything is built around leverage the power of logical names.  A small example is the way a logical name can point to more than one place at once.  And they are recursive. If I set up such a name for whatever reason, the parse and search services mean that everything in the system understands what I mean.  It's hard to explain except with a lot of examples.  Kind of like pipes.  You  J > Well these faults you mentioned are a problem with that particular Unix,E > not the device=file model.  In addition to open() close(), etc. youiI > forgot ioctl().  For things like serial ports you use it for baud rate,lI > parity, etc (one file per port).  For a cdrom for eject, etc.  For taper@ > rewind, etc.  For sound card - sampling rate, bits per sample,H > mono/stereo, etc.  There is no need to have multiple files per device.  Once you add ioctl(), you have much of the complexity of VMS devices.  There is still a trivial issue of how they are presented to the user or system manager. (As a list of files, or as a separate list of devices?)  I expect good practice in unix should lead to proper functionality for devices, but I've seen many examples of crippling.  It _seems_ to result from programmers looking for simple shortcuts and trying to fit the device into the wrong cubbyhole.  K I guess I've been deprived of examples of well-done device support in unix.t  E > "looking out the window at coeds". Please, this is not a flame war.y" > Let's keep it from becoming one.   Ok.  I just have trouble coming up with a mental image of those unix guys designing.  I was only guessing about possible alternatives.   I'll put in a few extra smileys just in case. :-) :-) :-)  wG > You asserted that you need to change kernel source code and recompileeJ > the kernel for "simple" changes in OS configuration.  Do you still stand
 > by this?  v I'd don't.  The guys down the hall do, frequently.  Maybe they are doing it wrong, but they claim to be linux experts.   I do claim that there is seldom need to change VMS kernel code.  Yes, there are things that I'd like to change or add, but they are design modifications or new features, not tweeking the size of some data structure.      C > What your telling me is that you can run the same binaries acrosseH > different VMS versions and VAXes provided that the CPU instruction set > is the same.   Perhaps a bit more than that.  The VAX instruction set did change over the years, of course.  Some instructions were moved from hardware to emulation, and new stuff was added.  Old, single-cpu code will work on SMP systems, or on vector systems.i  But the OS changed much more than the hardware in many ways. It was done is such a way that user-mode code would not break, and often got the new features for free, without compiling or linking.  With rare exceptions in user mode, and this was almost never true of inner-mode code.  I think this goes back to the better OS support for common programming problems.  That support grew underneath the compiled programs as needed.  I > Same for Linux -- same binary will run on kernel 2.0.x, 2.2.x, 2.4.x onp > 386, 486, 586 686.  o I believe linux is ahead of most unixes here, which is not uncommon.  But what about linux 1.whatever binaries?n   -- c Robert Deininger rdeininger@mindspring.com.   ------------------------------    Date: 26 Jan 2001 00:15:05 +0800, From: Paul Repacholi <prep@prep.synonet.com>- Subject: Re: Linux worm and RedHat 7.0 brokeni- Message-ID: <87itn3622e.fsf@prep.synonet.com>i   sabolich@my-deja.com writes:  I > Same for Linux -- same binary will run on kernel 2.0.x, 2.2.x, 2.4.x onl > 386, 486, 586 686.  G You have missed 0.x and 1.x. And the answer is NO! First due to the ELF 0 change, then libc. Or was that 2 libc breakages?   -- t< Paul Repacholi                               1 Crescent Rd.,7 +61 (08) 9257-1001                           Kalamunda.0@                                              West Australia 6076. Raw, Cooked or Well-done, it's all half baked.   ------------------------------  # Date: Thu, 25 Jan 2001 15:59:34 GMTt" From: fooguy <jweisen@my-deja.com>3 Subject: Midrange I/O with VMS WAS: DS20 - Slow I/O ) Message-ID: <94pigv$io6$1@nnrp1.deja.com>d  E So I think we've all come to the same conclusion - this system sucks.-  E I'm looking around at other midrange options, and I don't see any. ItoF looks like my options are the KZPAC, and HSZ80. So what I'm hearing isH that Compaq and OpenVMS only support two backplane RAID controllers: one2 that is 5 years old and the other that is $25,000.  G I understand the industry trend is moving to detached storage, but here G I am, an admin of a VMS based financial system with 16Gigs of data, and E I can't buy a decent performing Ultra Wide 64bit RAID controller thateC can do RAID 5 with (3) 18gig disks? How do I justify a $25,000 RAIDiE controller for a system with 20 users? I can get a DS10 with a KZPCC,lF which would suit my needs perfectly, but what I'm seeing is that there0 is no support for VMS, and none planned for 7.3.  D If some could write back and say that I can get the KZPCC and a DS10D working under OpenVMS with a 4224 enclosure, I'd be the happiest man? alive. (C'mon people - today is my birthday. Help me out here.)a  F In any event, if anyone can offer their I/O success story (preferrablyG that doesn't involve the KZPAC) I'd appreciate. That could even includeuG performance of volume shadowing with faster, non-raid SCSI controllers.    Thanks,l John    ) In article <949rps$mcj$1@nnrp1.deja.com>,d%   fooguy <jweisen@my-deja.com> wrote: G > I know that the KZPAC is not supposed to be the shining stars of diskuE > performance, but we just purchased two DS20s, and I have to say I'm-H > really disappointed in the performance of the disk subsystem. I've runB > the gambit of configurations, and I have yet to get the DS20s to performb > faster than my desktop PC. >pD > My test was copying several large files (Oracle dumps - each about3 > 200MB) from one directory on the disk to another.v >p > DS20 >d > Number of Disks: (4) 9GB" > RAID Config: RAID 10 - 2 Channel > Cluster Size: 35 > MB/Sec: 1.13 >e > Number of Disks: (4) 9GB" > RAID Config: RAID 10 - 2 Channel > Cluster Size: 1t > MB/Sec: 1.48 >r > Number of Disks: (4) 9GB" > RAID Config: RAID 10 - 1 Channel > Cluster Size: 1  > MB/Sec: 1.20 >e > Number of Disks: (1) 9GB$ > RAID Config: SLED/JBOD - 1 Channel > Cluster Size: 1r@ > MB/Sec: Tested all four disks. Results range from 1.07 to 1.37 >s+ > Number of Disks: (1) 3GB Solid State Diskr- > RAID Config: N/A - attached to Symbios Cardo > Cluster Size: 1e > MB/Sec: 13.21hF > (Not Bad, but for a $30k hard drive it should be better. Compaq says itF > has a maximum sustainable transfer rate of ~60MB/Sec, I'd settle for 20 > in this test)d >e > Dell PowerEdge 4300u >b" > Number of Disks: (2) 9GB 7200RPM! > RAID Config: RAID 1 - 1 Channele > Cluster Size: N/As > MB/Sec: 8.38 >> > Dell Optiplex GX1p >r > Number of Disks: (1) 10GB IDEe > RAID Config: N/A > Cluster Size: N/A  > MB/Sec: 1.82 > H > Like I said, I know the KZPAC isn't the universe's answer to Fast I/O,G > but is there any sort of tuning I can do to improve performance? I'vehB > already played with cluster size, moving disks from the internal cabinethG > to the BA356, and cache type (write back vs. write through) with onlys anH > tiny difference. I'm not expecting a while lot, but is this the best I* > can expect? I'd be happy with 3-8MB/Sec. >  > Thanks in advance. >l > --/ > *********************************************x* > "All I every wanted from life was to see. > Larry Wall give Bill Gates a Perl Necklace." >. > Sent via Deja.com  > http://www.deja.com/ >    --- *********************************************u( "All I every wanted from life was to see, Larry Wall give Bill Gates a Perl Necklace."     Sent via Deja.comt http://www.deja.com/   ------------------------------    Date: 25 Jan 2001 17:40:51 +0100G From: Jan Vorbrueggen <jan@mailhost.neuroinformatik.ruhr-uni-bochum.de> 7 Subject: Re: Midrange I/O with VMS WAS: DS20 - Slow I/O H Message-ID: <y4snm77ffw.fsf@mailhost.neuroinformatik.ruhr-uni-bochum.de>  $ fooguy <jweisen@my-deja.com> writes:  G > I'm looking around at other midrange options, and I don't see any. It H > looks like my options are the KZPAC, and HSZ80. So what I'm hearing isJ > that Compaq and OpenVMS only support two backplane RAID controllers: one4 > that is 5 years old and the other that is $25,000.  3 So how about going to host-based shadowing instead?e   	Jan   ------------------------------    Date: 25 Jan 2001 12:08:03 -0500* From: young_r@eisner.decus.org (Rob Young)7 Subject: Re: Midrange I/O with VMS WAS: DS20 - Slow I/Oi+ Message-ID: <pLrhastjYbmz@eisner.decus.org>m  N In article <94pigv$io6$1@nnrp1.deja.com>, fooguy <jweisen@my-deja.com> writes:G > So I think we've all come to the same conclusion - this system sucks.  > G > I'm looking around at other midrange options, and I don't see any. ItaH > looks like my options are the KZPAC, and HSZ80. So what I'm hearing isJ > that Compaq and OpenVMS only support two backplane RAID controllers: one4 > that is 5 years old and the other that is $25,000. > I > I understand the industry trend is moving to detached storage, but here I > I am, an admin of a VMS based financial system with 16Gigs of data, andxG > I can't buy a decent performing Ultra Wide 64bit RAID controller thateE > can do RAID 5 with (3) 18gig disks? How do I justify a $25,000 RAIDWG > controller for a system with 20 users? I can get a DS10 with a KZPCC,yH > which would suit my needs perfectly, but what I'm seeing is that there2 > is no support for VMS, and none planned for 7.3. > F > If some could write back and say that I can get the KZPCC and a DS10F > working under OpenVMS with a 4224 enclosure, I'd be the happiest manA > alive. (C'mon people - today is my birthday. Help me out here.)r > H > In any event, if anyone can offer their I/O success story (preferrablyI > that doesn't involve the KZPAC) I'd appreciate. That could even includebI > performance of volume shadowing with faster, non-raid SCSI controllers.  > 	 > Thanks,o > John >     , 	Foo Guy... now we are getting some details.  A 	Now that we know the KZPAC does 2700 I/O per second and 20 MBytef? 	/sec bandwidth (max, bet it can't sustain that).  Tell us whatu> 	you need!  Do you *need* 10 MByte/sec sustained?  Do you need 	1000 I/O per sec sustained?  > 	Tell us how that solid state disk factors in, can you use it? 	How?e  ? 	Tell us how many drives you have, can you get more if need be? C 	(Used drives are pretty cheap, yes used.  With several RAID levels,? 	you can yawn when one takes a hit.  I have two bad ones on thei 	shelf behind me).  : 	You can pick up 4.3 Gig VW drives for $300, maybe pick up 	10 of them.  5 	How many open slots?  (Note: I'm unfamiliar with 230  	capacity).     @ 	You need I/O and bandwidth, RAID5 isn't the way to go.  I would> 	burn disks using Raid Software for OpenVMS (list price $1000,; 	you can do better than that!) and create 0+1 at the OS andp@ 	*maybe* use mirrors at the controller level (if possible... not> 	familiar with KZPAC, I do HSJ/HSZ).  (If you had 10 - 4.3 Gig= 	drives you could create 5 mirrorsets with the KZPAC and thenc= 	use RAID Software and create 5 stripes, do you have to breakd? 	shadowsets to quiece the database, or do you shut the app downt; 	and backup?).  With stripes across both your channels, yout 	should get 10+ MByte/sec.  < 	One gating factor... how heavy are your writes?  That KZPAC 	only has 4 Mbyte of cache.t  @ 	With your heavy I/O needs, I would definitely get RAID Software7 	for OpenVMS OR go to newer controller (HSZ) and createaB 	controller based 0+1.   But you are looking at a capital request.7 	You could always buy those 4.3 Gig drives 5 at a time,1 	if allowed.  ? 	Tell us more, we can do the best you can with what you have too 	work with.      				Robk   ------------------------------  # Date: Thu, 25 Jan 2001 17:42:39 GMTo" From: fooguy <jweisen@my-deja.com>7 Subject: Re: Midrange I/O with VMS WAS: DS20 - Slow I/O ) Message-ID: <94poic$p0v$1@nnrp1.deja.com>   G I'm not opposed to the idea, but with so many powerful inexpensive RAIDoB controllers out there, why would I opt for software mirroring overB hardware? Especially since we don't own Volume Shadowing Licenses,B and they're more expensive than a Mylex 64bit PCI RAID Controller.  G The other thing is that Volume Shadowing and bound volume sets give you @ RAID 0 and RAID 1, and any combo of the two, but not RAID 5. ForF something like a DB server (SQL/Oracle) I'm all about RAID 10, but for> an RMS based database and file sharing, my preference leans to& Distributed Data Guarding with Parity.    H In article <y4snm77ffw.fsf@mailhost.neuroinformatik.ruhr-uni-bochum.de>,C   Jan Vorbrueggen <jan@mailhost.neuroinformatik.ruhr-uni-bochum.de>h wrote:& > fooguy <jweisen@my-deja.com> writes: > F > > I'm looking around at other midrange options, and I don't see any. ItG > > looks like my options are the KZPAC, and HSZ80. So what I'm hearingd isH > > that Compaq and OpenVMS only support two backplane RAID controllers: onen6 > > that is 5 years old and the other that is $25,000. >s5 > So how about going to host-based shadowing instead?  >i > 	Jan >-   --- *********************************************n( "All I every wanted from life was to see, Larry Wall give Bill Gates a Perl Necklace."     Sent via Deja.com  http://www.deja.com/   ------------------------------  # Date: Thu, 25 Jan 2001 18:17:07 GMTa" From: fooguy <jweisen@my-deja.com>7 Subject: Re: Midrange I/O with VMS WAS: DS20 - Slow I/Oe) Message-ID: <94pqim$r1u$1@nnrp1.deja.com>   + In article <pLrhastjYbmz@eisner.decus.org>,a-   young_r@eisner.decus.org (Rob Young) wrote: H > In article <94pigv$io6$1@nnrp1.deja.com>, fooguy <jweisen@my-deja.com> writes:cB > > So I think we've all come to the same conclusion - this system sucks. > >sF > > I'm looking around at other midrange options, and I don't see any. ItG > > looks like my options are the KZPAC, and HSZ80. So what I'm hearing  isH > > that Compaq and OpenVMS only support two backplane RAID controllers: ones6 > > that is 5 years old and the other that is $25,000. > > F > > I understand the industry trend is moving to detached storage, but hereG > > I am, an admin of a VMS based financial system with 16Gigs of data,m andtD > > I can't buy a decent performing Ultra Wide 64bit RAID controller thatG > > can do RAID 5 with (3) 18gig disks? How do I justify a $25,000 RAIDgB > > controller for a system with 20 users? I can get a DS10 with a KZPCC,D > > which would suit my needs perfectly, but what I'm seeing is that therei4 > > is no support for VMS, and none planned for 7.3. > >sH > > If some could write back and say that I can get the KZPCC and a DS10H > > working under OpenVMS with a 4224 enclosure, I'd be the happiest manC > > alive. (C'mon people - today is my birthday. Help me out here.)  > >v= > > In any event, if anyone can offer their I/O success storyi (preferrablyC > > that doesn't involve the KZPAC) I'd appreciate. That could evenh include > > > performance of volume shadowing with faster, non-raid SCSI controllers. > >e > > Thanks,l > > John > >s >n. > 	Foo Guy... now we are getting some details. > C > 	Now that we know the KZPAC does 2700 I/O per second and 20 MByte A > 	/sec bandwidth (max, bet it can't sustain that).  Tell us whath@ > 	you need!  Do you *need* 10 MByte/sec sustained?  Do you need > 	1000 I/O per sec sustained? >i@ > 	Tell us how that solid state disk factors in, can you use it? > 	How?t >sA > 	Tell us how many drives you have, can you get more if need be?i> > 	(Used drives are pretty cheap, yes used.  With several RAID levelsA > 	you can yawn when one takes a hit.  I have two bad ones on thes > 	shelf behind me). >h< > 	You can pick up 4.3 Gig VW drives for $300, maybe pick up > 	10 of them. >t7 > 	How many open slots?  (Note: I'm unfamiliar with 230e
 > 	capacity).  >kB > 	You need I/O and bandwidth, RAID5 isn't the way to go.  I would@ > 	burn disks using Raid Software for OpenVMS (list price $1000,= > 	you can do better than that!) and create 0+1 at the OS andtB > 	*maybe* use mirrors at the controller level (if possible... not@ > 	familiar with KZPAC, I do HSJ/HSZ).  (If you had 10 - 4.3 Gig? > 	drives you could create 5 mirrorsets with the KZPAC and thene? > 	use RAID Software and create 5 stripes, do you have to breaknA > 	shadowsets to quiece the database, or do you shut the app downa= > 	and backup?).  With stripes across both your channels, you  > 	should get 10+ MByte/sec. >l> > 	One gating factor... how heavy are your writes?  That KZPAC > 	only has 4 Mbyte of cache.e > B > 	With your heavy I/O needs, I would definitely get RAID Software9 > 	for OpenVMS OR go to newer controller (HSZ) and create ; > 	controller based 0+1.   But you are looking at a capitalp request.9 > 	You could always buy those 4.3 Gig drives 5 at a time,  > 	if allowed. > A > 	Tell us more, we can do the best you can with what you have tod
 > 	work with.a >m	 > 				Robc >l >t  0 Some more details (from a compatriot of fooguy):  D The DS20 in question is intended to support an ad-tracking/AR systemE with an Oracle backend.  The Oracle database is quite poorly designedsH and relies on far too many full-table scans and disk i/o than necessary.' Redisigning the database is unfeasible.m  G The SSD is a StorageWorks module DS-EZ832 mounted in a BA356 shelf with E a -FA personality module.  The shelf is connected to a KZPCC (SymbiosdF 53c895) controller which is capable of 80MB/s LVD operation.  That i/o subsytem looks like:  G KZPCC (80MB/s LVD) --> BA356-FA (20MB/s Ultra) --> DS-EZ832 (40MB/s UW)m  G Let's play "spot the bottleneck", shall we?  In that config, the SSD isnD delivering ~13MB/s copying from dir --> dir within the device.  NoteB that there are no physical moving parts (head movement, rotationalF latency) to deal with.  Given the BA356-FA, that's a pretty good rate;H given the $10,000/GB price of the EZ832, I want all 40MB/s it is capableH of.  This i/o subsystem is destined to accomodate the Oracle indexes and* the tables showing the greatest hot spots.  # The other i/o subsystem looks like:o  F KZPAC-CA (20MB/s/channel) --> internal split shelf --> 4 x DS-RZ1DD-VW  C At the fastest (RAID 10 across two channels) we are seeing ~1.5MB/ssF copying dir --> dir within the array and ~6.5MB/s in a BACKUP to NLA0:G Given that a two channel array should have a peak performance of 40MB/siA (to NLA0:), this is pathetic.  For comparison, a Sun Netra T1-105eB with a single disk delivers ~39MB/s TARring to /dev/null, which is% about all a single drive can sustain.m  E The array is intended to handle boot/OS, application, and non-hotspots Oracle data.  F The entire system only needs to support <20 users and less than 5GB ofH data in the production database; cost-justifying a larger system will be quite difficult.       Sent via Deja.comt http://www.deja.com/   ------------------------------    Date: 25 Jan 2001 13:27:28 -0500* From: young_r@eisner.decus.org (Rob Young)7 Subject: Re: Midrange I/O with VMS WAS: DS20 - Slow I/OT+ Message-ID: <cscU7a6IG7Xk@eisner.decus.org>r  N In article <94poic$p0v$1@nnrp1.deja.com>, fooguy <jweisen@my-deja.com> writes:I > I'm not opposed to the idea, but with so many powerful inexpensive RAID D > controllers out there, why would I opt for software mirroring overD > hardware? Especially since we don't own Volume Shadowing Licenses,D > and they're more expensive than a Mylex 64bit PCI RAID Controller. > I > The other thing is that Volume Shadowing and bound volume sets give youiB > RAID 0 and RAID 1, and any combo of the two, but not RAID 5. ForH > something like a DB server (SQL/Oracle) I'm all about RAID 10, but for@ > an RMS based database and file sharing, my preference leans to( > Distributed Data Guarding with Parity. >   B 	You mentioned you had 3 - 18 Gig drives in a RAID5 configuration.? 	If you expect to get 400 I/O per second out of them, you would5 	be delusional.    	So what are your needs?   				Rob    ------------------------------   Date: 25 Jan 2001 10:33 CST ' From: carl@gerg.tamu.edu (Carl Perkins)-6 Subject: Re: Moving from VMS to Tru64 Unix, Any Ideas?- Message-ID: <25JAN200110334725@gerg.tamu.edu>   6 rdeininger@mindspring.com (Robert Deininger) writes...R }I ordered a freebo VMS-related CD from the web site, and it arrived the next day. } A }Then it turned out I couldn't use it, since it's a Pee Cee disk.t }  }--  }Robert Deiningero }rdeininger@mindspring.com    / VMS can read most "PC" formatted CDs just fine.m   MOUNTp     /MEDIA_FORMATt  
     CD-ROM  C          Mounts a volume assuming the media to be ISO 9660 (or High.          Sierra) formatted.             Formatd              /MEDIA_FORMAT=CD-ROMn   --- Carl   ------------------------------  % Date: Thu, 25 Jan 2001 12:04:13 +0000p9 From: "Miller, Daniel" <Daniel.Miller@nightfreight.co.uk>I Subject: MPACK/MUNPACK imagesoA Message-ID: <5E8A0E4920B0D411B1E900508BFCB24029947F@NF-HOUSE-NT1>I   Hi,t  K Does anyone have or know where i can get images for MPACK / MUNPACK for VMSg 7.2-1?   thanks,n  
 Daniel Miller,   ------------------------------    Date: 25 Jan 2001 09:20:53 -0500, From: koehler@eisner.decus.org (Bob Koehler)! Subject: Re: MPACK/MUNPACK images + Message-ID: <LHoBMy8cNukI@eisner.decus.org>y  } In article <5E8A0E4920B0D411B1E900508BFCB24029947F@NF-HOUSE-NT1>, "Miller, Daniel" <Daniel.Miller@nightfreight.co.uk> writes:l > Hi,K > M > Does anyone have or know where i can get images for MPACK / MUNPACK for VMSh > 7.2-1? >   2 According to the readme's from my copy of munpack:  K "The canonical FTP site for this software is ftp.andrew.cmu.edu:pub/mpack/"   F ----------------------------------------------------------------------? Bob Koehler                     | Computer Sciences Corporation-= NASA GSFC Flight Software       | Federal Sector, Civil GroupmE                                 | please remove ".aspm" when replyingS   ------------------------------    Date: 25 Jan 2001 20:03:26 +0800, From: Paul Repacholi <prep@prep.synonet.com>% Subject: Re: Multiple path devices...e- Message-ID: <87y9vzvnxt.fsf@prep.synonet.com>   1 brooks@cuebid.zko.dec.nospam (Rob Brooks) writes:   ? > 	You can't; as someone else noted correctly, it is simply the F > first path that was found when this device was configured.  The term  D Rob, if you boot from a MP devise, does does the primitive IO system5 have all paths for dump/reboot, or is the primary it?)   --  < Paul Repacholi                               1 Crescent Rd.,7 +61 (08) 9257-1001                           Kalamunda.7@                                              West Australia 6076. Raw, Cooked or Well-done, it's all half baked.   ------------------------------  # Date: Thu, 25 Jan 2001 17:03:29 GMTe% From: Uwe Zessin <zessin@my-deja.com>u% Subject: Re: Multiple path devices...c) Message-ID: <94pm8q$mim$1@nnrp1.deja.com>   - In article <87y9vzvnxt.fsf@prep.synonet.com>,)/   Paul Repacholi <prep@prep.synonet.com> wrote: 3 > brooks@cuebid.zko.dec.nospam (Rob Brooks) writes:p > A > > 	You can't; as someone else noted correctly, it is simply thelH > > first path that was found when this device was configured.  The term >hF > Rob, if you boot from a MP devise, does does the primitive IO system7 > have all paths for dump/reboot, or is the primary it?   D I don't know about dump, but the console has (should have) all paths
 for a reboot:(   [...]k Testing the NetworkiB AlphaServer ES40 Console V5.8-43, built on Aug 21 2000 at 16:22:48 P00>>>show devE dga101.1001.0.4.1          $1$DGA101                      HSG80  V85F.E dga101.1002.0.4.1          $1$DGA101                      HSG80  V85FeE dgb101.1003.0.3.0          $1$DGA101                      HSG80  V85FeE dgb101.1004.0.3.0          $1$DGA101                      HSG80  V85F*D dka0.0.0.1.1               DKA0              COMPAQ BD00962A66  B007   [...] " I/O paths to device              4@ Path PGB0.5000-1FE1-000B-F0C2 (xyz), primary path, current path.F  Error count              0    Operations completed               4319$ Path PGA0.5000-1FE1-000B-F0C4 (xyz).F  Error count              0    Operations completed                 15$ Path PGA0.5000-1FE1-000B-F0C3 (xyz).F  Error count              0    Operations completed                 15$ Path PGB0.5000-1FE1-000B-F0C1 (xyz).F  Error count              0    Operations completed                 14 $   C There are also environment variables (FFAUTO,FFNEXT) to modify that 
 behaviour.   --
 Uwe Zessin3 (If you want to send mail, please use user "zessin" / who lives at "decus.decus.de", not my-deja.com)i     Sent via Deja.comi http://www.deja.com/   ------------------------------  % Date: Thu, 25 Jan 2001 10:55:29 +0200 6 From: Graham van der Vaart <Graham.vd.Vaart@ast.co.za> Subject: NetbackupG Message-ID: <0B3DA236FD2E624299206C529081363401F0D3@platinum.ast.co.za>t   Hi  F I am busy evaluating backup/restore products for the OpenVMS platform.  8 I have a few questions about the Netbackup offering.....  ? 1.)	Exactly which Netbackup Configuration options (bp.conf) areo available for VMS?D 2.)	In the client CD text files there is only mention of a couple of$ notify scripts for pre and post exec0 	Is that all there is on this particular client.? 3.)	If true image backup is not supported. What about open filee backup. : 	VMS supports /IGNORE=3DINTERLOCK on their BACKUP command.< 4.)	I wonder if anybody out there has more info than what is available on the CD.   Regardsn Graham Van der vaart   ------------------------------  % Date: Thu, 25 Jan 2001 09:20:12 -0500a& From: Ken Robinson <ksrobin@erenj.com># Subject: NETBIOS Aborting (PW 6.0C)h7 Message-ID: <5.0.0.25.2.20010125091819.0377e050@clmail>a   Environment:             OpenVMS 7.2-1 on a DS204          Pathworks 6.0C, TCP/IP set as the transport.          TCP/IP for VMS 5.0a with PWIP enabled          Decnet Phase 5 1          The node in question is defined as a PDCl   Problem:  D          Upon starting Pathworks, the NetBios process dies with the 9 following  messages in the log file: (some lines wrapped)   M %NB-W-ERRSIGNAL, exceptional conditional detected at 25-JAN-2001  19:41:16.80r- -SYSTEM-W-NOSUCHDEV, no such device available / %TRACE-W-TRACEBACK, symbolic stack dump follows K    image    module    routine             line      rel PC           abs PC @   NETBIOS  NBPROC  NBSIGNAL              10739 000000000000ECB0  000000000003ED00@   NETBIOS  NBPROC  STARTUP                1320 000000000000079C  00000000000307EC/                                              0 a" FFFFFFFF9AC813F4  FFFFFFFF9AC813F4L %NB-F-ERRSIGNAL, exceptional conditional detected at 25-JAN-2001 19:41:18.19% -SYSTEM-F-IVCHAN, invalid I/O channelp/ %TRACE-F-TRACEBACK, symbolic stack dump followsyK    image    module    routine             line      rel PC           abs PC @   NETBIOS  NBPROC  NBSIGNAL              10739 000000000000ECB0  000000000003ED00@   NETBIOS  NBPROC  GETETHERPARAMETER      3040 0000000000002F00  0000000000032F50@   NETBIOS  NBPROC  STARTUP                1337 000000000000084C  000000000003089C@                                              0 FFFFFFFF9AC813F4  FFFFFFFF9AC813F4   Thanks for any help in advance.o   Ken Robinson ksrobin@erenj.com    ------------------------------    Date: 25 Jan 2001 16:36:43 +0100* From: eplan@kapsch.net (Peter LANGSTOEGER)' Subject: Re: NETBIOS Aborting (PW 6.0C)t* Message-ID: <3a70480b$1@news.kapsch.co.at>  ` In article <5.0.0.25.2.20010125091819.0377e050@clmail>, Ken Robinson <ksrobin@erenj.com> writes:
 >Environment:o > ! >         OpenVMS 7.2-1 on a DS20g  B I hope you have seen/checked/installed the umpteen ECOs for 7.2-1.  5 >         Pathworks 6.0C, TCP/IP set as the transporti  E I hope you have seen/checked/installed the ECO 1 for PATHWORKS V6.0C.dD I hope you also know of Advanced Server V7.2A (with ECO3 and patchesH thereafter) and have heard of PATHWORKS V6.1 and ASOVMS V7.3 coming real soon now...   / >         TCP/IP for VMS 5.0a with PWIP enableda  N I hope you have seen/checked/installed the ECO 1 for TCPIP V5.0-11 (aka V5.0A)   >         Decnet Phase 5  F I hope you have seen/checked/installed the ECO 2 for DNVOSI (V7.2-1) ?  2 >         The node in question is defined as a PDC >c	 >Problem:  > E >         Upon starting Pathworks, the NetBios process dies with the o: >following  messages in the log file: (some lines wrapped) >aN >%NB-W-ERRSIGNAL, exceptional conditional detected at 25-JAN-2001  19:41:16.80. >-SYSTEM-W-NOSUCHDEV, no such device available  > Do you have more than one network interface installed/in-use ?M Do you have defined logicals to force vms/osi/pathworks to not use them all ?r   eg.h  $ $	DEF/SYS/EXE	DNS$DEVICE		FWA0:	!OSI$ $	DEF/SYS/EXE	DTS$DEVICE		FWA0:	!OSI& $!	DEF/SYS/EXE	LAST$DEVICE		FWA0:	!VMS$ $	DEF/SYS/EXE	LAT$DEVICE		FWA0:	!VMS) $	DEF/SYS/EXE	NETBIOS$DEVICE		FWA0:	!PWRKf/ $	DEF/SYS/EXE	PWRK$KNBDAEMON_DEVICE	FWA0:	!PWRK - $	DEF/SYS/EXE	PWRK$NETBEUI_DEVICE	FWA0:	!PWRKu   HIHb   -- e< Peter "EPLAN" LANGSTOEGER           Tel.    +43 1 81111-2651; Network and OpenVMS system manager  Fax.    +43 1 81111-888n< <<< KAPSCH AG  Wagenseilgasse 1     E-mail  eplan@kapsch.netH A-1121 VIENNA  AUSTRIA              "I'm not a pessimist, I'm a realist"   ------------------------------  # Date: Thu, 25 Jan 2001 12:04:36 GMTi From: Didier.Morandi@gmx.frt Subject: Re: Newbie question) Message-ID: <94p4oh$6an$1@nnrp1.deja.com>*  A > 	After some time playing with different unixes, I'd like to trya
 > Open VMS OpenVMS is not a game.  3 > CPU: Alphaserver 4/233 32 Mb RAM (guess it's OK).dF Good guess. All Alphaservers can run OpenVMS (afaik) but 32Mb seems to me a bit "light".o  G > Disk: Samsung or IBM SCSI disks (I've read somewhere that not all theS3 > SCSI disks could be used with VMS. Is it true ? ) E VMS doesn't care about disks as long as they are in a "regular" Alphao	 platform.n  H > Video card: Can I use any standard video card - ie S3 / Matrox / ATI ?> VMS doesn't care about Video, as long as you don't want to useF DECWindows, and anyway you don't need DECwindows. It's a (better) copyF of the MacroHard stuff... Use  a VT or a Telnet connection instead :-)  : > CDROM: Good ol' Mitsubishi 6x SCSI (guess that's OK too) Same answer than for disks  D > Network card: I've got 3Com 3c509/3c900 and a WD8003. Would one of6 > these run ? Or do I have to buy a DEC network card ?7 Same answer than for disks (but I'm not a HW expert :-)m   D.     Sent via Deja.coma http://www.deja.com/   ------------------------------  % Date: Thu, 25 Jan 2001 13:17:31 -0300I) From: fabio_compaq@ep-bc.petrobras.com.br( Subject: OpenVMS ConsortiumiL Message-ID: <OF5C6A43D7.A5D68F10-ON032569DF.00594842@ep-bc.petrobras.com.br>  G This initiative of the Linux world must be copied by the  OpenVMS worldi$ to put OVMS in its the correct place  E http://www.cnn.com/2001/TECH/computing/01/25/linux.lab.idg/index.htmld   Regards    FC   ------------------------------  % Date: Thu, 25 Jan 2001 15:41:16 +0000d) From: Christof Brass <brass@infopuls.com>p Subject: Re: OpenVMS Consortiume, Message-ID: <3A70491C.4D4F47A8@infopuls.com>  * fabio_compaq@ep-bc.petrobras.com.br wrote: > I > This initiative of the Linux world must be copied by the  OpenVMS worldl& > to put OVMS in its the correct place > G > http://www.cnn.com/2001/TECH/computing/01/25/linux.lab.idg/index.htmls > 	 > Regardsy >  > FC  @ Very interesting reading. The first what comes to my mind is are; there only Intel systems? Second question: why aren't there , Alphas? Third question: what will the OS be?   ------------------------------  # Date: Thu, 25 Jan 2001 16:05:20 GMTe From: mikeg_myson@my-deja.coms Subject: OpenVMS problem) Message-ID: <94piro$j0c$1@nnrp1.deja.com>   0 Can anyone help?? We are running OpenVMS version/ 7.2-1. When I try to make my terminal the opcom / by typing in REPLY/ENABLE at the command line Iy2 get the message %SYSTEM-F-MBFULL, mailbox is full.+ Anyone know how I can empty this mailbox or , indeed how I can identify what mailbox it is bleeting on about?   Mike.n     Sent via Deja.com' http://www.deja.com/   ------------------------------  # Date: Thu, 25 Jan 2001 16:09:16 GMTt From: mikeg_myson@my-deja.coms Subject: OpenVMS problem) Message-ID: <94pj33$jb0$1@nnrp1.deja.com>r  ? Can anyone help?? OpenVMS version 7.1-2. When trying to make myaF terminal the opcom by typing in REPLY/ENABLE I get the error message %G SYSTEM-F-MBFULL, mailbox is full. Anyone any idea how I can detect whats: mailbox is full and exactly what vms is bleeting on about?     Sent via Deja.come http://www.deja.com/   ------------------------------  # Date: Thu, 25 Jan 2001 11:13:09 GMT.= From: system@SendSpamHere.ORG (Brian Schenkenberger, VAXman-)0! Subject: Re: RW-CD drives for VMSt0 Message-ID: <009F6A1E.5A487EB4@SendSpamHere.ORG>  ] In article <EJEDILJIPPJOIEFOLDHLAEMCCEAA.tom@kednos.com>, Tom Linden <tom@kednos.com> writes:eE >When I posted my original question I was hoping to get some guidancelJ >from the troops that had crawled throught the muck. Not interested in DVD >so let's drop that topic. >5D >What can I buy off the shelf that will work with VMS?  Is there any >experiencewM >out there that can be shared?  A list of supported equipment would be a nicef	 >feature.t >t >   ; That certainly wasn't well expressed in your first posting.r  J That said, I have a Yamaha CRW4416S.  Yamaha has newer models which "burn"J at faster rates (The 4416 is 4xR 4xRW 16xRead). I believe there is a newerI model something like CRW8432S.  I have also had a Plextor model on my boxoI as well as a Smart&Friendly device.  Sorry, I don't recall the models buteH I might be able to find a reference to the Plextor I tested and post its
 model number.M  H If you are truly looking for models compatible with VMS, now that you've4 asked, I'm sure others will relay their experiences.   --O VAXman- OpenVMS APE certification number: AAA-0001     VAXman(at)TMESIS(dot)COM3            fO city, n., 1. a place where trees are cut down and streets are named after them.    ------------------------------  + Date: Thu, 25 Jan 2001 09:31:50 +0000 (UTC)p' From: Osmo Kujala <kujala@tukki.jyu.fi>i0 Subject: Re: Source for used StorageWorks disks?, Message-ID: <94orq6$t8j$1@mordred.cc.jyu.fi>  - Paul Repacholi <prep@prep.synonet.com> wrote:e+ > Osmo Kujala <kujala@tukki.jyu.fi> writes:e  F >> How about the drive connector. Is there easy way to find out if the& >> connector is CSA type for example?   C > SCA. 80 way, SCSI power select etc all in one plug. In a standardh > place!  J Yes. For example. If such even exists, but there are different flex cables@ in containers. Is there easy way to find out which model of flexH cable it is (without opening the container of course, when trying to buy these things)? l  J >> If the drive is ultra wide and markings are 20,F10,W,W (fast, wide, not                      ^D >> ultra), is it possible to get to ultra wide (40 MB/s) by changing@ >> SCSI adapter only?                                          ^  G > No, ultra refers to the siganing on the wires. You have to have Ultrae > drives and controllers.w  # Paul, you must be very busy.... :-)-# I'll try to ask the right question:5J I have AS4000 with BA356-SB Storageworks and KZPDA-controllers(fast-wide).I I'd like to get SCSI faster, which means ultra. I have ultra-wide drives, ; but before investing to ultra controllers I'd like to know:y  E Does BA356-SB (and cable BN21K-02 and "connector piece" 70-31490-01) f support ultra??     (I'd guess yes, because cabling is wide, but cable lenght?)r   regards            Osmo Kujalan   ------------------------------  # Date: Thu, 25 Jan 2001 13:34:54 GMTt% From: Uwe Zessin <zessin@my-deja.com> 0 Subject: Re: Source for used StorageWorks disks?) Message-ID: <94pa1u$agh$1@nnrp1.deja.com>R  , In article <94orq6$t8j$1@mordred.cc.jyu.fi>,*   Osmo Kujala <kujala@tukki.jyu.fi> wrote:/ > Paul Repacholi <prep@prep.synonet.com> wrote:d- > > Osmo Kujala <kujala@tukki.jyu.fi> writes:  > H > >> How about the drive connector. Is there easy way to find out if the' > >> connector is CSA type for example?  > E > > SCA. 80 way, SCSI power select etc all in one plug. In a standardi
 > > place! >hE > Yes. For example. If such even exists, but there are different flexeD > cables in containers. Is there easy way to find out which model ofA > flex cable it is (without opening the container of course, when2 > trying to buy these things)?  A I am afraid not. Also, I have seen disk carriers with a small fand8 in it(!), but I don't recall any disk names. Be careful!  B > >> If the drive is ultra wide and markings are 20,F10,W,W (fast, > wide, notk >                      ^F > >> ultra), is it possible to get to ultra wide (40 MB/s) by changingB > >> SCSI adapter only?                                          ^  B No, I don't think so, but how do you know the drive is ultra-wide?  9 Hm. '20,F10,W,W' I find on a DS-RZ28M-VZ (green) carrier.E% Are you sure you have an U-SCSI disk?-= On a DS-RZ1DB-VW (top gun blue) carrier I read: '40,F20,W,W'.0  C > > No, ultra refers to the siganing on the wires. You have to have:! > > Ultra drives and controllers.   F And I/O modules (enough power for the U-SCSI module - 180W are needed,C if I recall correctly), appropriate cables and a suitable backplane 9 (the U-SCSI variant has additional shielding) in the box.-  % > Paul, you must be very busy.... :-)-% > I'll try to ask the right question:oF > I have AS4000 with BA356-SB Storageworks and KZPDA-controllers(fast-F > wide). I'd like to get SCSI faster, which means ultra. I have ultra-D > wide drives, but before investing to ultra controllers I'd like to > know:  >iF > Does BA356-SB (and cable BN21K-02 and "connector piece" 70-31490-01) > support ultra?  C Is that a grey box? Then I would not try this out. At least you areeC violating FCC (or what ever they are called in your company) rules./2 Such a configuration might also corrupt your data.  A >     (I'd guess yes, because cabling is wide, but cable lenght?)S  A Wide cable means 'nothing', sorry. DEC _did_ sell U-SCSI _narrow_r% SBBs (e.g. DS-RZ1CF-VA, DS-RZ1DF-VA).   A Also, there are _three_ different U-SCSI I/O modules (also called-C 'personality modules') for the BA356 U-SCSI box: 2 single-ended and<@ a differential one that allows cable length of up to 20.5 meters for a point-to-point segment.s   --
 Uwe Zessin3 (If you want to send mail, please use user "zessin" . who lives at "decus.decus.de", not my-deja.comD  Please understand that I don't have time for individual consulting)     Sent via Deja.com, http://www.deja.com/   ------------------------------    Date: 26 Jan 2001 00:29:05 +0800, From: Paul Repacholi <prep@prep.synonet.com>0 Subject: Re: Source for used StorageWorks disks?- Message-ID: <87elxr61f2.fsf@prep.synonet.com>   ) Osmo Kujala <kujala@tukki.jyu.fi> writes:   % > Paul, you must be very busy.... :-)o  % > I'll try to ask the right question:   L > I have AS4000 with BA356-SB Storageworks and KZPDA-controllers(fast-wide).K > I'd like to get SCSI faster, which means ultra. I have ultra-wide drives,g= > but before investing to ultra controllers I'd like to know:f   > G > Does BA356-SB (and cable BN21K-02 and "connector piece" 70-31490-01) o > support ultra?A >     (I'd guess yes, because cabling is wide, but cable lenght?)o  A Cable length you will have to check against the exact config. ButtB remember the big plus of the original ultra was extra length, then< more speed very soon after. And if a 2' cable is too long...  < Compaq REALLY ought to get the cables catalogue back, and an? SOC that contains enough info to enable peole to sort this sort 
 of stuff out.n   -- e< Paul Repacholi                               1 Crescent Rd.,7 +61 (08) 9257-1001                           Kalamunda.$@                                              West Australia 6076. Raw, Cooked or Well-done, it's all half baked.   ------------------------------  % Date: Thu, 25 Jan 2001 12:32:44 -0500g2 From: "Carmine Castiglia" <ccastiglia@engsint.com> Subject: Telnet, yes; HTTP, no+ Message-ID: <94pnvp$ff1$1@bob.news.rcn.net>A  ' OpenVMS 7.1-1H2,  Decnet Phase IV (UCX)   L Connecting an HP JetDirect device to the network using TCP/IP services.  DidF the SET HOST thing in UCX, set the IP address in the JetDirect device.  J I can telnet to the device, read the configuration table, etc.  I *cannot*; reach the device's built-in web server - attempting to loadxK http://151.106.50.20 (as described in JetDirect docs) from Netscape browser J results in "unable to connect to server" type messages.  Also (and this isI why we're here), I cannot make the device function as a server queue - itsK seems to setup correctly but then attempts to print result in %UCS-F-NOMSG,t Message number 030AEBB4.  B Note that I have a similar setup for a networked Xerox machine andJ everything works fine.  I assume that I have forgotton to do something but2 can't for the life of me figure out what it was...  
 Some details:r     UCX> sho hosto        LOCAL databasep     Host address    Host name9     151.106.50.20   HPDIR, hpdir&   127.0.0.1       LOCALHOST, localhost   151.106.50.10   XER01, xer01   151.106.50.1    esp98, ESP98   UCX>     UCX> sho arp1                                 ARP table entrieso  C   Ethernet           Internet        Host name           ARP statusnC   08-00-3E-30-D0-DB  151.106.50.10   XER01               INUSE CMPLfC   00-01-E6-2D-17-C9  151.106.50.20   HPDIR               INUSE CMPL.   UCX>  '  $ type sys$manager:ucx$lpd_startup.comb
     <snip>     $ init/queue/start-t                 /owner=[1,4]-B                 /retain=error-                 /device=server-h%                 /process=ucx$lpd_smb-a5                 /protection=(s:e,g:r,w:w,o:d) HPDIR --)                 /library=emsdevctl_hplj -0                 /form=normal --                 /default=(nofeed,form=normal)h     $!
     <snip>   ------------------------------  # Date: Thu, 25 Jan 2001 17:51:16 GMTS= From: system@SendSpamHere.ORG (Brian Schenkenberger, VAXman-)a" Subject: Re: Telnet, yes; HTTP, no0 Message-ID: <009F6A55.F86EF90F@SendSpamHere.ORG>  ` In article <94pnvp$ff1$1@bob.news.rcn.net>, "Carmine Castiglia" <ccastiglia@engsint.com> writes:( >OpenVMS 7.1-1H2,  Decnet Phase IV (UCX)   Hi Carmine!e  M >Connecting an HP JetDirect device to the network using TCP/IP services.  DidcG >the SET HOST thing in UCX, set the IP address in the JetDirect device.i >lK >I can telnet to the device, read the configuration table, etc.  I *cannot*o< >reach the device's built-in web server - attempting to loadL >http://151.106.50.20 (as described in JetDirect docs) from Netscape browserK >results in "unable to connect to server" type messages.  Also (and this is7J >why we're here), I cannot make the device function as a server queue - itL >seems to setup correctly but then attempts to print result in %UCS-F-NOMSG, >Message number 030AEBB4.H  E Question.  Can you use Netscape to view other web sites?  Do you haveyA a proxy setup in  Netscape?  Options/Network Preferences/Proxies.   H Are you sure you access the built-in web server via the normal http port5 (port 80) or is there another port number to be used.R   --O VAXman- OpenVMS APE certification number: AAA-0001     VAXman(at)TMESIS(dot)COM             0O city, n., 1. a place where trees are cut down and streets are named after them.F   ------------------------------    Date: 25 Jan 2001 18:58:24 +0100* From: eplan@kapsch.net (Peter LANGSTOEGER)" Subject: Re: Telnet, yes; HTTP, no( Message-ID: <3a706940@news.kapsch.co.at>  ` In article <94pnvp$ff1$1@bob.news.rcn.net>, "Carmine Castiglia" <ccastiglia@engsint.com> writes:( >OpenVMS 7.1-1H2,  Decnet Phase IV (UCX) >0M >Connecting an HP JetDirect device to the network using TCP/IP services.  DidoG >the SET HOST thing in UCX, set the IP address in the JetDirect device.i >aK >I can telnet to the device, read the configuration table, etc.  I *cannot*h< >reach the device's built-in web server - attempting to loadL >http://151.106.50.20 (as described in JetDirect docs) from Netscape browser9 >results in "unable to connect to server" type messages. 6  J 1.) There is not one HP JetDirect device. There are umpteen different ones: 	in three different classes (External, MIO-Slot, EIO-Slot)I 2.) Not all of the HP JetDirect Devices have webservers built in (memory)oI 3.) Only a more recent F/W version for the card will contain a webserver.   ? Start with http://www.hp.com/cposupport/nonjsnav/netprn.html or B http://www.hp.com/cposupport/eschome.html to find the HP JetDirectI device type, description, Firmware and Release Notes appropriate for yourR model.  L >                                                          Also (and this isJ >why we're here), I cannot make the device function as a server queue - itL >seems to setup correctly but then attempts to print result in %UCS-F-NOMSG, >Message number 030AEBB4.h  B Sorry, unknown to me (except %UCX-F-NOMSG instead of %UCS-F-NOMSG)  C >Note that I have a similar setup for a networked Xerox machine andIK >everything works fine.  I assume that I have forgotton to do something butK3 >can't for the life of me figure out what it was...T  N TELNET to port 80 (http) or 515 (lpd) to see, if the HPJD accepts a connectionG at all. Really old HPJD didn't have LPD, only stream (TCP Port 9100)...1  	 Good luck    -- A< Peter "EPLAN" LANGSTOEGER           Tel.    +43 1 81111-2651; Network and OpenVMS system manager  Fax.    +43 1 81111-888 < <<< KAPSCH AG  Wagenseilgasse 1     E-mail  eplan@kapsch.netH A-1121 VIENNA  AUSTRIA              "I'm not a pessimist, I'm a realist"   ------------------------------  % Date: Thu, 25 Jan 2001 13:52:44 -0500 2 From: "Carmine Castiglia" <ccastiglia@engsint.com>" Subject: Re: Telnet, yes; HTTP, no+ Message-ID: <94pslq$7t1$1@bob.news.rcn.net>g  K Boy, it's tough when you don't do these things often enough to remember allh of the details...s  / One thing led to another until I took a look at I SYS$SPECIFIC:[UCX_LPD}HPDIR.LOG and found an error message concerning the I spool directory.  Uh-oh, didn't manually create it (why the heck should I K have to do that manually anyway?).  Created the directory and can now printo via server queue.d  I FWIW, you were correct about the web server - this JetDirect EX Plus doesEJ not have one built-in.  The docs (on CD-ROM like everything else nowadays)J cover multiple JetDirect devices and do not clearly indicate which ones do# not have the web server capability.s  I Thanks Peter and Brian (who helped me a great deal yesterday with another  issue).       5 Peter LANGSTOEGER <eplan@kapsch.net> wrote in messagef" news:3a706940@news.kapsch.co.at...A > In article <94pnvp$ff1$1@bob.news.rcn.net>, "Carmine Castiglia"-  <ccastiglia@engsint.com> writes:* > >OpenVMS 7.1-1H2,  Decnet Phase IV (UCX) > >eJ > >Connecting an HP JetDirect device to the network using TCP/IP services. Did6I > >the SET HOST thing in UCX, set the IP address in the JetDirect device.L > >wD > >I can telnet to the device, read the configuration table, etc.  I *cannot*> > >reach the device's built-in web server - attempting to loadF > >http://151.106.50.20 (as described in JetDirect docs) from Netscape browsera: > >results in "unable to connect to server" type messages. >mL > 1.) There is not one HP JetDirect device. There are umpteen different ones; > in three different classes (External, MIO-Slot, EIO-Slot)sK > 2.) Not all of the HP JetDirect Devices have webservers built in (memory)tK > 3.) Only a more recent F/W version for the card will contain a webserver.0 >5A > Start with http://www.hp.com/cposupport/nonjsnav/netprn.html orsD > http://www.hp.com/cposupport/eschome.html to find the HP JetDirectK > device type, description, Firmware and Release Notes appropriate for yourl > model. > K > >                                                          Also (and thism isL > >why we're here), I cannot make the device function as a server queue - it@ > >seems to setup correctly but then attempts to print result in
 %UCS-F-NOMSG,  > >Message number 030AEBB4.p > D > Sorry, unknown to me (except %UCX-F-NOMSG instead of %UCS-F-NOMSG) >EE > >Note that I have a similar setup for a networked Xerox machine andoI > >everything works fine.  I assume that I have forgotton to do somethinge but 5 > >can't for the life of me figure out what it was...- >-E > TELNET to port 80 (http) or 515 (lpd) to see, if the HPJD accepts ay
 connectionI > at all. Really old HPJD didn't have LPD, only stream (TCP Port 9100)...  >a > Good luck? >p > --> > Peter "EPLAN" LANGSTOEGER           Tel.    +43 1 81111-2651= > Network and OpenVMS system manager  Fax.    +43 1 81111-888,> > <<< KAPSCH AG  Wagenseilgasse 1     E-mail  eplan@kapsch.netJ > A-1121 VIENNA  AUSTRIA              "I'm not a pessimist, I'm a realist"   ------------------------------  # Date: Thu, 25 Jan 2001 14:31:26 GMT / From: "John Nixon" <jorlnixon@worldnet.att.net>. Subject: UCX Route problemE Message-ID: <2NWb6.9541$1m.601346@bgtnsc04-news.ops.worldnet.att.net>8   We are running a mixed cluster.e  . The VAXes have VMS 7.1 with TCP/IP V4.2 eco 4.+ The Alphas have VMS 7.2-1 with TCP/IP V5.0Ao  C We have a dynamic network,  ie, there are always changes being made 
 somewhere.  L Here is a little history.  Some of the items are easily explainable, even by me.  Others are not.  K We had a single gateway, and this  allowed our VAXes and Alphas to access at network printer via TCP/IP. 1 We added another gateway, and everybody was fine.tL About a month ago, the network guys removed the old gateway.  The VAXes wereL still able to print to a printer, but the Alphas could not.  That was ok. We, had to define the new gateway to the Alphas.   Now it gets curious.K About a week ago I lost my telnet connections to one of the Alphas, but notw# the other one (in the same cluster)yL UCX SHOW ROUTE showed the old gateway address (which has since been assigned to a new node).nK We redefined the route to point to the new gateway.  About an hour later, Iw% lost connectivity again, and we againrK redefined the route, and it worked again.  This was only on one Alpha.  Theo other Alpha node seemed stable.T  K However, this morning, I again lost my sessions, but this time, both Alphas K showed the old gateway address.  None of the VAXes are having any problems.,  G We obviously have some sort of TCP/IP configuration problem, but we areiK stumped.  If anyone has any ideas, or if I need to provide any more details@ or information, please ask.p   ------------------------------  % Date: Thu, 25 Jan 2001 11:05:13 -0500 ' From: "Dale A. Marcy" <dqm@y12.doe.gov>t6 Subject: Re: Urgent:Error recovering files from backup+ Message-ID: <3A704EB8.DD1D874F@y12.doe.gov>S   Nivlesh,  M      I am afraid that I cannot help you with your current restore, but I havenK seen similar errors when trying to restore a save set that was created by ayN user with improper process quotas.  The backup command to produce the save setG did not produce an error, but when you tried to restore it, it producedwM similar errors.  If the backup command to create the save set was done with atI /VERIFY, then you would see the errors during the verify portion.  In ourpM case, it was not fatal, because we were moving executables from a developmenthM system to a production system, so we could go back and try again until we got  it correct.t  
 Dale Marcy. Science Applications International Corporation   Nivlesh Chandra wrote: > N > I have a tape that contains a  backup that was done using the backup commandG > in VMS. Now when I try and recover the file I get the following erroro > 6 > excessive error rate reading $2$MKA300:[000000]*.*;* > software header crc error0 > + > I used the following to restore the files  > " > $ mount $2$mka300 /media=compactN > $ backup $2$mka300:[000000]bak-13-12-99.bck;1 /save disk44:[recover...]*.*;*0 > /log /list=disk$scratch:bak.lis /nocrc /rewind > J > Then since I was getting so much error.. I thought I would just copy theH > saveset onto the disk and then recover .. I used the following command > A > $ backup $2$mka300:[000000]*.*;* disk44:[test]*.*;* /log /nocrco > % > but received the same error messageo > L > Please I really need to get the files in this saveset and if someone could* > help me I would really appreciate it ... >  > Please respond...a >  > Niv    ------------------------------  % Date: Thu, 25 Jan 2001 16:57:26 +0000 - From: Tim Llewellyn <tim.llewellyn@bbc.co.uk>i6 Subject: Re: Urgent:Error recovering files from backup) Message-ID: <3A705AF6.B92D5A43@bbc.co.uk>e   >   Nivlesh Chandra wrote: > >pP > > I have a tape that contains a  backup that was done using the backup commandI > > in VMS. Now when I try and recover the file I get the following erroro > >e8 > > excessive error rate reading $2$MKA300:[000000]*.*;* > > software header crc error1 > > - > > I used the following to restore the files  > >u$ > > $ mount $2$mka300 /media=compactP > > $ backup $2$mka300:[000000]bak-13-12-99.bck;1 /save disk44:[recover...]*.*;*2 > > /log /list=disk$scratch:bak.lis /nocrc /rewind > >   M Did you try mounting the tape /foreign? If not, and the tape was written withi blocks greaterL than the max disk block size allowed by RMS, then some very weird errors are	 produced,' certainly on VMS 7.1.e       > L > > Then since I was getting so much error.. I thought I would just copy theJ > > saveset onto the disk and then recover .. I used the following command > >cC > > $ backup $2$mka300:[000000]*.*;* disk44:[test]*.*;* /log /nocrcS > >a' > > but received the same error messagex >   M Hmmm, sdavesets with blocksize larger than the max disk block size allowed by M RMS (a little less than 32kbytes, the exact value was pisted here recently in. another + thread) can't be copied off to disk either.   & Worth trying a MOUNT/FOREIGN, I'd say.  L I won't recount the full story of the day I was fooled by this into thinkingI my bacups were toast! It was my birthday and I really did wish I'd stayedy= at home that day when I found out my worried were illfounded.n     > > N > > Please I really need to get the files in this saveset and if someone could, > > help me I would really appreciate it ... > >a > > Please respond...t > >e > > NivS   --6 Tim Llewellyn, OpenVMS Infrastructure, Remarcs Project0 MedAS at the BBC, Whiteladies Road, Bristol, UK.A Email tim.llewellyn@bbc.co.uk. Home tim.llewellyn@cableinet.co.uk-  A I speak for myself only and my views in no way represent those of< MedAS or the BBC.e   ------------------------------  % Date: Thu, 25 Jan 2001 07:35:22 +0000t) From: Christof Brass <brass@infopuls.com>m@ Subject: Re: VMS vs UNIX (was: Linux worm and RedHat 7.0 broken), Message-ID: <3A6FD73A.582D2AEA@infopuls.com>   Charles Sebold wrote:cG > I don't think that Linus himself had any political or anti-commercial-I > vision originally, in fact I don't think he does now, either.  A bit of<E > an anti-Microsoft agenda at times, perhaps, but the few things that?I > Linus has said about Free Software suggest that he's not a fanatic; thekB > original GPL-ing of the kernel was more of a convenience, and anG > incentive for others to come hack the kernel with him without fear ofg# > losing their work, than anything.e  ; I clearly remember that he said in an interview that he wast3 pissed off that the UNIX sources weren't available.t  H > I think that originally his challenge wasn't to create a new operatingH > system but to create something usable and recognizable (as you pointedH > out earlier, as a college student, no less), and possibly just to bestG > Tannenbaum and Minix with a monolithic kernel.  Now it's such a beastiH > that truly new innovations are hard to work into the scheme of things;D > they have to keep up with the Joneses (Solaris Jones and NT Jones,
 > mostly).  > Agreed. But I think he missed a big chance because it had been> better to explore new territory than to repeat the experiences@ others already had made with this sort of designless approach of OS.K) What do you mean by "Joneses" or "Jones"?   H > But of course Linus claims that VMS was all engineering and no design,  > so what does he know? *smiles*  < "Linus" did claim that? I've heard that he knew how to spell "VMS". Are you sure he did?.   > --$ > Charles Sebold, Systems Specialist& > LCMS - Office of Information Systems7 > *** Opinions expressed herein are not necessarily *** 7 > *** those of the Lutheran Church - Missouri Synod ***  > -- > 29th of Teveth, 5761 > --I > A consultant is a person who borrows your watch, tells you what time itd5 > is, pockets the watch, and sends you a bill for it.e   ------------------------------  % Date: Thu, 25 Jan 2001 10:43:08 +0100d= From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>"@ Subject: Re: VMS vs UNIX (was: Linux worm and RedHat 7.0 broken)) Message-ID: <3A6FF52C.827CA548@gtech.com>c   Christof Brass wrote:i= > I clearly remember that he said in an interview that he wasi5 > pissed off that the UNIX sources weren't available.   2 Unix sources being available is not a new concept.  , You have always been able to get BSD source.  * You have also been able to get VMS source.  3 SUN either has made Solaris source available or are  planning to do.o   The important issues are:    - what do the sources cost(   - what are you allowed to do with them  7 Linux adopted the GNU/GPL/FSF/Stallman ideas going back1 to the 1980's.   Arne,   ------------------------------  % Date: Thu, 25 Jan 2001 10:26:49 +0000 ) From: Christof Brass <brass@infopuls.com> @ Subject: Re: VMS vs UNIX (was: Linux worm and RedHat 7.0 broken), Message-ID: <3A6FFF69.61A3CA45@infopuls.com>   sabolich@my-deja.com wrote:  > B > > Stating that Linux (or should we say Linus Thorvalds?) startedA > > without any grand vision seems to kill the enthusiasm and the < > > trust of getting something real innovative with Linux. I= > > wouldn't go that far but I think the vision wasn't on the D > > technical, specifically the architectural, the algorithmical and> > > the UI side. Could we agree that there was a vision on the: > > social, political and eventually anti-commercial side? > G > No.  Linux (strictly speaking) is just a kernel while GNU/Linux is anrH > OS.  UI has no place in the kernel (which is the only part Linus did).I > When it was started I doubt there was any grand vision on Linus's part. D > The vision you speak of is the GNU part, Free Software Foundation.I > Have a look at www.gnu.org.  GNU had libraries compilers, editors, etc.CI > At the time they had no kernel but were working on the HURD.  The LinuxsJ > kernel was put under GPL and put together with software from GNU to make > a working OS.   > A kernel without API and UI is useless. And I mentioned the UI= part only as the last technical aspect. I wasn't too specificg@ but I never heard that the kernel was innovative with respect to# its architecture or its algorithms.a  C > > Agreed. I don't know. But I regret that even in better informede= > > circles it is not mentioned that much. And with technicaltD > > discussions Linux is much more mentioned, analysed and compared.A > > So apart from any hype I think there is a level of popularityc) > > which could be helpful and desirable.  > A > have a look at http://uptime.netcraft.com/up/today/top.avg.htmlr  @ Checked that and was first impressed of the frequently mentioned@ FreeBSD. A second look raised some doubts about how useful these figures are.; What does this mean "For performance reasons, we limit this)? monitoring process to the most frequently requested sites." Who @ does request these sites? I personally doubt that the third site@ on the list "Universitaetsklinikum Rudolf Virchow" is frequently
 requested.= If you take a closer look at the last column "Netblock Owner"@> you won't be very happy if you detect that the same owner e.g.@ "Hopemoon Internet" but others also show up several times. It is7 in no way clear to me that the results are independent.D? Technically the different URLs in the first column belonging toT? the same "Netblock Owner" could be served with the same server. ? But the worst: I don't see any relation with what I was talkingc about.  A > What is talked about and what is used are not allways the same.   9 True but not helpful in our situation. I doubt that thesee? netcraft numbers are reliable or useful in any respect. In facts: I think that there are about 1'000'000 times as many Linux? system in usage than FreeBSD systems. What is the point of what,	 you said?E  C > > So, here it comes: is it not a little bit strange that a system B > > which has its goal of maximum performance on x86 isn't good at > > SMP? > D > No.  FreeBSD kernel is written in such a way that it cannot safelyG > execute on multiple CPUs simultaneously.  To provide protection a big J > kernel lock is used that enforces this.  However user processes can takeC > advantage of multiple CPUs.  So for example if compiling multipleAG > programs simultaneously multiple CPUs will be used since most time isaI > spent processing outside the kernel.  But under certain conditions likesI > high network throughput (FTP server) only one CPU can be utilized sincex= > much time will be system time in kernel (disk i/o, tcp/ip). I > To allow the kernel to run on multiple CPUs simultaneously fine grained J > locking is required.  Look for this in FreeBSD 5.x.  Linux 2.0 has a bigF > kernel lock like FreeBSD.  Linux 2.2 removed the big lock but is notG > really fine grained - so SMP performance is not that good.  Linux 2.4hI > does have fine grained locking and SMP performance is quite good - bestN > of the free *nixes.   @ You are completely right but it seems to me that you are missing@ my point. I expect a system which is optimised for a certain cpu? type with respect to performance to be perfomant. Unfortunatelys> it is not performant when it comes to systems of that specific= cpu type which are built for high performance requirements. Ic4 would call that a clear example of missing the goal.  = > > Are you sure about this FreeBSD's goal? I remember BSD386o0 > > and think that you might confuse these both. > G > FreeBSD focus - high performance on x86, also runs on Alpha (check to 5 > make sure Alpha port solid before you commit to it)e > E > OpenBSD focus - security (acomplished by continuous code audits andtA > secure default install).  Also quite portable and runs on Alphar > I > NetBSD focus - portability (go to www.netbsd.org and check out the listr > of supported platforms). > H > All the BSDs are good performers and hold up well under load.  Any one > will make a fine server.  ? Thanks, good overview. Hope that the other variants better liveo* up to their respective goals than FreeBSD.  J > > > VMS clean in every of the three important dimensions? What are these > > > three dimensions?A > > >E > >l0 > > 1.Kernel - how things internally are handled > C > There are many implementations of Unix, varies by implementation.e  ? How do they excel? The internal VMS architecture excels in many 	 respects.s   > 5 > > 2.API - how the OS is presented to the programmer  > C > This was my main point.  Unix has a short, simple, effective API.h  > Can't buy that. Short, simple, effective are prerequisites but> have to be measured against what they provide and this has not= much to do with quality of design. As I pointed out later theh> UNIX design goals were rather strange. Simplicity according to@ the tasks that have to be accomplished is a virtue. But lowering? the goalpost instead of thinking about what almost every useful > program will need is a no-brainer. The UNIX people did exactly that.t   > 3 > > 3.CLI - how the OS is presented to the sysadmine > I > This has more to do with the shell, scripts, and directory organiztion.r< > Varies from Unix to Unix although generally quite similar.  > Unfortunately yes. This is the clearest indication for lack of> planning and design. The fact that there are so many different; and incompatible shells out there is an implicit proof thato? there was/is a piece missing and no satisfying solution around.   N > > > Unix on the other hand has much of its roots in academia where a lack ofG > > > market pressure gives more time for planning, debate, and design.i > > >l > > C > > If it were so! In fact UNIX didn't really start in academia. To C > > my knowledge UNIX was the outcome of a project that was startedhD > > by AT&T, GE and the MIT. The DoD's ARPA was also a major funder.@ > > To make is short: MULTICS was the goal, never finished, AT&TA > > canceled the project. Ken Thompson, Dennis Ritchie and JosephN< > > Ossanna, all AT&T, had nothing to do, so they decided toC > > implement a game called Space Travel on a PDP-7 which was theremC > > and unused. They realised that it would be faster to develop anGB > > OS for the PDP-7 suitable for their task than cross-developingB > > the game with one of the other computers around where the gameC > > was already running. The goals of UNIX were minimalism in every * > > respect: kernel hardware requirements, > F > Humble beginnings heh?  Let's make an OS for Space Travel.  The name> > Unix is a pun on Multics.  Multics was designed to be an allI > encompassing do all OS.  Perhaps Thompson, Ritchie, and Ossanna came tonG > the conclusion that Multics was overkill and a simple system could doC# > everything the complex one could.   @ Yes, MULTiplexed Information and Computer System was a tank. But> to be honest I checked the story and it is true. No "perhaps".> No thinking ahead. No design. And the decision to stop Multics= was made by management not by these three (young) programmersr  who would had loved to continue.   > > number and completenessaA > > of services, documentation, guidelines a.s.o.. And the sourcet( > > wasn't by no means freely available. > I > Because of the simple minimalist approach the source code from AT&T wasaG > used in computer science courses at universities for study.  A lot of H > work was contributed back, much of it from Berkley (therefore the name% > BSD, Berkley software distribution)e  ; According to publicly available sources the source text wash> studied because there wasn't any useful documentation and this= was the only way to find out how to access and use the systemc@ services or what options a certain system program might have and8 how they would influence the behaviour of the program in@ question. Sorry, but according to the publicly available sources? the source text was even so incomprehensible that a lot of works= was dublicated because no one was able to understand what wash= going on. There was virtually no comment in the source texts.m= Some of the comments were "you are not expected to understand ; this" or "this function is recursive" as if recursion was a-> difficult to understand concept. Instead of having appropriate@ comments the variable names were cryptic like p, pp and ppp used7 for a lot of different purposes in one single function.d  gI > Later when AT&T realized what they had did trouble start.  Lawyers camenD > in did a lot of damage.  Enter Linus and his kernel - written fromD > scratch and free of AT&T code with no legal cloud hanging over it.   True.c  ? > > Honestly, I don't see any design on the sysadmin dimension.yB > > There is no consistent option and parameter handling. There is& > > even no system in naming commands. > H > I'm a programmer so I tend to place more weight on the API.  It is notG > like every one is different though - ls, ps, find, etc are all pretty  > standard.   @ Obviously I wasn't clear at all, so you completely misunderstood@ what I wanted to explain but it might not easy to understand for< a person who is accustomed to UNIX. I claimed that different< standard commands should have consistent options in that the8 same aspect should be controled with the same option for? different command. An example with Linux system commands. ls -r3= shows the directory reverse sorted. ls -R shows the directorye> and all its subdirectories and all their subdirectories and so@ on. cp -r and cp -R both copy one directory and all nested files= and directories in there whereas the man page gives differentc? descriptions for both options. rm -r and rm -R do the same, butt8 mv -r and mv -R don't exist (e.g. mv -R *.map /archiv/).? ls, ps and escpecially find have a lot of different options and[9 different meaning of the "same" options on different UNIXe? variants with ps and find beeing of the worse. Examples needed?h  = And with respect to naming commands: cp, ls, mv, rm but sort,e8 grep, cat, man, biff; where is the logic or consistency?  C > > I'm really interested in understanding the "design" of UNIX and @ > > find a way to like the API. I tried it for several years butD > > without success. And I'm close to give up. If you could show howD > > I can learn to like UNIX this would be really great (no irony!). > B > I'll try to give a good example here.  Suppose you have two fileI > descriptors (fds which is an open serial port, and fdf which is an openeH > file)  The program xyz takes input on standard in and writes it outputF > to standard out.  You want to run xyz with fds for input and fdf for= > output.  You also want to run it at boosted priority level.3 > H > In VMS you will have one call to the OS the create process xyz.  ThereH > is a big structure you pass containing information about input, outputI > handles (for redirection), priority, priviledge, etc.  Everything aboutm3 > that process must have a place in that structure.r > ( > Unix is completely different, example: > ) > x = fork() /* create copy of process */u
 > if (x == 0)o > {I > /* we are child */ > /* make fds standard in */ > dup2(fds, 0);  > /* make fdf standard out */c > dup2(fdf, 1);$  > /* adjust my priority to -5 */# > setpriority(PRIO_PROCESS, 0, -5);e. > /* replace current process image with xyz */ > execl("xyz", NULL);e > /* never get here */ > }e > else { > /* we are parent */t > }2 > I > So I've replaced a complex system call (in VMS) which takes a structureeB > with a few simple ones in Unix (none of which take a structure). > I > That is what I like about unix.  Simplicity and flexibility.  It's kindiJ > of like a lego or erector set.  The basic components are very simple butN > you can put them together any way you want to achieve exactly what you want.  ? Honestly I don't think that having multiple calls is simpler oro@ safer than decorating a structure and executing only one call. I; personally like structures much more but this is probably aV@ matter of taste. But distributing calls which should be executed9 as a bundle is obviously less safe. But what is much more.< important: the VMS interface is much richer. The UNIX API is@ like a lego set for three year old children, whereas the VMS API5 is like a Fisher Technik set for 15 year old teenies.2  C > > Very interesting thought. I never looked at it that way. ThoughvD > > it seems that this description will fit for many other OSs. If ID > > look close than it seems to me that the differences are strongerB > > than the similarities. Could we agree that VMS to UNIX is like. > > MacOS to M$DOS? VMS establishes standards, > - > A quote from the book Linux Device Drivers:p > J > .... role of a device driver is providing mechanisms, not policies.  TheJ > distinction between mechanism and policy is one of the best ideas behindJ > the Unix design.  Most programming problems can be split into two parts:E > "what needs to be done" (the mechanism) and "how can the program beeI > used" (the policy).  If the two issues are addressed by different partssH > of the program, or even by different programs altogether, the softwareE > package is much easier to develop and to adapt to particular needs.n > + > A quote from Beginning Linux Programming:  > I > You can't anticipate exactly how ingenious users will use your program.wG > .... Never assume that you know everything the user might want to do.t > C > Again, philosophy here is similar to programming, give the user awI > collection of simple programs and let him pipe them together to do whath > he wants.   @ These are two very good and convincing quotes but we need to see how it is accomplished.c   > > a rich set of full? > > blown services and guidelines to achieve uniform and systemnB > > adapted behaviour of the apps written according to these rules: > > and using these services, to prevent each developer to< > > re-implement the same things again and again like record# > > handling and parameter parsing.d > F > Parameter parsing functions are in the C library.  On the whole codeF > reuse is good.  Suppose you want to see all unique processes running > excluding bash, output sortedh  ? Misunderstanding. I'm not talking about code reuse with respects: of parameter parsing. I'm talking of implementation effort: because each program has to check its parameters by itself= whereas with VMS this is done generically and the program has1@ only to use the set of supplied values. Did you ever implement a; program under VMS which had to use command line parameters?e   > $ > ps -A | sort | unique | grep -v sh > I > As you can see a few simple commands piped together.  ps (process show)nJ > does not need to have sort logic programmed in, etc.  ps need not have aF > million options trying to anticipate everything the user could want.# > Connect them up however you want.   = Sorry, but I completely disagree. First you can't combine alld< the programms using pipes because there are problems how the> parameters are interpreted. E.g. if you pipe to grep the input; is interpreted as a file but you can't pipe to the pattern./: Another example? "find . -name '*.text' | grep foo" simply> doesn't work. Why? More examples needed? I have the impression= that pipeing is a patch work approach which works most of theV< time but not all the time and it's not well thought through.@ Sorting is anyway one of the worst examples for pipes. Sort will@ wait until it got all the data and as it doesn't know which data? is to sort it is difficult to specify how to sort. This reveals ; a more fundamental lack of system architecture of UNIX: the ? exchange format is a stream of unstructured characters. This iso8 a complete waste of resources and knowledge. The user is? responsible for telling the programs how they should understands what's feeded.> Code reuse is at least as good if each program uses the system9 calls for that. E.g. sorting can be done in the DIRECTORYN: program and in the SORT program both using the same system= calls. Why does "ls" offer sorting options - couldn't that ber@ done with pipeing and sort (should be srt!)? Why does "ls" offer< output format selection selection options - couldn't that be@ done with pipeing and grep (should be grp!)? Wouldn't it be much? more efficient if the "ls" like the VMS DIRECTORY command couldu? all sorts of sorting and selecting because it knows best how tom interpret the data?k   " >  If I compare this with the UNIX9 > > approach then I think if there was design that it haseC > > disappeared over time or it was an anti-design (the decision torA > > be without any design) or the design was intentionally hiddenh > > and remained in that state.  > J > Let's call it anti policy.  The design is to be flexible through lack of	 > policy.w  > Good reply but it's not clear until now how this pays back and how VMS breakes theses rules?t= It is common knowledge that style guidelines are an essentialc@ concept even for programming languages where you really could do: without. I don't see how a consistent and efficient system/ behaviour could be achieved without guidelines.   E > > > I'm just curious, how much have your programmed for Unix vs VMS = > > > (sometimes familiarity with one system skews judgement)i > > >i
 > > > Fran > > >h > > > Sent via Deja.come > > > http://www.deja.com/ > >e? > > I programmed about 2000 h on VMS using MODULA-2 and heavilyeB > > exploiting the RMS services (we didn't use a database) and the? > > I/O features to communicate with exernal devices like SPSs. D > > I programmed about 500 h on different UNIX variants using C, C++D > > and Java for very different tasks one of them recently beeing to? > > implement a servlet based management system. But I did moreR* > > system management on UNIX than on VMS. > >eB > > And yes, familiarity with one system skews judgement: I hadn'tB > > believed that I scorned UNIX so much if I hadn't gotten really > > acquainted with. > K > What I consider to be 'clean' about Unix is the fact that you can achieveIJ > so much with such a simple system.  Just a guess but I'll bet people whoH > like Unix like building things from kits, etc.  I guess a lot of it is > just personal taste. >  > Fran >  > Sent via Deja.com  > http://www.deja.com/  @ While I completely agree with your conclusion I'm sure because I: have facts and numbers that this is not sufficient. Like a: typical C programmer who isn't able to do real programming9 because he or she is subconsciously absorbed by masteringe@ difficulties which are a waste of time to be mastered by a human? beeing and by that are unable to achieve a level of abstractionl9 and overview which is necessary to do real programming asa? opposed to dumb coding, it is this attitude of putting together > simple but insufficient parts which creates this vast majority= of unusable, unmaintainable, inefficient and architectureless > programs. This C programmer attitude is exactly what makes the@ UNIX programmers happy by re-inventing the wheel again and again@ instead of trying to reach a higher level of programming. Why is> parsing an unstructured text file a pleasure (UNIX approach to; configuration files) while fetching a value from structuredf: storage is a pain (VMS approach with RMS key files)? Is it@ because one has to think a little bit more before coding starts?" Or is it simply lack of knowlegde?   ------------------------------  % Date: Thu, 25 Jan 2001 11:30:25 +0000i) From: Christof Brass <brass@infopuls.com>r@ Subject: Re: VMS vs UNIX (was: Linux worm and RedHat 7.0 broken), Message-ID: <3A700E51.E309E3FE@infopuls.com>   Arne Vajhj wrote: >  > Christof Brass wrote:e? > > I clearly remember that he said in an interview that he wasl7 > > pissed off that the UNIX sources weren't available.t > 4 > Unix sources being available is not a new concept. > . > You have always been able to get BSD source. > , > You have also been able to get VMS source. > 5 > SUN either has made Solaris source available or are  > planning to do.  >  > The important issues are:c >   - what do the sources cost* >   - what are you allowed to do with them > 9 > Linux adopted the GNU/GPL/FSF/Stallman ideas going back. > to the 1980's. >  > Arnes  ; Yes, of course, he was pissed of that the AT&T UNIX sourcesf weren't freely available.e Does this matter for the topic?i   ------------------------------  % Date: Thu, 25 Jan 2001 08:39:55 -0500v0 From: David Beatty <David.Beatty.NOSPAM@sas.com>@ Subject: Re: VMS vs UNIX (was: Linux worm and RedHat 7.0 broken)2 Message-ID: <uytwOjez5SP9uQThs27NeiZL0arm@4ax.com>  
 <BIG SNIP>   >i: >True but not helpful in our situation. I doubt that these@ >netcraft numbers are reliable or useful in any respect. In fact; >I think that there are about 1'000'000 times as many Linux @ >system in usage than FreeBSD systems. What is the point of what
 >you said? >i  9 Considering we have at least 100 systems running FreeBSD,h9 I doubt there are over 100,000,000 systems running Linux.   
 <BIG SNIP>   ------------------------------  % Date: Thu, 25 Jan 2001 15:34:43 +0000.) From: Christof Brass <brass@infopuls.com>o@ Subject: Re: VMS vs UNIX (was: Linux worm and RedHat 7.0 broken), Message-ID: <3A704793.855B090F@infopuls.com>   David Beatty wrote:d >  > <BIG SNIP> >  > >o< > >True but not helpful in our situation. I doubt that theseB > >netcraft numbers are reliable or useful in any respect. In fact= > >I think that there are about 1'000'000 times as many LinuxcB > >system in usage than FreeBSD systems. What is the point of what > >you said? > >  > ; > Considering we have at least 100 systems running FreeBSD,r; > I doubt there are over 100,000,000 systems running Linux.y >  > <BIG SNIP>  < Yep, drop at least two zeros. The point made is still valid:= there are probably x times more Linux systems in use. What is  your guess?    ------------------------------  % Date: Thu, 25 Jan 2001 11:12:20 -0500p2 From: rdeininger@mindspring.com (Robert Deininger)@ Subject: Re: VMS vs UNIX (was: Linux worm and RedHat 7.0 broken)L Message-ID: <rdeininger-2501011112210001@user-2iveal5.dialup.mindspring.com>  E In article <94n6i5$hi1$1@nnrp1.deja.com>, sabolich@my-deja.com wrote:m   My news server seems to have discarded this, so I'm reposting.  Slightly out of time order, but still replying to the right post, I hope.s  B > I'll try to give a good example here.  Suppose you have two fileI > descriptors (fds which is an open serial port, and fdf which is an open H > file)  The program xyz takes input on standard in and writes it outputF > to standard out.  You want to run xyz with fds for input and fdf for= > output.  You also want to run it at boosted priority level.e   Ok.n  H > In VMS you will have one call to the OS the create process xyz.  ThereH > is a big structure you pass containing information about input, outputI > handles (for redirection), priority, priviledge, etc.  Everything about 3 > that process must have a place in that structure.R  9Well, no.  I guess you are describing SYS$CREPRC, which is the kitchen-sink method of creating a process.  I count 13 arguments to this routine, which sure is complex.  Only two of them are "structures", IIRC. But they are ALL optional.  (Perhaps not all at the same time.)  So you are overstating the complexity.i  sTo compare the API complexity to create a process, you ought to note that a VMS process has a lot more to it than a unix one.  The extra features have to be specified (or defaulted) somehow.  VMS supports very fine granularity in both privleges and quotas, for example.  Your little example doesn't use such features, so we'll let VMS do it's default thing.  That's easy.a   But this is the wrong way to make this process, probably.  LIB$SPAWN might be more suitable.  Same number of arguments, all of them optional, none of them a structure.i   One specifies input, one specifies output, one specifies command to execute.  Some others help with interprocess communication, which we don't seem to need here.s   Since we don't clone processes in VMS, the new process will need to set it's own priority, if we use lib$spawn.  sys$creprc lets the creator specify priority.  J Depending on the context, we might not need another process at all in VMS.  ( > Unix is completely different, example: > ) > x = fork() /* create copy of process */ 
 > if (x == 0)i > {l > /* we are child */ > /* make fds standard in */ > dup2(fds, 0);e > /* make fdf standard out */h > dup2(fdf, 1);t  > /* adjust my priority to -5 */# > setpriority(PRIO_PROCESS, 0, -5);p. > /* replace current process image with xyz */ > execl("xyz", NULL);t > /* never get here */ > }o > else { > /* we are parent */o > }n    I > So I've replaced a complex system call (in VMS) which takes a structure B > with a few simple ones in Unix (none of which take a structure).  m So we might do something like this in VMS (Fortran, just to remind folks that VMS is very language-flexible):   +       status = lib$spawn('xyz','fds','fdf')n   We'll have to move the priority-change into xyz if we do it this way, or wrap xyz in a little script to change the priority if we don't want to change xyz.w   I don't think your example is simpler than mine.  Yours has comments; that's nice.  Maybe I should have done mine in Ada, with named parameters.  Then I wouldn't need any comments.  - > A quote from the book Linux Device Drivers:t > J > .... role of a device driver is providing mechanisms, not policies.  TheJ > distinction between mechanism and policy is one of the best ideas behindJ > the Unix design.  Most programming problems can be split into two parts:E > "what needs to be done" (the mechanism) and "how can the program besI > used" (the policy).  If the two issues are addressed by different partstH > of the program, or even by different programs altogether, the softwareE > package is much easier to develop and to adapt to particular needs.e   Though the terminology is a bit strange to me, I think it describes VMS device drivers as well.  They provide a set of functions that (hopefully) match the capabilities of the device.   Programmers use those functions as needed.  ,"What needs to be done" inside the driver is to interact nicely with the rest of the OS, while getting all the performance and functionality out of the device.  And we must often program around little problems in the device itself.  All of that can get non-trivial, but the rules are straightforward.     F > Parameter parsing functions are in the C library.  On the whole code > reuse is good.     Partly.  But there is little uniformity in the switches from one program to another.  Perhaps a little more OS support would have encouraged uniformity, to make the user's life easier.  Or maybe not.  The unix culture does not value such things as much as the VMS (or Mac) community does.  6 > Suppose you want to see all unique processes running > excluding bash, output sortedt > $ > ps -A | sort | unique | grep -v sh > I > As you can see a few simple commands piped together.  ps (process show)xJ > does not need to have sort logic programmed in, etc.  ps need not have aF > million options trying to anticipate everything the user could want.# > Connect them up however you want.e   Yes, pipes are sometimes useful.  That's why VMS stole them.  But they came late, so a lot of "pipe" stuff is typically done in other ways.I   In VMS we wouldn't program sort logic into ps.  We'd use the callable sort facility, which takes a few lines of effort for a typical simple case.-  /J > Let's call it anti policy.  The design is to be flexible through lack of	 > policy.    Too much flexibility makes it harder for the end user.  Too little makes it harder for the programmer.  Nothing in VMS stops you from doing it your own way.  H > What I consider to be 'clean' about Unix the fact that you can achieveJ > so much with such a simple system.  Just a guess but I'll bet people whoH > like Unix like building things from kits, etc.  I guess a lot of it is > just personal taste.   You can achieve a lot easily in VMS as well.  Not exactly the same things.  It helps to not think in unix when you do it.  Maybe VMS programmers make a bloody mess of things when they try unix..   -- s Robert Deininger rdeininger@mindspring.comp   ------------------------------  % Date: Thu, 25 Jan 2001 11:44:31 -0500d0 From: David Beatty <David.Beatty.NOSPAM@sas.com>@ Subject: Re: VMS vs UNIX (was: Linux worm and RedHat 7.0 broken)2 Message-ID: <PVdwOpqXwDwH=5gAYLdcnex+iF0Q@4ax.com>  = I'm not involved in either the FreeBSD or Linux community, sod7 my guess would be worth less than the electrons used tod5 transmit this message.  However, your point is valid.p  2 On Thu, 25 Jan 2001 15:34:43 +0000, Christof Brass <brass@infopuls.com> wrote:    >David Beatty wrote: >>  
 >> <BIG SNIP>r >> . >> >= >> >True but not helpful in our situation. I doubt that theseaC >> >netcraft numbers are reliable or useful in any respect. In factp> >> >I think that there are about 1'000'000 times as many LinuxC >> >system in usage than FreeBSD systems. What is the point of whatf
 >> >you said?  >> > >> h< >> Considering we have at least 100 systems running FreeBSD,< >> I doubt there are over 100,000,000 systems running Linux. >> e
 >> <BIG SNIP>  >d= >Yep, drop at least two zeros. The point made is still valid:a> >there are probably x times more Linux systems in use. What is >your guess?   ------------------------------  % Date: Thu, 25 Jan 2001 12:52:01 -0500 5 From: "Fred Kleinsorge" <kleinsorge@star.zko.dec.com>o@ Subject: Re: VMS vs UNIX (was: Linux worm and RedHat 7.0 broken), Message-ID: <94pp40$809p$1@lead.zk3.dec.com>  ! I'm not sure who this troll is...   F sabolich@my-deja.com wrote in message <94n6i5$hi1$1@nnrp1.deja.com>.../ >> 1.Kernel - how things internally are handled  >eB >There are many implementations of Unix, varies by implementation. >g    I That is: non-uniform implemenations and quality from really bad to really  good.w    4 >> 2.API - how the OS is presented to the programmer > B >This was my main point.  Unix has a short, simple, effective API. >     G Eh?  It's API is a random assortment of routines that have been cobbled  togetherJ together over the years.  Some subset of these have been "standardized" at variousnJ times (POSIX, UNIX98), but each vendor has it's own peculiar differences -	 even witht routines of the same name.  2 >> 3.CLI - how the OS is presented to the sysadmin >nH >This has more to do with the shell, scripts, and directory organiztion.< > Varies from Unix to Unix although generally quite similar. >>    E UNIX has a poorly designed, ad-hoc UI.  It is short, cryptic, and not  designedD and not really even evolved.  If anything it ensures that nobody but computer geeks will ever make it mainstream.a      	      SNIPs    ' >Unix is completely different, example:h > ( >x = fork() /* create copy of process */ >if (x == 0) >{ >/* we are child */o >/* make fds standard in */c >dup2(fds, 0); >/* make fdf standard out */ >dup2(fdf, 1); >/* adjust my priority to -5 */ " >setpriority(PRIO_PROCESS, 0, -5);- >/* replace current process image with xyz */i >execl("xyz", NULL); >/* never get here */g >} >else {i >/* we are parent */ >} >sH >So I've replaced a complex system call (in VMS) which takes a structureA >with a few simple ones in Unix (none of which take a structure).e >xH >That is what I like about unix.  Simplicity and flexibility.  It's kindI >of like a lego or erector set.  The basic components are very simple butcG >you can put them together any way you want to achieve exactly what youh want.h >g    / Do you know what the last S in KISS stands for?i  K Nonetheless, while the code sequence above is "fairly standard" on UNIX, ita doesn'toI mean it is simple or obvious, unless of course you are a UNIX programmer.l
 It is justD *different* and it reflects the design choices made by the operating systems.   ------------------------------   Date: 25 Jan 2001 09:32 CSTe' From: carl@gerg.tamu.edu (Carl Perkins)g& Subject: Re: VMS writable CD on Win2K?- Message-ID: <25JAN200109320281@gerg.tamu.edu>h  ! tim.llewellyn@bbc.co.uk writes...hI }Yeah, I want one! However DVD-R is alledgedly almost shipping (in hi endg }macs).uI }Can burn CDR's and DVD-R's (not the same as DVD-RAM). Mind you they seem  } ! }to be only 1x or 2x write speed.h } G }I sure hope those 100GB rewriteables, when they arrive, can be written  }at anG }decent rate, otherwise you're gonna NEED and VMS system to have enoughu9 }guaranteed uptime to write that 100 gigs at 100k/sec :-)n }  } --7 }Tim Llewellyn, OpenVMS Infrastructure, Remarcs Projecth  F A 1x DVD is not the same speed as a 1x CD - this is obviously the caseH as a DVD at 1x can stream high quality video plus CD quality (or better)D audio. I don't recall the actual figure, but it is at least 10 timesG faster, so while a 1x CR-R writes at about 160k/sec (the raw speed is atJ bit faster, but I don't know how much overhead the filesystem has comapredG to raw audio data - 44.1 samples/sec * 2 bytes/sample and 2 channels ishG 176.4KB/sec) which would make DVD at 1x be at least 1.6MB/sec. Anythingt- newer than a DVD is likely to be even faster.   F BTW, a 72x CD drive (which is, I think, the fastest I've ever seen) isG reading data at over 11MB/sec. That's pretty fast. In fact I think it'siH significantly faster than any of the hard drives in my VMScluster at the3 moment, the fastest of which is currently an RZ29B.    --- Carl   ------------------------------  % Date: Thu, 25 Jan 2001 16:22:07 +0000 - From: Tim Llewellyn <tim.llewellyn@bbc.co.uk>a& Subject: Re: VMS writable CD on Win2K?) Message-ID: <3A7052AF.179A43A2@bbc.co.uk>    Carl Perkins wrote:a  I >  A 1x DVD is not the same speed as a 1x CD - this is obviously the case J > as a DVD at 1x can stream high quality video plus CD quality (or better)F > audio. I don't recall the actual figure, but it is at least 10 timesI > faster, so while a 1x CR-R writes at about 160k/sec (the raw speed is a L > bit faster, but I don't know how much overhead the filesystem has comapredI > to raw audio data - 44.1 samples/sec * 2 bytes/sample and 2 channels iseI > 176.4KB/sec) which would make DVD at 1x be at least 1.6MB/sec. Anythingn/ > newer than a DVD is likely to be even faster.i >s  L OK, I did engage brain later on after posting and realized that 1x DVD wouldN be faster than 1x CD. 1.6 meg/sec is better than 8* CDR but not as good as the   16* CDR's that are out now.o   > H > BTW, a 72x CD drive (which is, I think, the fastest I've ever seen) isI > reading data at over 11MB/sec. That's pretty fast. In fact I think it's J > significantly faster than any of the hard drives in my VMScluster at the5 > moment, the fastest of which is currently an RZ29B.s >i  @ Wow, you got 4 gig drives ! You must be almost as overdue for an3 upgrade as me (I bet you ain' getting R3'd though).v  D I get over 20 MByte/sec peak, approx 15MByte/sec sustained from 72008 ATA 66 and ATA 100 drives currently, but not on VMS :-(. --6 Tim Llewellyn, OpenVMS Infrastructure, Remarcs Project0 MedAS at the BBC, Whiteladies Road, Bristol, UK.A Email tim.llewellyn@bbc.co.uk. Home tim.llewellyn@cableinet.co.ukt  A I speak for myself only and my views in no way represent those of  MedAS or the BBC.o   ------------------------------  % Date: Thu, 25 Jan 2001 18:19:37 +0000f) From: Christof Brass <brass@infopuls.com>tR Subject: Warning: No type converter registered for 'String' to 'Pixmap'conversion., Message-ID: <3A706E39.8449AFC9@infopuls.com>  6 Warning: No type converter registered for '%s' to '%s' conversion.   > I receive these warnings when XEmacs opens a dialog box to let me choose a file.i  > Env: DECwindows Motif Version 1.2-3 for OpenVMS 6.2 is used as front-end (X display server).t@ XEmacs is started from within an xterm running on a remote Linux; box (kernel 2.2) and also displayed on the local DECwindowse= Motif X display server. The warnings appear in the xterm fromf9 what XEmacs has been started. Otherwise XEmacs is workingc= flawlessly, especially the dialog box for chosing a file also 9 works flawlessly. It is a little anoying to get the xterm > cluttered by these warnings. They appear every time I open the9 dialog box for chosing a file (the system obviously isn'tu2 learning that this functionality isn't available).  < Is there anybody around who can direct me where to search or; even suggest a solution? Is it the XEmacs config? Is it thee( DECwindows config? Is it something else?   TIAT   ------------------------------  % Date: Thu, 25 Jan 2001 12:47:40 -0500c- From: "Peter Weaver" <peter.weaver@stelco.ca>h* Subject: Where is the latest MIME for AXP?4 Message-ID: <AFZb6.123442$Z2.1530569@nnrp1.uunet.ca>  E While testing DSN 3.0 (it still has some of the problems that were inPE DSN 2.0, but I still prefer DSN 1.2) I stumbled across VAXMIME02_072.nD It was posted 19-OCT-2000 and it replaced VAXMIME01_072. Last summerA Paul Mosteika posted several messages saying that the latest MIME'C fixed a lot of problems but the version he had was not available to E the general public. I figured most of the fixes Paul mentioned should0D be in this release and decided to try it on our test ES40. But I can not find AXPMIMExx_072.S  , I have searched using DSN and I have gone toF http://ftp.service.digital.com/patches/public/vms/axp/v7.2-1/ but I doC not spot it. Does anyone out there know where I can find the latest)	 AXP MIME?D     --   RULES OF THE AIR   -----------------y;  #18. If all you can see out of the window is ground that's =       going round and round and all you can hear is commotione>       coming from the passenger compartment, things are not at       all as they should be.   ------------------------------  # Date: Thu, 25 Jan 2001 13:50:41 GMT  From: graculuss@my-deja.com * Subject: Re: Why is the tape write locked?) Message-ID: <94pave$bev$1@nnrp1.deja.com>1   Here's a story:r  B When I was an operator (a long time ago) we had all our productionE backups fall over one night shift complaining that the tape drive wasm write locked. We tried:g  0 Checked that the procedures hadn't been changed.+ Checked the write rings -they were all O.K.g3 Taped the sensor switch on the drive down - no joy.S Got field service in and:n$ Replaced the sensor switch - no joy.E Switched a known working drive from another system - no joy, but thisc gave us a BIG clue..  D Then we found that some twat had put a symbol in SYLOGIN.COM so that1 MOUNT did a MOUNT/NOWRITE and hadn't told anyone.   E Having spent half of our nightshift trying to sort this one out theretG was almost a fight when the operator who did the change came on for thes following morning shift.   <SNIP>  B > Try SHOW SYMBOL MOUNT. Maybe there's a hidden /NOWRITE in there. > F > Add a SHOW DEVICE C2 /FULL command before and after the SORT commandB > and see what its output says about C2 being write-locked or not. >o     Sent via Deja.com  http://www.deja.com/   ------------------------------  % Date: Thu, 25 Jan 2001 11:08:50 -0700u+ From: "Barry Treahy, Jr." <treahy@mmaz.com>pR Subject: Why still VAX? Was: Re: Is Compaq dropping support for vms on vaxplatform( Message-ID: <3A706BB2.EA304BF5@mmaz.com>   JF Mezei wrote:   K > If they had done such a great job, they wouldn't have been demand for VAXaM > hardware for 8 years after the introduction of the Alpha. Had they providedOK > VEST with every alpha sold, and made it work on any executable (includingnN > software packages which Digital had decided not to port to Alpha), I suspectB > that the migration to Alpha would have been much more succesful.  O Absolutely!!!  Additionally, many of the layered products from the VAX era thatiG run with legacy softwares are no longer owned by Compaq which makes thep4 proposition of relicensing/rehosting a costly event.  R Charon-VAX on an Alpha may never produce true native hardware performance results,R but I have to believe that it does demonstrate that Digital/Compaq could have done a better job with VEST.a  Q I, for one, would be on an Alpha today if my existing applications would run withe' better performances at a lessor cost...t   Barry.   --  ? Barry Treahy, Jr  *  Midwest Microwave  *  Vice President & CIOo  A E-mail: Treahy@mmaz.com * Phone: 480/314-1320 * FAX: 480/661-7028s   ------------------------------    Date: 25 Jan 2001 13:40:21 -0500/ From: jordan@lisa.gemair.com (Jordan Henderson)n2 Subject: Window for Alpha success still seems open* Message-ID: <94prul$f25$1@lisa.gemair.com>   Take a look at this article:  %  http://www.tweakers.net/reviews/204/i  > It seems that the Itanium is benchmarking unbelievably poorly,; especially when executing 32 bit code.  A 100Mhz Pentium I  , outbenches a 1.5 Ghz Itanium in some cases!?  @ As the Alpha still seems to be improving and the IA architecture> may be hitting a serious bump in the road going to 64 bits, it@ might be that we're looking for the window for success for Alpha to be open a few more years.  = Hey, it's possible that IA-64 could even be a godsend to the c= Alpha.  IT organizations might commit to 64 bit architecturesM> for long term strategic reasons (now that IA-64 is here/almost? here) and then recoil from the horror of having to take such a a( performance hit with the Intel products.   Anything's possible...     -Jordan Hendersont jordan@greenapple.com    ------------------------------  % Date: Wed, 24 Jan 2001 23:30:59 -0800h) From: Wayne Holland <wholland@tscnet.com> ! Subject: Re: Wizard Disappears!!!aO Message-ID: <390B48DB93D0B53B.AB010BC847E71E0E.00327F7A567A85C6@lp.airnews.net>e   jlsue wrote: > 3 > On Wed, 24 Jan 2001 14:13:15 -0800, Wayne Hollandd > <wholland@tscnet.com> wrote: > I > >     Boy, are you right!  I'm running Solaris now and can't even get a H > >straight answer out of the docs, faqs, or whatever on getting a local$ > >printer to work thru ghostscript!= > >I have to fight the O/S more than I can get any work done.h > E > Hey Andrew, are you reading this?  By your calculating methods, SunyF > can't say that their systems, in general, are easy and reliable.  We0 > have evidence right here that that isn't true.  9 Oh, by the way...  I was running Caldera's OpenLinux 2.4.iD They finally got their act together so that when everything installsE you can get on the internet easily, (without having to read a 18 pagesA document from Kempston.org for Solaris).  It even has an internal  databaseG for common isps so that all you have to do is provide a phone number tod dialF into and thats that.  Also provides for OpenGl (Mesa) that Solaris x86 does nA not have, or that I couldn't find anywhere.  I don't need it tho. . It was fairly easy... too easy and got boring.H It had IBM's Jikes onboard as well (Java) that converted to native code.  8 But If I could justify another computer I'd get OpenVMS!   ------------------------------    Date: 26 Jan 2001 01:13:31 +0800, From: Paul Repacholi <prep@prep.synonet.com>0 Subject: Re: Writes flushed to Disk on Dismount?- Message-ID: <87wvbj4ksk.fsf@prep.synonet.com>c  , young_r@eisner.decus.org (Rob Young) writes:  Q > In article <3A6FA83E.88F5751E@tmisnet.com>, Koloth <koloth@tmisnet.com> writes:uU > > I would not count on it.  I think you have to wait for the Cache timeout (default  > > is 10 seconds) > > on the controller. > >  > B > 	But I want to count on it.  Here's the scenario... as you point< > 	out CACHE_FLUSH_TIMER is default of 10 seconds.  But what? > 	about the poor sap that has it set to 120 seconds?  He makese< > 	sure all file streams are closed (or they are anyhow) and: > 	dismounts a drive.  If writeback cache is turned on, is= > 	there a mechanism that signals the controller to flush thes9 > 	cache?  If not, where is the big doc warning that sayseA > 	"don't do that" or "beware of?"  Same question ... more stuff.  > 	 > 				Rob   F Yes, this is judt the sort of issue that really scares me with all theG SAN noise. The dismount sending the drive unit-offline REQUIRES all theuI data on the disk before the offline completes. Foo knows what will happenc< out there in happy multi vendor, standard de jour, SAN land.     --  < Paul Repacholi                               1 Crescent Rd.,7 +61 (08) 9257-1001                           Kalamunda. @                                              West Australia 6076. Raw, Cooked or Well-done, it's all half baked.   ------------------------------    Date: 25 Jan 2001 13:32:53 -0500* From: young_r@eisner.decus.org (Rob Young)0 Subject: Re: Writes flushed to Disk on Dismount?+ Message-ID: <NNRkEzje7ZH8@eisner.decus.org>r  \ In article <87wvbj4ksk.fsf@prep.synonet.com>, Paul Repacholi <prep@prep.synonet.com> writes:. > young_r@eisner.decus.org (Rob Young) writes: > R >> In article <3A6FA83E.88F5751E@tmisnet.com>, Koloth <koloth@tmisnet.com> writes:V >> > I would not count on it.  I think you have to wait for the Cache timeout (default >> > is 10 seconds)d >> > on the controller.  >> > 3 >> 0C >> 	But I want to count on it.  Here's the scenario... as you point:= >> 	out CACHE_FLUSH_TIMER is default of 10 seconds.  But what @ >> 	about the poor sap that has it set to 120 seconds?  He makes= >> 	sure all file streams are closed (or they are anyhow) anda; >> 	dismounts a drive.  If writeback cache is turned on, ish> >> 	there a mechanism that signals the controller to flush the: >> 	cache?  If not, where is the big doc warning that saysB >> 	"don't do that" or "beware of?"  Same question ... more stuff. >> .
 >> 				Rob > H > Yes, this is judt the sort of issue that really scares me with all theI > SAN noise. The dismount sending the drive unit-offline REQUIRES all thesK > data on the disk before the offline completes. Foo knows what will happen > > out there in happy multi vendor, standard de jour, SAN land. >   = 	And ... at the all day SAN session I sat in on at CETS 2000,D> 	the speaker mentioned a top StorageWorks engineer - that most> 	of us are familiar with (by name anyhow) - recommends setting? 	CACHE_FLUSH_TIMER to 65535 seconds "for performance."  So thate@ 	is why I am *assuming* it surely flushes the writes on dismount3 	as many folks are doing shadow splits for backups.   - 	So there is actually a two-fold reason here.    				Rob-   ------------------------------   End of INFO-VAX 2001.050 ************************