1 INFO-VAX	Fri, 21 Jul 2000	Volume 2000 : Issue 405       Contents:3 Re: %PCSI-E-PARAMB, ambiguous location for file ... 3 Re: %PCSI-E-PARAMB, ambiguous location for file ... > Re: ANALYZE/DISK/REPAIR (was Re: VMS equivalent of UNIX 'ln'?)> Re: ANALYZE/DISK/REPAIR (was Re: VMS equivalent of UNIX 'ln'?)> Re: ANALYZE/DISK/REPAIR (was Re: VMS equivalent of UNIX 'ln'?)> Re: ANALYZE/DISK/REPAIR (was Re: VMS equivalent of UNIX 'ln'?) Re: DECServer 90 to DB-9 PC  Re: DECServer 90 to DB-9 PC  Re: DECServer 90 to DB-9 PC # determining free disk space via DCL ' Re: determining free disk space via DCL ' Re: determining free disk space via DCL # Re: Dialup internet sharing routers  Re: DS10 IDE vs U2W SCSI# How to speed up RMS files reading ? ' Re: How to speed up RMS files reading ? ' Re: How to speed up RMS files reading ? ' Re: How to speed up RMS files reading ? ' Re: How to speed up RMS files reading ? 4 Re: I need a precise definition of "layered product"4 Re: I need a precise definition of "layered product" Re: INFO ON A VT320   Re: Intraserver SCSI Controllers Join the INFe-Investors Network . Re: Limit a UIC group access to a number of 50 re: Linux "clusters"1 Re: Mlucas performance probelm - need expert help 1 Re: Mlucas performance probelm - need expert help 1 Re: Mlucas performance probelm - need expert help 1 Re: Mlucas performance probelm - need expert help  OpenVMS Registry Configuration Printer installation problems $ Re: Question on MX mailer relay file- Re: Quick answer time.... how fast is a TK70? $ Re: Removing <LF><CR> from text file$ Re: Removing <LF><CR> from text file$ Re: Removing <LF><CR> from text file$ Re: Removing <LF><CR> from text file RE: Sun may GPL StarOffice RE: Sun may GPL StarOffice RE: Sun may GPL StarOffice Re: SYSUAF.DAT Re: SYSUAF.DAT RE: SYSUAF.DAT; Re: VMS Com file to write a Form Feed Character --- THANKS!   Re: VMS equivalent of UNIX 'ln'?  Re: VMS equivalent of UNIX 'ln'? Re: Web browser and VMS mail Re: Web browser and VMS mail Re: Web browser and VMS mail3 Re: Where can I get older version of Apache for VMS 3 Re: Where can I get older version of Apache for VMS 3 Re: Where can I get older version of Apache for VMS 3 Re: Where can I get older version of Apache for VMS 1 Re: [Fwd: FW: Wish List:  I/O vs. CPU Priorities] 1 Re: [Fwd: FW: Wish List:  I/O vs. CPU Priorities]   F ----------------------------------------------------------------------  % Date: Fri, 21 Jul 2000 20:10:43 +0010 % From: paddy.o'brien@zzz.tg.nsw.gov.au < Subject: Re: %PCSI-E-PARAMB, ambiguous location for file ...5 Message-ID: <01JS1I1VV3QQ0021RR@tgmail.tg.nsw.gov.au>    Larry Kilgallen wrote:Q >In article <009ED5A0.CCD29AF1@SendSpamHere.ORG>, system@SendSpamHere.ORG (Brian    Schenkenberger, VAXman-) writes:$ >> Where are these error documented? >  >Not in HELP/MESSAGE PARAMB .  >  > What does this error mean? > K >It means you should complain to DEC about the shortcoming in HELP/MESSAGE.  > E >(Side note to bystanders, I know Brian, and he is quite competent at  >complaining to DEC :-).    N It is probably quite obvious that Brian is not averse to complaining :-)  But < also one of our gurus.  I too have had minor correspondense.  K The salient point of your note is what I get the sh*ts about. HELP/MESSAGE  M leaves many flaws in what I expected as help.  Many of the messages are just  M translations of f$fao.  Leaving little to be desired as erudition, i.e. !FAO  $ something within the error response.   Regards, Paddy   Paddy O'Brien, Transmission Development, 
 TransGrid, PO Box A1000, Sydney South,  NSW 2000, Australia    Tel:   +61 2 9284-3063 Fax:   +61 2 9284-3050& Email: paddy.o'brien@zzz.tg.nsw.gov.au  M Either "\'" or "\s" (to escape the apostrophe) seems to work for most people, ; but that little whizz-bang apostrophe gives me little spam.    ------------------------------    Date: 21 Jul 2000 09:00:01 -05009 From: Kilgallen@eisner.decus.org.nospam (Larry Kilgallen) < Subject: Re: %PCSI-E-PARAMB, ambiguous location for file ...+ Message-ID: <C2jmtjmvKhfL@eisner.decus.org>   ] In article <01JS1I1VV3QQ0021RR@tgmail.tg.nsw.gov.au>, paddy.o'brien@zzz.tg.nsw.gov.au writes:   M >                                                               HELP/MESSAGE  O > leaves many flaws in what I expected as help.  Many of the messages are just  O > translations of f$fao.  Leaving little to be desired as erudition, i.e. !FAO  & > something within the error response.  > The point is that in the DEQ scheme of things, HELP/MESSAGE is= documentation.  If you complain (through proper channels, not ? here) they will fix it.  There is no way to fix it "in general" ? because they need to know specific point items that are broken.   B Now it may be that since Brian's was a PCSI HELP/MESSAGE complaintD and the text was not there at all some additional "not there at all"D problems will get fixed as a result, but only for PCSI HELP/MESSAGE.  A VMS documentation is one of the easiest components about which to @ file a proper complaint -- just send email to the address listed$ in the early pages of all the books.   ------------------------------    Date: 21 Jul 2000 10:05:32 +0200G From: Jan Vorbrueggen <jan@mailhost.neuroinformatik.ruhr-uni-bochum.de> G Subject: Re: ANALYZE/DISK/REPAIR (was Re: VMS equivalent of UNIX 'ln'?) H Message-ID: <y4ya2v3o37.fsf@mailhost.neuroinformatik.ruhr-uni-bochum.de>  L At least in the confirm mode, ANAL could offer to re-enter the file into itsM original directory. What would also help if a DUMP/HEADER translated the back H link into human-readable format, and SET FILE/ENTER was augmented with a /BACKLINK qualifier.    I Automatically having ANAL doing this operation without human intervention # or confirmation is a security risk.    	Jan   ------------------------------  % Date: Fri, 21 Jul 2000 10:40:23 +0100 5 From: "Adrian Lumsden" <A.Lumsden@spamtrap.xdt.co.uk> G Subject: Re: ANALYZE/DISK/REPAIR (was Re: VMS equivalent of UNIX 'ln'?) / Message-ID: <8l983s$1vr$1@newsg3.svr.pol.co.uk>   8 I think that I favour VERIFY putting the recovered files; in [SYSLOST]. If it restores them to the original directory = you could run into all sorts of problems with version numbers " if there's a version number clash.  B If VERIFY printed the directory backlink ("Was lost from [xxx]" ?)? as part of its normal logging of operations then you'd have the  information D you'd need to deal with the file and put it back where it (possibly): was lost from. Same goes for RSX if Mentec's listening :-)    
 with regards,    Adrian --( Adrian Lumsden, XDT Computer Systems, UK" A dot Lumsden at xdt dot co dot uk    @ David J. Dachtera <djesys.nospam@earthlink.net> wrote in message' news:3977B89B.7CB5806A@earthlink.net...  > Hoff Hoffman wrote:  > > E > > In article <39766725.5C7607E9@earthlink.net>, "David J. Dachtera" % <djesys.nospam@earthlink.net> writes: G > > :Since the file header indicates the FID of the directory where the B > > :"primary" entry should be found, why does VERIFY (the program
 behind theF > > :ANALYZE/DISK command) put a "lost" file (a file with no directory entry)F > > :into [SYSLOST] instead of the file's original directory (assuming the  > > :directory still exists)?  > > C > >   The directory backlink may or may not be current, even if the 	 directory D > >   listed is valid.  Would you rather have all files -- those not found inH > >   any directory -- consistently dropped into SYSLOST for resolution, > F > Sure! ...as long as the directory back-link still points to where it did H > before VERIFY came along, and doesn't point to the [SYSLOST] directory1 > (technically, an "invalid directory backlink").  >  > > orD > >   would you rather have some files dropped into SYSLOST and some files B > >   potentially randomly inserted back into (outdated) directory
 locations,B > >   directories that the files had previously been removed from? > H > Typically, you find a newbie SysAdmin, or even an experienced one on aF > new site, who finds the need to run VERIFY, then finds TONS of files inH > [SYSLOST], many with the same name and extension (but VASTLY differentF > sizes, attributes, organizations, etc.), and no clue where they came@ > from. (This has happened to me more often than I care to say!) > E > True, a file usually gets "lost" because the directory got trashed.  > E > Then again, you also get the over-privileged coder who goes hacking = > armed only with the on-line help not understanding that SET  FILE/REMOVE A > and DELETE are not to be toyed with ("Chill, dude! Just use the  UNDELETE
 > command!").  > G > As long as we got a message stating that VERIFY had re-entered a file H > back into it's indicated directory, I don't see it as a large problem,* > rather something of a time(life?)-saver. >  > I dunno - whaddaya think?  >  > -- > David J. Dachtera  > dba DJE Systems  > http://www.djesys.com/ > < > Unofficial Affordable OpenVMS Home Page and Message Board:! > http://www.djesys.com/vms/soho/    ------------------------------   Date: 21 Jul 2000 10:25:23 GMT' From: david20@alpha2.mdx.ac.uk (D.Webb) G Subject: Re: ANALYZE/DISK/REPAIR (was Re: VMS equivalent of UNIX 'ln'?) 0 Message-ID: <8l98ej$k8v$1@aquila.news.mdx.ac.uk>  g In article <3977B89B.7CB5806A@earthlink.net>, "David J. Dachtera" <djesys.nospam@earthlink.net> writes:  >Hoff Hoffman wrote: >>  j >> In article <39766725.5C7607E9@earthlink.net>, "David J. Dachtera" <djesys.nospam@earthlink.net> writes:F >> :Since the file header indicates the FID of the directory where theL >> :"primary" entry should be found, why does VERIFY (the program behind theL >> :ANALYZE/DISK command) put a "lost" file (a file with no directory entry)I >> :into [SYSLOST] instead of the file's original directory (assuming the  >> :directory still exists)? >>  L >>   The directory backlink may or may not be current, even if the directoryL >>   listed is valid.  Would you rather have all files -- those not found inG >>   any directory -- consistently dropped into SYSLOST for resolution,  > I >Sure! ...as long as the directory back-link still points to where it did G >before VERIFY came along, and doesn't point to the [SYSLOST] directory 0 >(technically, an "invalid directory backlink"). >  >> or I >>   would you rather have some files dropped into SYSLOST and some files L >>   potentially randomly inserted back into (outdated) directory locations,A >>   directories that the files had previously been removed from?  > G >Typically, you find a newbie SysAdmin, or even an experienced one on a H >new site, who finds the need to run VERIFY, then finds TONS of files inG >[SYSLOST], many with the same name and extension (but VASTLY different E >sizes, attributes, organizations, etc.), and no clue where they came ? >from. (This has happened to me more often than I care to say!)  > D >True, a file usually gets "lost" because the directory got trashed. > D >Then again, you also get the over-privileged coder who goes hackingH >armed only with the on-line help not understanding that SET FILE/REMOVEI >and DELETE are not to be toyed with ("Chill, dude! Just use the UNDELETE  >command!"). > F >As long as we got a message stating that VERIFY had re-entered a fileG >back into it's indicated directory, I don't see it as a large problem, ) >rather something of a time(life?)-saver.  >  >I dunno - whaddaya think? >  >--     3 Might be an idea for a future version of verify to    " 1) Put all files in syslost as nowD 2) Create a listing file with a standard name somewhere on the disk $    (either top level or in syslost).    The listing file to contain    ,    Filename (inc version) of file in SYSLOSTN    FID of file (to doubly make sure we know which file we are talking about)  &    Original directory of file if known  7    This file should be in a format so it can be edited.   L 3) Add new qualifier to ANALYZE/DISK  to prompt on whether to put each file I    which is listed in this new file back into their original directories.     J Hence a system manager will know easily where the files came from and can I decide whether to put them back or not  and can do so in a semi-automated  fashion.   Just a thought  
 David Webb VMS and Unix team leader CCSS Middlesex University   ------------------------------  % Date: Fri, 21 Jul 2000 12:18:04 -0400 ' From: "Bill Todd" <billtodd@foo.mv.com> G Subject: Re: ANALYZE/DISK/REPAIR (was Re: VMS equivalent of UNIX 'ln'?) ( Message-ID: <8l9t1u$1a6$1@pyrite.mv.net>  > Adrian Lumsden <A.Lumsden@spamtrap.xdt.co.uk> wrote in message) news:8l983s$1vr$1@newsg3.svr.pol.co.uk... : > I think that I favour VERIFY putting the recovered files= > in [SYSLOST]. If it restores them to the original directory ? > you could run into all sorts of problems with version numbers $ > if there's a version number clash. > D > If VERIFY printed the directory backlink ("Was lost from [xxx]" ?)A > as part of its normal logging of operations then you'd have the 
 > information F > you'd need to deal with the file and put it back where it (possibly)< > was lost from. Same goes for RSX if Mentec's listening :-)  L After this many years my memory could well be faulty, but my recollection is4 that the directory back-link did not exist in ODS-1.   - bill   >  >  > with regards,  >  > Adrian > --* > Adrian Lumsden, XDT Computer Systems, UK$ > A dot Lumsden at xdt dot co dot uk >  > B > David J. Dachtera <djesys.nospam@earthlink.net> wrote in message) > news:3977B89B.7CB5806A@earthlink.net...  > > Hoff Hoffman wrote:  > > > G > > > In article <39766725.5C7607E9@earthlink.net>, "David J. Dachtera" ' > <djesys.nospam@earthlink.net> writes: I > > > :Since the file header indicates the FID of the directory where the D > > > :"primary" entry should be found, why does VERIFY (the program > behind theH > > > :ANALYZE/DISK command) put a "lost" file (a file with no directory > entry)H > > > :into [SYSLOST] instead of the file's original directory (assuming > the  > > > :directory still exists)?  > > > E > > >   The directory backlink may or may not be current, even if the  > directory F > > >   listed is valid.  Would you rather have all files -- those not
 > found inJ > > >   any directory -- consistently dropped into SYSLOST for resolution, > > H > > Sure! ...as long as the directory back-link still points to where it > did J > > before VERIFY came along, and doesn't point to the [SYSLOST] directory3 > > (technically, an "invalid directory backlink").  > >  > > > orF > > >   would you rather have some files dropped into SYSLOST and some > files D > > >   potentially randomly inserted back into (outdated) directory > locations,D > > >   directories that the files had previously been removed from? > > J > > Typically, you find a newbie SysAdmin, or even an experienced one on aH > > new site, who finds the need to run VERIFY, then finds TONS of files > inJ > > [SYSLOST], many with the same name and extension (but VASTLY differentH > > sizes, attributes, organizations, etc.), and no clue where they cameB > > from. (This has happened to me more often than I care to say!) > > G > > True, a file usually gets "lost" because the directory got trashed.  > > G > > Then again, you also get the over-privileged coder who goes hacking ? > > armed only with the on-line help not understanding that SET 
 > FILE/REMOVE C > > and DELETE are not to be toyed with ("Chill, dude! Just use the 
 > UNDELETE > > command!").  > > I > > As long as we got a message stating that VERIFY had re-entered a file J > > back into it's indicated directory, I don't see it as a large problem,, > > rather something of a time(life?)-saver. > >  > > I dunno - whaddaya think?  > >  > > -- > > David J. Dachtera  > > dba DJE Systems  > > http://www.djesys.com/ > > > > > Unofficial Affordable OpenVMS Home Page and Message Board:# > > http://www.djesys.com/vms/soho/  >  >    ------------------------------  % Date: Fri, 21 Jul 2000 01:19:43 -0400 * From: David A Froble <davef@tsoft-inc.com>$ Subject: Re: DECServer 90 to DB-9 PC- Message-ID: <3977DD6F.CD320661@tsoft-inc.com>    quayle@pobox.com wrote:  > C > How do you hook a PC-style serial port (actually, a 500au running G > OpenVMS) to a DECserver 90M or 90TL?  I'm having trouble figuring out  > the connections.  7 Well, since I just dug the Fine Manual out for another.:  * 90TL and 90M use RJ45s.  Pin out for RJ45:   	1 - rxd ground-	 	2 - recvm 	3 - txd groundf 	4 - unused  	5 - unusedI
 	6 - transmitR 	7 - DTR 	8 - DSR  F And, the FM is wrong on page A-6 of the DECserver 90TL owner's manual.  K As for the PC style DB9, I don't have a FM handy.  The FAQ should have some/P info, and your 500au owner's manual should be helpful.  The pins you really needP are transmit, receive, and both grounds.  The grounds should be tied together atP the DB9 end.  If I had to guess, I think that the transmit is pin 2, and receiveO is pin 3.  Remember to cross the transmit and receive, ie; xmit to recv in both  directions.u   Dave   -- d4 David Froble                       Tel: 724-529-04504 Dave Froble Enterprises, Inc.      Fax: 724-529-0596> DFE Ultralights, Inc.              E-Mail: davef@tsoft-inc.com6 T-Soft, Inc.  170 Grimplin Road  Vanderbilt, PA  15486   ------------------------------  % Date: Fri, 21 Jul 2000 13:08:24 +0100 - From: "Rob van Lopik" <lopik@mail.telepac.pt>F$ Subject: Re: DECServer 90 to DB-9 PC* Message-ID: <8l9ehg$ero$1@duke.telepac.pt>  # Having just soldered my own cables:m   On a 9-pin connector:  2 = receive data 3 = transmit datae 5 = signal grounds   I found all info at:$ http://www.airborn.com.au/rs232.html   regards,
 rob van lopikl    7 "David A Froble" <davef@tsoft-inc.com> wrote in messagel' news:3977DD6F.CD320661@tsoft-inc.com...e > quayle@pobox.com wrote:  > >aE > > How do you hook a PC-style serial port (actually, a 500au running<I > > OpenVMS) to a DECserver 90M or 90TL?  I'm having trouble figuring out< > > the connections. snip > H > And, the FM is wrong on page A-6 of the DECserver 90TL owner's manual. >YH > As for the PC style DB9, I don't have a FM handy.  The FAQ should have someF > info, and your 500au owner's manual should be helpful.  The pins you really need'F > are transmit, receive, and both grounds.  The grounds should be tied together at J > the DB9 end.  If I had to guess, I think that the transmit is pin 2, and receivelL > is pin 3.  Remember to cross the transmit and receive, ie; xmit to recv in both
 > directions.  >e > Dave >s > --6 > David Froble                       Tel: 724-529-04506 > Dave Froble Enterprises, Inc.      Fax: 724-529-0596@ > DFE Ultralights, Inc.              E-Mail: davef@tsoft-inc.com8 > T-Soft, Inc.  170 Grimplin Road  Vanderbilt, PA  15486   ------------------------------  % Date: Fri, 21 Jul 2000 14:25:57 +0100g- From: "Rob van Lopik" <lopik@mail.telepac.pt>1$ Subject: Re: DECServer 90 to DB-9 PC* Message-ID: <8l9j37$rn0$1@duke.telepac.pt>  # Having just soldered my own cables:n    On a 9-pin connector:  2 = receive datan  3 = transmit data  5 = signal ground    I found all info at:P%  http://www.airborn.com.au/rs232.htmls  	  regards,   rob van lopik    8  "David A Froble" <davef@tsoft-inc.com> wrote in message(  news:3977DD6F.CD320661@tsoft-inc.com...F  > > How do you hook a PC-style serial port (actually, a 500au runningJ  > > OpenVMS) to a DECserver 90M or 90TL?  I'm having trouble figuring out  > > the connections.   > Davee  >  > --c7  > David Froble                       Tel: 724-529-0450I7  > Dave Froble Enterprises, Inc.      Fax: 724-529-0596nA  > DFE Ultralights, Inc.              E-Mail: davef@tsoft-inc.comt9  > T-Soft, Inc.  170 Grimplin Road  Vanderbilt, PA  15486i   ------------------------------  # Date: Fri, 21 Jul 2000 13:35:42 GMTd From: operagost@my-deja.coml, Subject: determining free disk space via DCL) Message-ID: <8l9jje$b9p$1@nnrp1.deja.com>9  F Is there a lexical or something that I can use to determine free spaceF on a disk? The problem is that, using VMS 5.5-2 we're limited to ~8 GBH disks. What we're doing is storing Oracle archive logs on a raid-5, thenG moving the older ones to another disk with a daily batch file. However, @ sometimes database activity increases and the logs overflow diskF capacity. I'd like to optimize the use of this array and avoid lockingD up the archiver by moving files to a second, non-redundant disk only when necessary.'    & Sent via Deja.com http://www.deja.com/ Before you buy.    ------------------------------  # Date: Fri, 21 Jul 2000 14:15:31 GMT , From: "J. Scott Greig" <jsgreig@geminaq.com>0 Subject: Re: determining free disk space via DCL4 Message-ID: <7WYd5.5656$GO4.37242@quark.idirect.com>  + Look at F$GETDVI("<diskname>","FREEBLOCKS")s  ? This should give the number of free blocks on the logical unit.    ScottmL <operagost@my-deja.com> wrote in message news:8l9jje$b9p$1@nnrp1.deja.com...H > Is there a lexical or something that I can use to determine free spaceH > on a disk? The problem is that, using VMS 5.5-2 we're limited to ~8 GBJ > disks. What we're doing is storing Oracle archive logs on a raid-5, thenI > moving the older ones to another disk with a daily batch file. However,.B > sometimes database activity increases and the logs overflow diskH > capacity. I'd like to optimize the use of this array and avoid lockingF > up the archiver by moving files to a second, non-redundant disk only > when necessary.0 >2 >2( > Sent via Deja.com http://www.deja.com/ > Before you buy.u   ------------------------------   Date: 21 Jul 2000 14:08:35 GMT2 From: hoffman@xdelta.zko.dec.nospam (Hoff Hoffman)0 Subject: Re: determining free disk space via DCL6 Message-ID: <8l9lh3$fa2$1@mailint03.im.hou.compaq.com>  G In article <8l9jje$b9p$1@nnrp1.deja.com>, operagost@my-deja.com writes: G :Is there a lexical or something that I can use to determine free space  :on a disk?   B   Maybe some combination of the system time and the following two    DCL lexical functions:  9     $ freeblocks = f$getdvi("sys$sysdevice","freeblocks")   5     $ maxblock = f$getdvi("sys$sysdevice","maxblock")1  E   would help figure out if/when you are rapidly approaching a problemu;   with your available storage on the specified disk device.X  A   Seriously consider an upgrade to your OpenVMS release, as well. <   (OpenVMS VAX V5.5-2 was released back in 1992, after all.)  N  --------------------------- pure personal opinion ---------------------------L    Hoff (Stephen) Hoffman   OpenVMS Engineering   hoffman#xdelta.zko.dec.com   ------------------------------  + Date: Fri, 21 Jul 2000 12:35:14 -0500 (EST) 1 From: "Robert J. Slover" <slover@Rose-Hulman.Edu>t, Subject: Re: Dialup internet sharing routersW Message-ID: <Pine.LNX.3.96.1000721123155.20499B-100000@rocinante.admin.rose-hulman.edu>   A I just (as in 2 days ago) bought a Netgear RM356 which does this.s? I've not tried this capability yet, just know that it is there.d= I, too, plan to have my VMS/Alpha box be the designated host.5  8 RM356 56K INET GTW RTR W/ V90MODEM & 4PT ETH HUB (RM356)   $182.00 from PCSTOP.com@   --Robert  * On Thu, 20 Jul 2000, Richard Jordan wrote:  G > I've been reviewing some earlier discussion about the dialup interneteG > access sharing boxes (dialup router with NAT, allows multiple clientstE > on a local LAN to access the internet via a PPP dialup link) and I  A > want to use one to get my systems online.  Can anyone recommendf8 > one that has the capability of supporting what amounts# > to a 'server' on the client side?t > E > Situation:  home network with my VMScluster on it (three nodes, twotD > VAXen and one Alphastation).  I just got a nice network-only laser> > printer (no standard ports, just 10Base-T with IP protocol).6 > I've test printed to it from the Alpha and it works. > fine.  > F > What I'd like to be able to do is use the dialup box to set up a PPPH > (RAS, but not to an NT box) link to work and have the work machines beC > able to print to that laser printer.  Since the dialup box almostLJ > certainly will only present one external address (probably DHCP-derived)G > are any of them capable of redirecting inbound packets to an internal F > host based on the port number being used?  Or some similar facility?E > I know I could probably set up some chain of LPR/LPD queues through 9 > one of the VMS systems, but if possible I'd rather not.m > K > I've got the book for a Kingston EtheRx, which cannot apparently do this.tE > Anything else short of a full blown (expensive) router provide this 
 > capability?r >  > Thanks...  > 
 > Rich Jordany > rjordan@mcs.nete >    ------------------------------   Date: 21 Jul 2000 15:25:05 GMT2 From: mathog@seqaxp.bio.caltech.edu (David Mathog)! Subject: Re: DS10 IDE vs U2W SCSIt, Message-ID: <8l9q0h$29s@gap.cco.caltech.edu>  ` In article <39774C8B.1E17E6B8@digital.com>, Rich Dischler <richard.dischler@digital.com> writes:? >The DS10 IDE chip runs at 16MHz DMA mode 2. This allows burstssO >in the 25MB/s range.  For many of our customers, this, plus the lower price of  >IDE >disks, is ok for them.n  * Ah, technical information at last, thanks. >r> >If not, we have several flavors of SCSI PCI cards to plug in.  K Sure, but SCSI controllers and disks are much more expensive than IDE/ATA -/F even when you look at ATA/66 and ATA/100 devices.   You have to spend D around $250 for an Intraserver (linux variant) SCSI controller, but C ATA/100 controllers run as low as $45 (heck, you can buy an Athlon $G motherboard with built in ATA100 for less than the one SCSI controller!eF Really, see:  http://www.cdw.com/shop/products/default.asp?EDC=229075)E DS10 and DS10L systems often only have one disk in them so, so SCSI'saK ability to handle more devices wasn't relevant.  Thus the original questionaK was, how did the built in IDE disk controller compare in performance with abK SCSI controller - and the answer was, not well.  But the IDE implementationuI in these machines seems to be archaic, so one starts to wonder if insteaddF of SCSI it might be possible to add an ATA/100 controller and disk andB still spend less than SCSI, but achieve "comparable" performance. I Supposedly ATA/100 is supported in the Linux 2.4 kernels, if not by Tru64lK and OpenVMS.  That would at least allow Compaq to test such a configurationLK and possibly even offer it for Linux variants of these machines.  (It isn'tpK like Compaq doesn't sell millions of these controllers and disks every yearSC already!) Ideally one would like to be able to boot from the second C controller though, and that may require work on SRM to support it. T  G For smallish systems (other than Alphas, due to lack of drivers) ATA isi4 looking more and more attractive, see for instance:   M    http://www.gamepc.com/reviews/hardware_review.asp?review=cheetahx15&page=2e/    http://www.promise.com/Products/products.htms5    http://www.promise.com/Docs/Datasheets/FT100DS.pdf[  LI The Promise FastTrak100 card maxes out at 4 drives per controller, but it K is very competitive on price and performance against SCSI up to that limit.h   Regards,   David Mathog mathog@seqaxp.bio.caltech.eduu? Manager, sequence analysis facility, biology division, Caltech     ------------------------------  % Date: Fri, 21 Jul 2000 10:24:03 +0200a( From: "Web2.lyon" <Web2.lyon@wanadoo.fr>, Subject: How to speed up RMS files reading ?* Message-ID: <397808A2.50187A72@wanadoo.fr>   Hi  H  I am trying to read a RMS file (Indexed, with one key) from a C programG under VMS. To get good I/O performances, I tried to read it as a binaryi file, by using  )          fopen(fp, "ctx=bin", "rfm=var");v and then'          fread(buf, BUF_LENGTH, 1, fp);    but it gave bad results.  E Does anybody know how to perform a speedy read of VMS indexed files ?      FT   ------------------------------  % Date: Fri, 21 Jul 2000 10:31:25 +0200 = From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>i0 Subject: Re: How to speed up RMS files reading ?) Message-ID: <39780A5D.53D2F0FF@gtech.com>    "Web2.lyon" wrote:J >  I am trying to read a RMS file (Indexed, with one key) from a C programI > under VMS. To get good I/O performances, I tried to read it as a binarye > file, by using > + >          fopen(fp, "ctx=bin", "rfm=var");t
 > and then) >          fread(buf, BUF_LENGTH, 1, fp);g >  > but it gave bad results. > G > Does anybody know how to perform a speedy read of VMS indexed files ?   0 I would drop the C RTL and do direct RMS calls !   Arne   ------------------------------  # Date: Fri, 21 Jul 2000 09:11:21 GMT-% From: rlb@fdhoekstra.nl (Richard Bos)00 Subject: Re: How to speed up RMS files reading ?4 Message-ID: <39780f57.678475704@news.worldonline.nl>  ) "Web2.lyon" <Web2.lyon@wanadoo.fr> wrote:8  J >  I am trying to read a RMS file (Indexed, with one key) from a C programI > under VMS. To get good I/O performances, I tried to read it as a binaryI > file, by using > + >          fopen(fp, "ctx=bin", "rfm=var");e  C This is not an ISO-compliant way of using fopen(). I suspect you'reoC using a VMS-specific extension to ISO C (or maybe even a completely.= non-ISO replacement); as such, it's off-topic to comp.lang.c.m   Richard    ------------------------------  % Date: Fri, 21 Jul 2000 09:05:24 -0400m= From: "Stephane Paquin" <stephane.paquin@nospam.ispatnet.com>e0 Subject: Re: How to speed up RMS files reading ?3 Message-ID: <pUXd5.24593$227.443973@nnrp1.uunet.ca>y  I If you have a Fortran compiler installed, you could use its great support-K for indexed files to create subroutines to access your indexed files. I useiL this way in C programs where I have made file access IO routines in Fortran,H compiled  and linked the objects to the C program. Works great. Then youI could use the BLOCKSIZE and BUFFERCOUNT values to improve file access. Ifn you need help, contact me.       Stphane PaquinE  1 Web2.lyon <Web2.lyon@wanadoo.fr> wrote in messagep$ news:397808A2.50187A72@wanadoo.fr... > Hi >lJ >  I am trying to read a RMS file (Indexed, with one key) from a C programI > under VMS. To get good I/O performances, I tried to read it as a binaryY > file, by using >a+ >          fopen(fp, "ctx=bin", "rfm=var");r
 > and then) >          fread(buf, BUF_LENGTH, 1, fp);l >r > but it gave bad results. > G > Does anybody know how to perform a speedy read of VMS indexed files ?  >  >  > FT >t >v >i   ------------------------------   Date: 21 Jul 2000 13:26:03 GMT2 From: hoffman@xdelta.zko.dec.nospam (Hoff Hoffman)0 Subject: Re: How to speed up RMS files reading ?6 Message-ID: <8l9j1b$emo$1@mailint03.im.hou.compaq.com>  s In article <pUXd5.24593$227.443973@nnrp1.uunet.ca>, "Stephane Paquin" <stephane.paquin@nospam.ispatnet.com> writes: J :If you have a Fortran compiler installed, you could use its great supportH :for indexed files to create subroutines to access your indexed files...M :file access IO routines in Fortran, compiled  and linked the objects to the iH :C program... Then you could use the BLOCKSIZE and BUFFERCOUNT values...  F   This capability to specify the multiblock and multibuffer counts is H   available in a variety of languages, including the Compaq C (formerly '   DEC C) RTL and via direct RMS access.t  G   If you choose to get full access to the record management system, it gE   is usually best to go directly after RMS itself -- what Fortran andpC   C provide is only part of what is available via direct RMS calls.u  <   An example of using RMS directly (from C) is available at:  F     http://www.openvms.compaq.com/freeware/srh_examples/RMS_EXAMPLES.C  3   There are various other examples around the 'net.   2 :Web2.lyon <Web2.lyon@wanadoo.fr> wrote in message% :news:397808A2.50187A72@wanadoo.fr... K :>  I am trying to read a RMS file (Indexed, with one key) from a C programtJ :> under VMS. To get good I/O performances, I tried to read it as a binary :> file, by usingg  D   This binary access does not look to be an approach I would choose.H   If you have an indexed file, you should access it as an indexed file,    and not as a raw/binary file.o  I   Note that the classic performance vs features trade-off exists here --  K   the lighter-weight and faster the access, the fewer controls and sharing  K   features you will typically have.  And by the time you get done layering .L   in the controls and features your application needs (and debugging them), D   the lighter-weight access does not necessarily look quite as good.    , :>          fopen(fp, "ctx=bin", "rfm=var");  G   The options "mbc={multi-block_value}" and "mbf={multi-buffer_value}"  H   are available using C, and can assist with I/O performance -- this as $   was suggested by the other poster.  J   If access to the file is shared, then enabling and using global buffers    can help with performance.  E   You will also want to take a careful look at your system parametersnE   and particularly at your process quotas, as well as the performanceF    of your particular I/O system.  H :> Does anybody know how to perform a speedy read of VMS indexed files ?  F   Exactly what sort of read?  Random access reads, or sequential read?  N  --------------------------- pure personal opinion ---------------------------L    Hoff (Stephen) Hoffman   OpenVMS Engineering   hoffman#xdelta.zko.dec.com   ------------------------------  % Date: Fri, 21 Jul 2000 20:26:06 +0010 % From: paddy.o'brien@zzz.tg.nsw.gov.auI= Subject: Re: I need a precise definition of "layered product"E5 Message-ID: <01JS1IKYMJN60021SA@tgmail.tg.nsw.gov.au>R   John Nixon wrote:1  L >I had thought this was an easy question, but I see from the other responses thatJ >I must have been naive.   I would have thought that a layered product was any H >product that is "layered" on top of the OS.  If you install the OS, and there isI something you need that is not there, you install a "layered" product.  Ig
 have dealtI >with layered products with various manufacturers that have nothing to doa
 >with DECpaq.D  O That is indeed my perception.  I think of my compilers as "layered products".  tG Since I work in a commercial environment, I am not sure what DECpaq is.s  5 Discussion below is ellided regarding SCAN and LATeX.)  O SCAN is no longer proprietary (as also Bliss), Digital gave these to the world .& on the freeware CD's including source.  Q LATeX has never been proprietary,  To my knowledge Leslie Lamport gave it to the hQ world as did Knuth with his original TeX.  LL and DK maintain a copyright on the hK source, I believe, along with WEB and WEAVE as discussed in another thread.t   Regards, Paddy   Paddy O'Brien, Transmission Development, 
 TransGrid, PO Box A1000, Sydney South,  NSW 2000, Australiad   Tel:   +61 2 9284-3063 Fax:   +61 2 9284-3050& Email: paddy.o'brien@zzz.tg.nsw.gov.au  M Either "\'" or "\s" (to escape the apostrophe) seems to work for most people,T; but that little whizz-bang apostrophe gives me little spam.2   ------------------------------  % Date: Fri, 21 Jul 2000 13:38:38 +0200m= From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>-= Subject: Re: I need a precise definition of "layered product" ) Message-ID: <3978363E.B7682EAF@gtech.com>e  & paddy.o'brien@zzz.tg.nsw.gov.au wrote:I > Since I work in a commercial environment, I am not sure what DECpaq is.m   DECpaq = DEC/Compaq    Arne   ------------------------------  % Date: Fri, 21 Jul 2000 10:24:13 +0200f= From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>r Subject: Re: INFO ON A VT320) Message-ID: <397808AC.C7C90D10@gtech.com>t   miguel ramirez wrote:DF >      I am running a Microvax 3100 computer model 20.  My question isJ > that I looking for a vt320 for this computer.   I have found some vt320,H > but the thing is that they have different models.  I need to know whatE > model of the vt320 will I need to connect it with the microvax 3100b > computer model 20.  E Any VT220/VT240/VT241/VT320/VT330/VT340/VT420/VT520/VT525 should do !e   Arne   ------------------------------  # Date: Fri, 21 Jul 2000 07:31:36 GMT-( From: Terry Kennedy <terry@gate.tmk.com>) Subject: Re: Intraserver SCSI Controllers@' Message-ID: <Fy1E8o.9L4@spcuna.spc.edu>i  5 Eric Dittman <dittman@narnia.int.dittman.net> writes:m> > Does anyone know of a vendor that sells the Intraserver SCSI6 > controllers that support OpenVMS for less than list?  H   Compaq list or IntraServer list? IntraServer's list price is less than Compaq's...   I   If you look at IntraServer's list of distributors, you'll probably findHH some that will sell to you at a better price. Of course, it helps if youG are looking for more than 1, and you don't act end-userish. I've gotten F nice discounts from some of them in the past (on the generic, not VMS, cards).a  - 	Terry Kennedy             http://www.tmk.com 5         terry@tmk.com             Jersey City, NJ USAa   ------------------------------  , Date: Fri, 21 Jul 2000 14:53:45 +0300 (EEST)! From: INFE <noAddress@infext.com>s( Subject: Join the INFe-Investors Network8 Message-ID: <200007211153.OAA18455@sh1.swisshosting.net>   Dear Investor,  2 INFe.com, Inc. is proud to invite you to join the 2 INFe-Investors Network. Membership to this unique ; network of international investors and information is free.l  $      http://www.infe.com/infe-in.asp  : Recently, Stockreporter began coverage of INFe.com, Inc.  0 (OTC/BB: INFE) with a STRONG BUY recommendation 8 and price target of $7.00 per share in 12 to 18 months. 9 For the full story: http://www.infe.com/infe-n-y2k-19.aspf  # To learn more about INFe.com go to:t    http://www.infe.com!    http://www.infe.com/infe-n.asp    Have a great day!u  
 Sincerely, INFe.com, Inc.  < NOTICE: If you would like to be removed from this email list$ then enter your email address in at:-   http://www.infe.com/infe-removeFromList.asp    ------------------------------  , Date: Fri, 21 Jul 2000 17:30:04 +0200 (CEST): From: "Gotfryd Smolik, VMS lists" <gotfryd@stanpol.com.pl>7 Subject: Re: Limit a UIC group access to a number of 50 I Message-ID: <Pine.LNX.4.21.0007211719040.4023-100000@irys.stanpol.com.pl>1  - On Wed, 19 Jul 2000 trdorr@my-deja.com wrote:h< [...your answer to Larry Kilgallen message... Please: answer,  *after* included (original) post/mail !...]  + +Thanks for the suggestion for MAXACCTJOBS.CH +However,MAXACCTJOBS does not address the solution. I want to be able toD +limit the number of users in a UIC group to not be able to login if0 +there are more than 50 users in that UIC group.  !  Can you be more specific, why ??i(  IMHO it is suposed do exactly that job.)  Supose group 743 and account name LIM50:u   UAF> MOD [743,*]/ACCOUN=LIM50O UAF> MOD [743,*]/MAXACCTJOBS=50c  6  To be more specific: if you have the 743 group named,6  then by default the account name is the same as group4  (identifier) name when user is added with AUTHORIZE3  (check HELP/ADD, ADD_IDENT *also* adds the accounta1  name). Regardless of that you can "insert" *any*R  username to *any* account.-  6  And you can set differrent limit to selected users...4  MAXACCTJOBS is "connected" to username, not account! - even if is by-account checked !     Regards - Gotfryd   --  E =====================================================================yF $ ON F$ERROR("LANGUAGE","ENGLISH","IN_MESSAGE").GT.F$ERROR("NORMAL") - 		THEN EXCUSE/OBJECT=ME5. $!                        GS@stanpol.zabrze.plE =====================================================================2   ------------------------------  % Date: Fri, 21 Jul 2000 11:40:49 +0100 / From: Nigel Arnot <sysmgr@maxwell.ph.kcl.ac.uk>s Subject: re: Linux "clusters"s7 Message-ID: <009ED690.FAB54B12.23@maxwell.ph.kcl.ac.uk>e  C > On 20 Jul 2000 15:04:10 GMT, mathog@seqaxp.bio.caltech.edu (Davidn > Mathog) wrote: > H > >Here's an article which describes development work (by SGI!) on LinuxI > >clusters specifically designed to behave like VMS clusters.  Given thesM > >astronomical price of VMS clusters it isn't at all surprising that this is I > >happening - there's plenty of market below the VMS cluster price pointtN > >which could benefit from VMS cluster capabilities.   On the other hand, it H > >sounds like they are miles behind TruCluster, let alone VMS clusters. > >u6 > >  http://www.theregister.co.uk/content/1/12022.html > >eQ > >(Warning:  some of the statements you'll read there are laughable inaccurate!)  > F > Yeah, you pretty much have to take anything The Register says with a > grain of salt. >   J Grain is about right ... compared to the spoonful that most of the rest ofH the trade press merits. IMO "The Register" is one of the better IT news H sources, though they do like to stir things up on purpose and sometimes 	 unfairly.n  F Looks as if Linux clusters right now are about on a par with MicrosoftH clusters: node failover. But where they are now isn't what's important, F it's where they want to go: genuine shared-everything Linux clusters. G And the Linux community has a pretty fine track record of making things5 that work well.   G In passing: if a node is as cheap as a PC is, and the software is free,SF then node failover isn't really all that bad. For a version 1, anyway.G How many PCs can you buy for the price of two full VMScluster licenses?i   	Yours,D
 		Nigel Arnotn- 		NRA@MAXWELL.PH.KCL.AC.UK                      7 		"In the beginning there was nothing, which exploded."e   ------------------------------  % Date: Thu, 20 Jul 2000 22:45:24 -0700o! From: Koloth <koloth@tmisnet.com> : Subject: Re: Mlucas performance probelm - need expert help+ Message-ID: <3977E374.91A10596@tmisnet.com>e   Just taking a stab./  Z It does sound like a cache issue.  Cache uses some hash of the physical address to convertU a memory address to a cache address.  I remember that there was a fortran pre or post E processor (Knuth???) that would take cache architecture into account.s  ? One of the examples that they gave was the summing of two arrays I.e. C[i] = A[i] + B[i]EN if Array A and B were of a size modulo of the cache hash value then A[i] wouldN hash to the same cache address as B[i].  This would casue the cache to thrash.\ By placing a small array between the two arrays it would cause the the hash value to differ.  F Hope this helps.  You may want to investigate how the cache is set up._ Digital did a great job of explaining this in there Digital Technical Journals when they talked ! about new processors and systems.a   Regards, Cass Witkowski   Robert Deininger wrote:"   > Hello folks, >sH >    I am trying to understand (and hopefully fix) a performance problem/ > that's appeared in the VMS version of Mlucas.t >cA > (See ftp://hogranch.com/pub/mayer/README.html for Mlucas info.)f >eI > (This is probably a post for wizards and experts.  You've been warned.)l >s > A little background info:d > C > Mlucas uses the Lucas-Lehmer algorithm to hunt for Mersenne primerE > numbers.  It spends most of its time squaring huge integers using aiC > fancy Fast Fourier Transform method -- which means it is entirelyn > CPU bound.  It grinds through the same memory again and again.  It uses otherwise wasted idle time, and a typical run takes 40 to 50 days on ourL > fastest system, depending on how much idle time is available.  I'm chasingI > a seemingly-random 10% performance hit, which means a savings of 4 or 5l( > days per run if I can find a solution. >dD > The problem I see is this.  If I stop the program and then restartD > it on the same system, it gets faster or slower by as much as 10%.8 > And stays that way forever (until I restart it again). >-G > Here is one illustration.  Mlucas writes a line of status informationnA > and writes a checkpoint file every 2000 iterations.  It reports-G > how much time elapsed during the 2000 iterations.  Here is an excerptlE > from the status file, showing a few iterations before and after thep
 > restart: >t; > M( 10591183 ) iteration =  4282000 clocks = 00:13:24.920.n; > M( 10591183 ) iteration =  4284000 clocks = 00:13:24.760.U; > M( 10591183 ) iteration =  4286000 clocks = 00:13:25.123.y; > M( 10591183 ) iteration =  4288000 clocks = 00:13:24.874. ; > M( 10591183 ) iteration =  4290000 clocks = 00:13:24.762.0; > M( 10591183 ) iteration =  4292000 clocks = 00:13:25.141.e0 >  Restarting M(10591183 ) at iteration  4292000; > M( 10591183 ) iteration =  4294000 clocks = 00:12:04.509.m; > M( 10591183 ) iteration =  4296000 clocks = 00:12:01.806.I; > M( 10591183 ) iteration =  4298000 clocks = 00:12:01.468.o; > M( 10591183 ) iteration =  4300000 clocks = 00:12:01.861. ; > M( 10591183 ) iteration =  4302000 clocks = 00:12:01.591. ; > M( 10591183 ) iteration =  4304000 clocks = 00:12:01.655.u; > M( 10591183 ) iteration =  4306000 clocks = 00:12:01.770.m >sG > When I captured this status output, the machine wasn't doing anythingtJ > except running Mlucas and chatting with other cluster members.  It was aD > Saturday morning, and there was NO user activity.  Nothing changedC > around the restart except a new process was started and the image0M > was reactivated.  Work that used to take 13 minutes and 25 seconds now only ! > takes 12 minutes and 2 seconds.  >h > The hardware is:7 >    Digital Personal Workstation 600au - 600 MHz 21164d+ >    320 MB RAM (in 2 banks, 64MB + 256 MB) # >    2 MB external (level 3) cache.m >r > Software:i >    OpenVMS 7.1-2  >    DIGITAL Fortran 90 X7.1-574 >uE > Other Mlucas users have reported similar behavior on Alphas runninglL > linux, but NOT on Alphas running Digital Unix  (I don't know the versions,K > but I know the Digital Unix program is compiled with Digipaq Fortran 90.) I > We have two of these DPW systems, and they both show the same behavior.: >wK > After reading through the Alpha Architecture Handbook, the 21164 HardwareuJ > Reference Manual, and the OpenVMS 1.5 Internals and Data Structures bookN > (NOT the whole thing), I suspect that the program's behavior in the L3 cacheJ > is changing, depending on where the image activator places it in memory.H > But I'm over my head.  I don't know how to pinpoint the problem, and IJ > REALLY don't know how to fix it.  I just have a bunch of dumb questions, > and a general plea for help. >mE > Our systems have enough memory, and sufficiently dainty users, that F > Mlucas usually claims the memory it needs and then never page faultsF > again.  I've seen it run 2 or 3 weeks without the page fault counter > incrementing.O >rI > Question(s) 1:  If I incur no page faults, does than mean I'm occupying C > exactly the same physical pages of memory?  Or does VMS have some G > way of suffling memory where it does not bump the page fault counter?k > H > Question(s) 2:  Does the Performance and Coverage Analyzer (PCA) allowC > me to measure details of program behavior like this?  Can I learnGG > anything about cache misses and other hardware-related details?  Fromt$ > my hunting so far, it appears not. >DL > Question(s) 3:  The 21164 has some internal processor registers that allowF > things like cache misses to be measured.  Is there any way to accessF > these processor registers in VMS?  (It seems to me that PCA ought toI > provide an interface to these features.)  If I figure out how to tickle / > the performance counters, will I blow up VMS?t >,K > Question(s) 4:  Is there any way a VMS process can control (or influence/tG > affect/bribe) which _physical_ pages it occupies?  Can a process eveneE > find out which physical pages it's using?  From groping in the darkgH > with SDA, the page tables do not appear to be readable from user mode.A > Is there a system service that will give me a virtual->physicalm@ > translation for a particular address?  Would it be possible toA > make a user-written system service to provide this information?l >iH > Question(s) 5:  Why do VMS (and linux) programs have this problem, butD > Digital Unix programs don't?  Isn't the Compaq F90 compiler almost > identical on VMS and Unix? >iD > Thanks for any advice and answers you can offer, including answers0 > to the questions I should ask, but didn't. :-) >.C > I'm willing to tiptoe into Kernel mode if necessary to make toolso8 > to cure this disease, but I don't know where to start. > @ > The URL at the top of the post will lead you to Mlucas, if youG > want to try it on your own alpha systems.  Object files are provided,iE > so you don't need an F90 compiler.  If you have VMS alphas that arerD > always up, and have some unused CPU cycles, you should try finding- > the largest known prime number with Mlucas.a >f > -- > Robert Deininger > rdeininger@mindspring.come   ------------------------------    Date: 21 Jul 2000 10:02:03 +0200G From: Jan Vorbrueggen <jan@mailhost.neuroinformatik.ruhr-uni-bochum.de>t: Subject: Re: Mlucas performance probelm - need expert helpH Message-ID: <y41z0n52tg.fsf@mailhost.neuroinformatik.ruhr-uni-bochum.de>  4 rdeininger@mindspring.com (Robert Deininger) writes:  D > The problem I see is this.  If I stop the program and then restartD > it on the same system, it gets faster or slower by as much as 10%.8 > And stays that way forever (until I restart it again).  K You are seeing the effects of page colouring or, rather, the fact that the  L VMS page fault handler doesn't do page colouring and DUNIX's does. (Neither K does LinuxALpha, at the moment). I think adding it is on the VMS to-do-listwK (Hoff?!). Do the following: modify the program to run a few iterations (forfJ a few minutes, say), and if the CPU time is longer than a machine-depedent& value, write a checkpoint and restart.  N I've heard a report of a factor of two slowdown for one particular application1 which was solved by doing what I described above.e   	Jan  J PS: So what is page colouring? The L2/L3 caches of the 21154 and 21264 areI larger than the page size multiplied by their associativity, and they useMI physical addresses. It thus can happen that two pages that wouldn't see aiK cache conflict when using virtual addresses happen to be mapped to physicaleK addresses that are identical in the relevant address bits and thus clash intI the cache. Page colouring means that the page fault handler endeavours to K obtain a page from the free list such that this is unlikely to happen, for aK instance by using a physical page whose relevant address bits are identical I to the virtual address being mapped. The VMS page fault handler currentlyiD operates FIFO on the free page list and is oblivious to this effect.   ------------------------------  % Date: Fri, 21 Jul 2000 10:44:04 -0400 2 From: rdeininger@mindspring.com (Robert Deininger): Subject: Re: Mlucas performance probelm - need expert helpL Message-ID: <rdeininger-2107001044040001@user-2ive7fg.dialup.mindspring.com>   In article <y41z0n52tg.fsf@mailhost.neuroinformatik.ruhr-uni-bochum.de>, Jan Vorbrueggen <jan@mailhost.neuroinformatik.ruhr-uni-bochum.de> wrote:t  6 > rdeininger@mindspring.com (Robert Deininger) writes: > F > > The problem I see is this.  If I stop the program and then restartF > > it on the same system, it gets faster or slower by as much as 10%.: > > And stays that way forever (until I restart it again). > M > You are seeing the effects of page colouring or, rather, the fact that the  N > VMS page fault handler doesn't do page colouring and DUNIX's does. (Neither M > does LinuxALpha, at the moment). I think adding it is on the VMS to-do-list-M > (Hoff?!). Do the following: modify the program to run a few iterations (foriL > a few minutes, say), and if the CPU time is longer than a machine-depedent( > value, write a checkpoint and restart.  I It should work.  I'll have to be careful to ignore the timing measurement=G if I'm not the only active process.  Re-activating just because someoneu4 else is using a bunch of cache isn't likely to help.  G I'd prefer to somehow find out which pages are badly mapped to physical D memory, and force only those to be remapped.  Some kind of iterativeF algorithm that would tend to converge on a good mapping.  I guess thatG would mean re-inventing a page coloring algorithm by hand.  (Turns pale  and hides under desk.)  L (Related question:  Does VMS give me a way to learn the size of the externalH cache(s)?  The console must know this, so it should be lurking in a data structure somewhere.)T  M It is a bit uglier than it seems at first.  Mlucas is actively using at leastrL twice as much data as will fit in the L3 cache.  (Our cache is 2 MB.)  So I C don't really know what the optimum cache behavior is.  Is it bettersP to keep half the data in cache all the time, and never cache the rest?  Or cacheJ different sections at different times in the course of a single FFT?  OverK a full iteration, all the data is accessed equally often, but there is muchr locality of reference over shorter times.  This is a high-powered multi-radix FFT, and I'm far from having my brain wrapped around it.  A stupid page & coloring algoithm might not help here.  F I don't think we have enough performance data to say whether the DunixC scheme results in faster execution times than the VMS one.  I don'tMA know of head-to-head comparisons on identical hardware.  So while I Dunix gives stable execution times, I'm not certain they're faster.  It's-E possible, though not likely, that Dunix picks the worst page mapping.oO Does it necessarily pick the best, in the case where the cache is significantly0I smaller than the data?  I don't see why it would, off the top of my head.i2 I guess it probably falls somewhere in the middle.    P > I've heard a report of a factor of two slowdown for one particular application3 > which was solved by doing what I described above.   M I don't know what the best case is on my platform yet.  I've seen a variation  of a bit more than 10%, but I don't know how close I've come to the best and worst cases.  I don't think a factor of 2 is likely.n    
 >         Janl > L > PS: So what is page colouring? The L2/L3 caches of the 21154 and 21264 areK > larger than the page size multiplied by their associativity, and they useyK > physical addresses. It thus can happen that two pages that wouldn't see asM > cache conflict when using virtual addresses happen to be mapped to physical-M > addresses that are identical in the relevant address bits and thus clash iniK > the cache. Page colouring means that the page fault handler endeavours to M > obtain a page from the free list such that this is unlikely to happen, for tM > instance by using a physical page whose relevant address bits are identicaleK > to the virtual address being mapped. The VMS page fault handler currentlypF > operates FIFO on the free page list and is oblivious to this effect.   Best explanation I've seen so far.  Thanks.  (Some folks like to toss page coloring out as a buzzword, and can't quite explain what they mean.)m   -- n Robert Deininger rdeininger@mindspring.como   ------------------------------  % Date: Fri, 21 Jul 2000 16:49:22 +0100u& From: Roy Omond <Roy.Omond@compaq.com>: Subject: Re: Mlucas performance probelm - need expert help* Message-ID: <39787101.C372FAFE@compaq.com>   Jan Vorbrueggen wrote:  6 > rdeininger@mindspring.com (Robert Deininger) writes: >eF > > The problem I see is this.  If I stop the program and then restartF > > it on the same system, it gets faster or slower by as much as 10%.: > > And stays that way forever (until I restart it again). > L > You are seeing the effects of page colouring or, rather, the fact that theM > VMS page fault handler doesn't do page colouring and DUNIX's does. (Neither=M > does LinuxALpha, at the moment). I think adding it is on the VMS to-do-list,M > (Hoff?!). Do the following: modify the program to run a few iterations (for L > a few minutes, say), and if the CPU time is longer than a machine-depedent( > value, write a checkpoint and restart. >-P > I've heard a report of a factor of two slowdown for one particular application3 > which was solved by doing what I described above." >s
 >         Jan- > L > PS: So what is page colouring? The L2/L3 caches of the 21154 and 21264 areK > larger than the page size multiplied by their associativity, and they usetK > physical addresses. It thus can happen that two pages that wouldn't see acM > cache conflict when using virtual addresses happen to be mapped to physicalfM > addresses that are identical in the relevant address bits and thus clash inpK > the cache. Page colouring means that the page fault handler endeavours tooL > obtain a page from the free list such that this is unlikely to happen, forM > instance by using a physical page whose relevant address bits are identicalsK > to the virtual address being mapped. The VMS page fault handler currentlyuF > operates FIFO on the free page list and is oblivious to this effect.  > Mmmm... VMS's page colouring is set to "monochrome" by default5 (SYSGEN parameter PFN_COLOR_COUNT).  You *can* set it H to be colourful;  if you do, please take ZERO_LIST_HI into consideration also !).  D There's no magic formula for the best settings for these parameters;A you'll have to test combinations until you find the best for yourw9 particular application.  Unfortunately PFN_COLOR_COUNT is./ not dynamic, so each change requires rebooting.   	 Roy Omonde Blue Bubble Ltd., (certainly not speaking on behalf of Compaq)                                 .    ------------------------------   Date: 21 Jul 2000 08:15:31 GMT= From: "$B?<C+(J $B0l9,(J" <kazuyuki.fukaya@jp.compaq.com>n' Subject: OpenVMS Registry Configurationo? Message-ID: <01bff2ea$111f5b80$3067a310@ngoequinox.ngo.dec.com>   	  Hello !!lK Does anybody why I got the following message while trying to list a certain  Key ? 0 I just installed OpenVMS and Advance Server 7.2.( I am trying to install DCOM for OpenVMS.  / ----------------------------------------------- ) REG> list key hkey_local_machine\software:  0 Key name:            HKEY_LOCAL_MACHINE\SOFTWARE' Security policy:     REG$K_POLICY_NT_40m Volatile:            REG$K_NONEd, Last written:        27-JUN-2000 14:16:01.75  
 Subkey(s):  "     Key name:            Microsoft+     Security policy:     REG$K_POLICY_NT_40h#     Volatile:            REG$K_NONEs0     Last written:        27-JUN-2000 14:10:42.42  4     Key name:            Compaq Computer Corporation+     Security policy:     REG$K_POLICY_NT_40+#     Volatile:            REG$K_NONEz0     Last written:        27-JUN-2000 14:10:42.42        Key name:            Classes+     Security policy:     REG$K_POLICY_NT_40y#     Volatile:            REG$K_NONEo0     Last written:        27-JUN-2000 16:06:35.82  '     Key name:            DCOMCNFGFORVMSe+     Security policy:     REG$K_POLICY_NT_40 #     Volatile:            REG$K_NONEr0     Last written:        27-JUN-2000 14:16:01.753 REG> list key hkey_local_machine\software\MicrosofttK Key HKEY_LOCAL_MACHINE\SOFTWARE\MICROSOFT, %REG-E-INVPATH, invalid key path   %REG-E-INVPATH, invalid key path  . ----------------------------------------------	  Regards,l Kazuyuki   ------------------------------  % Date: Fri, 21 Jul 2000 17:30:11 +0200 / From: klaus <klaus@kmb-715.ik.mb.uni-siegen.de> & Subject: Printer installation problems: Message-ID: <39786C83.BA899A9@kmb-715.ik.mb.uni-siegen.de>   Hallo,F I cannot install my printer correctly in VMS. The printer is a HPDesk= Jett 890C.sD If you could offer me any source of information or probably knows it	 yourself,P4 I would be very thankfull. Please send the answer to ralfenglisch@web.de.F If you just send the message back from where it came from(pressing th= esF answer button), it might get lost, because I=B4am writing this mail f= romm an official computerF of the institute of our university(University of Siegen, Germany) whe= re I work right now.h     Regards,  
 Ralf Englischs   ------------------------------  % Date: Fri, 21 Jul 2000 02:10:32 -0700r1 From: Vance Haemmerle <vance@toyvax.Tucson.AZ.US>r- Subject: Re: Question on MX mailer relay file 3 Message-ID: <3977B118.6C32D369@toyvax.Tucson.AZ.US>o   Lawrence Bleau wrote:c  , > Well, I tried just the single disable line > # > disable 0.0.0.0 with mask 0.0.0.00   How about trying   enable <your.ip.address>! disable 0.0.0.0 with mask 0.0.0.0s  7   I think it's testing both the source and destination.r   From SMTP_SERVER.B32: P                         !---- Retrieve the receiver IP addresses and check them.5                         !---- This is the last check.y   -- Vance Haemmerlet vance@toyvax.Tucson.AZ.US1   ------------------------------  # Date: Fri, 21 Jul 2000 11:35:59 GMT , From: "J. Scott Greig" <jsgreig@geminaq.com>6 Subject: Re: Quick answer time.... how fast is a TK70?4 Message-ID: <zAWd5.5034$GO4.36138@quark.idirect.com>  = "Charles Gilley" <gilley@nospam.bravewc.com> wrote in message09 news:lKPd5.12246$2M5.438785@news-west.usenetserver.com... L > Assuming the system can keep the TK70 busy, what kind of MB/minute might I	 > expect?8 >  >r  B Theorectical maximum throughput is 125 KB/sec, so (if K= 1024, and
 M=1,000,000),  the maximum is 7.68MB/minute.    Scotto   ------------------------------  % Date: Fri, 21 Jul 2000 09:09:48 +0200p, From: "Bart Zorn" <B.Zorn@TrueBit.nospam.nl>- Subject: Re: Removing <LF><CR> from text filee* Message-ID: <8l8t1j$agm$1@buty.wanadoo.nl>  ? Yes, but unfortunately, "real" keyboards don't have an ESC key!t   You can use ctrl - [ instead.   E The command is a bit dangerous, as ALL occurances of <cr><lf> will be ? removed, including the implicit ones's at the end of each line!h   Regardsi  	 Bart Zornt  F "Larry Kilgallen" <Kilgallen@eisner.decus.org.nospam> wrote in message% news:Q4WHBtxVQAxC@eisner.decus.org...(L > In article <5D2691D2D460D3118C0B009027719614964CEF@upiex1.uhcolorado.edu>,4 Monday Bill <Bill.Monday@UPI.UHColorado.Edu> writes: >rE > > I give up because I am under a deadline.  What's a quick and easyl2 > > way to remove the <CR><LF> s from a text file? >g
 > In TECO: >a > $ MCR TECO32 MYFILE.DATF > <n > $;fr$>ex$$ > 9 > Where the last 4 $ characters represent the Escape key.i   ------------------------------  % Date: Fri, 21 Jul 2000 09:44:36 +02004. From: Marcin Szczecinski <marcin@lodz.tpsa.pl>- Subject: Re: Removing <LF><CR> from text filet/ Message-ID: <20000721094436.B8250@lodz.tpsa.pl>a  G > > > I give up because I am under a deadline.  What's a quick and easy 4 > > > way to remove the <CR><LF> s from a text file?3 I am using zip/unzip (by Infozip) for this purpose.nH Zip file (check "-l" and "-ll" options) and unzip as text (yoy can force? it with "-a" option). Also fine when moving text files from VMSu to other platforms.r  = Marcin Szczecinski                        marcin@lodz.tpsa.plv   ------------------------------  % Date: Fri, 21 Jul 2000 10:29:53 +0200u= From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>n- Subject: Re: Removing <LF><CR> from text file ) Message-ID: <39780A00.C8926FEF@gtech.com>    Monday Bill wrote:K >         I give up because I am under a deadline.  What's a quick and easy 0 > way to remove the <CR><LF> s from a text file?I > I know how to add them in EDT, but how do you find/replace them or justtM > delete them?  What other VMS options are available if editing isn't the wayt
 > to go......    $ EDIT/TPU file-name <DO>' TPU EVE_REPLACE(ASCII(13)+ASCII(10),"")  <RET>t Aa  > It can be done much more elegant, but this one is easy to use.   Arne   ------------------------------    Date: 21 Jul 2000 08:43:17 -05009 From: Kilgallen@eisner.decus.org.nospam (Larry Kilgallen) - Subject: Re: Removing <LF><CR> from text file + Message-ID: <kiuQqu$wQLc8@eisner.decus.org>V  Y In article <8l8t1j$agm$1@buty.wanadoo.nl>, "Bart Zorn" <B.Zorn@TrueBit.nospam.nl> writes:eA > Yes, but unfortunately, "real" keyboards don't have an ESC key!   B They do once you change the keyboard setup (DECterm settings for a
 workstation).h   ------------------------------  % Date: Fri, 21 Jul 2000 10:59:50 +0100s/ From: Nigel Arnot <sysmgr@maxwell.ph.kcl.ac.uk>m# Subject: RE: Sun may GPL StarOfficee7 Message-ID: <009ED68B.41744617.17@maxwell.ph.kcl.ac.uk>m   > C > Any truth to the rumor that the release's internal name is Jason?n >     He who stole the golden fleece?   ) Wouldn't "David" be better? Or "Sampson"?o   	Yours,m
 		Nigel Arnot - 		NRA@MAXWELL.PH.KCL.AC.UK                   y  7 		"In the beginning there was nothing, which exploded."t   ------------------------------   Date: 21 Jul 2000 10:36 CSTt' From: carl@gerg.tamu.edu (Carl Perkins)k# Subject: RE: Sun may GPL StarOffice - Message-ID: <21JUL200010363867@gerg.tamu.edu>t  k In article <009ED68B.41744617.17@maxwell.ph.kcl.ac.uk>, Nigel Arnot <sysmgr@maxwell.ph.kcl.ac.uk> writes...r }> vD }> Any truth to the rumor that the release's internal name is Jason? }> r } ! }He who stole the golden fleece? P }	Yours, }		Nigel Arnot  D I expect not. Probably more like "he who wears hockey mask and killsF teenagers, and anybody else at hand, in Friday the 13th movie series".   --- Carl   ------------------------------   Date: 21 Jul 2000 17:24:43 GMT1 From: bill@triangle.cs.uofs.edu (Bill Gunshannon)u# Subject: RE: Sun may GPL StarOffice , Message-ID: <8la10r$14e9$1@info.cs.uofs.edu>  7 In article <009ED68B.41744617.17@maxwell.ph.kcl.ac.uk>,o2  Nigel Arnot <sysmgr@maxwell.ph.kcl.ac.uk> writes: |> > .F |> > Any truth to the rumor that the release's internal name is Jason? |> > e |> s# |> He who stole the golden fleece?   |> w, |> Wouldn't "David" be better? Or "Sampson"? |> e  8 I think he meant Jason with the 666 tattoo on his head!!   bill   -- eJ Bill Gunshannon          |  de-moc-ra-cy (di mok' ra see) n.  Three wolvesD bill@cs.scranton.edu     |  and a sheep voting on what's for dinner. University of Scranton   |A Scranton, Pennsylvania   |         #include <std.disclaimer.h>   s   ------------------------------  # Date: Fri, 21 Jul 2000 06:18:43 GMTa( From: Rainer Giese <giese_r@my-deja.com> Subject: Re: SYSUAF.DATs) Message-ID: <8l8pvv$r22$1@nnrp1.deja.com>c   You can do some things like :r   $ open uaf sysuaf.datS# $ read/end=no_more_records uaf line30 $ username = f$edit(f$extract(4,32,line),"trim")0 $ account = f$edit(f$extract(52,32,line),"trim")= $ owner = f$extract(85,f$cvui(0,8,f$extract(84,1,line)),line)e8 $ mem = f$fao("!OW",f$cvui(0,16,f$extract(36,2,line)))/18 $ grp = f$fao("!OW",f$cvui(0,16,f$extract(38,2,line)))/1C $! ... and now try something on the exchange server via rsh and thea exchange import tool  E We create exchange accounts from VMS on this way. Of course you can'te get the password of a user !   Rainer Giese      & Sent via Deja.com http://www.deja.com/ Before you buy.b   ------------------------------    Date: 21 Jul 2000 08:44:10 -05009 From: Kilgallen@eisner.decus.org.nospam (Larry Kilgallen)s Subject: Re: SYSUAF.DATs+ Message-ID: <kns$lTdyFiPZ@eisner.decus.org>   T In article <8l8pvv$r22$1@nnrp1.deja.com>, Rainer Giese <giese_r@my-deja.com> writes: > You can do some things like :c >  > $ open uaf sysuaf.dat   < Don't try that on a live system, as the lack of sharing will prevent people from logging in.n   ------------------------------  % Date: Fri, 21 Jul 2000 10:26:02 -0400@, From: Scott McNeilly <smcneilly@bridgew.edu> Subject: RE: SYSUAF.DATn2 Message-ID: <009ED686.88511DF5.7@fred.bridgew.edu>  B We have been using a BASIC program written by Brian Lomasky called@ SYSUAF for years.  It's been so many years that I have forgotten> where I got it.  Maybe some one else remembers.  But you might try ftp.spc.edu or ftp.wku.edu.a    @ >From: "Carlos A. =??Q?Jim=E9nez?=" <C_JIMENEZ@UPR1.UPR.CLU.EDU> >e >Hello >e- >        Hello where I can get the sysuaf.datLP >information(username,password,group, etc.) using a priv account to export theseG >values to server echange(sam file MS) using a priv account in echange.  > N >Because we move from alphaserver to echange server to manage the mail and dns >utilities.o >  >  >Att.. >  >Carlos JimenezI >Alpha Manager> --------------------------------------------------------------; Scott Mc Neilly               email:  smcneilly@bridgew.edus2 Assistant Director            Phone:  508-697-12362 Information Services          FAX:    508-697-1774 Bridgewater State College  Bridgewater, MA 02325r? ---------------------------------------------------------------n   ------------------------------    Date: 21 Jul 2000 09:04:45 -0500 From: briggs@eisner.decus.orgTD Subject: Re: VMS Com file to write a Form Feed Character --- THANKS!+ Message-ID: <SLJSaBL0nw7+@eisner.decus.org>   X In article <s975c921.016@hmcpsugwia.hmc.psu.edu>, Scott Brooks <SBROOKS@psu.edu> writes:N > Thanks for your answers to my question on how to get a form feed character =* > into an output file via a DCL com file!! >  > form_feed = f$fao("!^"). > 3 > worked for me and gave me the two Fs in Edit/TPU.o > J > I tried each of the following.  But they all seemed to give me a 12 in =9 > Edit/TPU.  So I'm not sure why they didn't work for me.e >  > form_feed[0,7] = 12  > form_feed[0,8] = 12a > form_feed[0,7] = %X0Ce  H The behavior of the  $ symbol[start,length]=value  syntax depends on the> type of the target symbol onto which the new data is overlaid.  : $ form_feed=""			! Creates a symbol with a string datatype> $ form_feed[0,8] = 12		! Sets the first 8 bits to the value 12 $ show sym form_feed*   FORM_FEED = "."		! It's a real form feed  ; $ form_feed=0			! Creates a symbol with an integer datatyped> $ form_feed[0,8] = 12		! Sets the first 8 bits to the value 12 $ show sym form_feed6   FORM_FEED = 12   Hex = 0000000C  Octal = 00000000014   And just for completeness...  : $ form_feed = 256		! A one in the ninth bit (bit number 8)> $ form_feed[0,8] = 12		! Sets the first 8 bits to the value 12 $ show sym form_feed7   FORM_FEED = 268   Hex = 0000010C  Octal = 00000000414   % 	John Briggs		briggs@eisner.decus.orgd   ------------------------------    Date: 21 Jul 2000 09:14:15 -0500 From: briggs@eisner.decus.orgt) Subject: Re: VMS equivalent of UNIX 'ln'?h+ Message-ID: <1TtRkvz5rd5N@eisner.decus.org>   ^ In article <sne5n7sa6tt58@corp.supernews.com>, "Peter Weaver" <peter.weaver@stelco.ca> writes:B > Thomas Dickey wrote in message <8l4kim$lhc$1@news1.Radix.Net>... > @ >>If there's no clue that an entry is a link, it's probably more	 > trouble , >>to use than you would want to put up with. > D > You can tell if an entry is a link since the file name in the file7 > header will point to the original file, not the link.   B That's true for links within a single directory.  But the rule for* cross-directory links has an extra caveat.  C    If the directory back-link in the file header does not match the D    directory in which the file's directory entry was found then that    entry is an alias.o  & 	John Briggs			briggs@eisner.decus.org   ------------------------------   Date: 21 Jul 2000 13:43:15 GMT/ From: Thomas Dickey <dickey@saltmine.radix.net>e) Subject: Re: VMS equivalent of UNIX 'ln'?a) Message-ID: <8l9k1j$r5$1@news1.Radix.Net>   3 Hoff Hoffman <hoffman@xdelta.zko.dec.nospam> wrote:s  E >   Um, you said you wanted a "reliable way to tell how many times a aF >   file's been entered", and that is available.  You did not specify 1 >   any particular performance requirements.  :-)   C ( well, implicit in that was the suggestion that it run in a timelyRC   fashion so that I don't have to worry about when the system might0   next be rebooted ;-)  I I knew I could use the system tools, but wanted to know if there was someiA library interface or data structure that I might have overlooked.    -- s= Thomas E. Dickey <dickey@radix.net> <dickey@herndon4.his.com>  http://dickey.his.comt ftp://dickey.his.com   ------------------------------  % Date: Fri, 21 Jul 2000 10:18:59 +0200i= From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>n% Subject: Re: Web browser and VMS mail() Message-ID: <39780773.118314E8@gtech.com>.   Mark Daniel wrote:O > Why build it, when I supply object modules allowing a link-only installation?   = I would expect 80-90% of VMS system-managers to want to buildi+ from sources even if objects are available.i  ; Objects are a (good) service to sites with no (C) compiler.c   Arne   ------------------------------   Date: 21 Jul 2000 09:27 CST ' From: carl@gerg.tamu.edu (Carl Perkins) % Subject: Re: Web browser and VMS mailo- Message-ID: <21JUL200009274061@gerg.tamu.edu>a  3 Mark Daniel <Mark.Daniel@wasd.vsm.com.au> writes...oK }> If I was using more recent versions of C and VMS, I expect it would havea) }> worked without the extensive fiddling.w }>S }> %SYSTEM-F-ACCVIO, access violation, reason mask=00, virtual address=00000000, PCc }> =80443A64, PS=0000001Bb2 }> %TRACE-F-TRACEBACK, symbolic stack dump followsP }>  Image Name   Module Name     Routine Name    Line Number  rel PC      abs PCQ }>                                                         0 80443A64    80443A64hQ }>                                                         0 805E3BB4    805E3BB4eQ }>                                                         0 805E470C    805E470ClQ }>                                                         0 805E61F0    805E61F0fQ }>  TMP          TMP             main                    681 000000C4    000200C4yQ }>  TMP          TMP             __main                    0 00000064    00020064rQ }>                                                         0 83CCFA50    83CCFA50u } L }A convincing demo.  I gues I should then ask, where was I giving it a NULL? } N }Why build it, when I supply object modules allowing a link-only installation? }  }Cheers, Mark.  M (BTW - anyvbody following this to see how easy/hard it is to implement really-N should keep in mind that I am using a version of both VMS and C that are olderH than the ones that are listed as working. If you use newer verions. like5 the ones it says it works with, it should just work.)u  K Unfortuantely, the link-only version didn't work. It is, of course, linkingw1 to the C RTL which produces the errors mentioned.a  H The original symptom was that the Javascript sent for the first page wasC bad, triggering an error message. The part that was supposed to be:   %         function doRange (rangeURL) { &            location.replace(rangeURL);            return false;	         }4            function doAddrList () {L            listURL = document.formPage.action + "?" + "act=address%20LISTS";	         }M  ! was showing up in the browser as:G  %         function doRange (rangeURL) {             location%20LISTS"; 	         }U  H which is just plain wrong. Over 100 bytes have gone missing. The initialE complaint by the browser was the missing definition of the doAddrListRD function, as I recall. I think it also mentioned the syntax probelms8 in the messed up doRange, but not until it was executed.  E The majority of the NULL sting problems actually occured in the debug F printf statements. Unfortuantely, this means that when you turn on theD debugging to find out what is going on, the problem changes location  to happen sooner and gets worse.  F NULL is passed as a parameter all over the place in the CGILIB module.K Things like: "UrlEscapedPtr = CgiLibHtmlEscape (UrlOriginal, -1, NULL, 0);"dD where that 4th parameter is a pointer to char data type. A statementI like "return (CgiLib__VarList (VarName, NULL, NonePtr));" appears several-H times: CgiLib__VarList calls with one or more NULL parameters show up 35D times in cgilib.c, and all 3 parameters for it are pointers to char.  K Once I split the one fprintf into two, and figured out what the NULL stringl1 problem was, it didn't take all that long to fix.s  G Actually, I just remembered another minor problem. The version of DEC ClJ that I used had no mailmsgdef.h (either in SYS$SYSROOT:[DECC$LIB...] or inF a TLB). Making one from one of the other langage versions was easy - II extracted the values for the undefined constants that yahMAIL needed fromaH the Fortran definitions (I didn't do the whole thing, just the values it needed).   The files I modified are:a cgilib.c other.ck	 private.ch yahmail.c and .h  L Which reminds me - I also created a few function prototypes where there wereF none, and modified several that had no arguments (sticking making themJ foo(void) instead of foo() - there are still a bunch that /WARN=ENABLE=ALL complains about, though).e  E I probably would have modified public.c too, but public mail archivesa( are currently unused and untested here.   ) I may have done some more fiddling too...    --- Carl   ------------------------------   Date: 21 Jul 2000 10:07 CSTi' From: carl@gerg.tamu.edu (Carl Perkins) % Subject: Re: Web browser and VMS mailc- Message-ID: <21JUL200010073428@gerg.tamu.edu>   = "Antony Wardle" <antony.wardle@nnnoospam.met.co.nz> writes...t }> }> Web server: OSU }> Mail software: yahMAIL- }> SSL: OpenSSLe } M }Not sure about ssl. How do I know if I am using it? The others are the same.j }  }Antonyk  M The easy way: If you can use a URL that starts with "https" instead of "http"r3 then you are running SSL. If not, then you are not.o  E The OSU server does not come with SSL. You have to go out and get it,cD build it, then rebuild the server to use it which involved replacingD a file or two and changing a build parameter or two and changing theH startup command slightly (adding the 5th parameter to specify port 443).E There are various places to look for information on this, including au list of places at:) http://www.levitte.org/~ava/vms_doc.htmlxm7 For adding SSL, the single most useful URL I found was:s# http://www.free.lp.se/openssl/docs/>' (which is on the list mentioned above).r  L (I should have kept notes on how I did it, but I didn't. Oops. This could beL trouble the next time I upgrade anything - I'll have to figure it out again./ Fortunately, it wasn't particularly difficult.)a   --- Carl   ------------------------------    Date: 21 Jul 2000 09:51:35 +0200G From: Jan Vorbrueggen <jan@mailhost.neuroinformatik.ruhr-uni-bochum.de>2< Subject: Re: Where can I get older version of Apache for VMSH Message-ID: <y44s5j53aw.fsf@mailhost.neuroinformatik.ruhr-uni-bochum.de>  8 Does it _have_ to be Apache? Why not use the OSU server?   	Jan   ------------------------------  % Date: Fri, 21 Jul 2000 10:25:52 +0200.= From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>e< Subject: Re: Where can I get older version of Apache for VMS) Message-ID: <3978090F.3B6D4EBF@gtech.com>m   Thomas Wirt wrote:J > I am desperate to get a version of Apache up on my VMS 7.1 system beforeJ > the boss makes me go with a Wintel Web Server.  Does anyone know where II > can get a version of Apache that might work with UCX 4.2.  I have tried>G > the new Apache 1.3.12 beta, but it will not run on my system.  I havenG > heard rumors that 1.3.9 ran on UCX 4.2.  I can upgrade my TCP latter,eI > but if the Web server for our intranet goes Wintel, I will never get itrA > back on VMS.  This is only needed for proof of concept, not ford > production use.-  G Why go for a beta Apache that requires latest VMS version, when you canrE install OSU or WASD that has been running in production for many manym+ years and will work with your VMS version ?3   Arne   ------------------------------   Date: 21 Jul 2000 13:08:09 GMT2 From: hoffman@xdelta.zko.dec.nospam (Hoff Hoffman)< Subject: Re: Where can I get older version of Apache for VMS6 Message-ID: <8l9hvp$ea0$1@mailint03.im.hou.compaq.com>  Q In article <8l80u4$bun$1@nnrp1.deja.com>, Thomas Wirt <twirt@kittles.com> writes:aI :I am desperate to get a version of Apache up on my VMS 7.1 system before+I :the boss makes me go with a Wintel Web Server.  Does anyone know where IeH :can get a version of Apache that might work with UCX 4.2.  I have triedF :the new Apache 1.3.12 beta, but it will not run on my system.  I haveF :heard rumors that 1.3.9 ran on UCX 4.2.  I can upgrade my TCP latter,H :but if the Web server for our intranet goes Wintel, I will never get it@ :back on VMS.  This is only needed for proof of concept, not for :production use.  G   OSU and Netscape FastTrack web servers are available on this release.s  3   You can also get to TCP/IP V5.0A on this release.-  E   I would not encourage use of an older release, as we found and put i@   in place these prerequisites for specific technical reasons...  N  --------------------------- pure personal opinion ---------------------------L    Hoff (Stephen) Hoffman   OpenVMS Engineering   hoffman#xdelta.zko.dec.com   ------------------------------   Date: 21 Jul 2000 12:18:27 CDT= From: wayne@tachysoft.xxx.293778.killspam.0223 (Wayne Sewell) < Subject: Re: Where can I get older version of Apache for VMS. Message-ID: <at1st9vEVuBB@tachxxsoftxxconsult>  i In article <3978090F.3B6D4EBF@gtech.com>, Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com> writes:c > Thomas Wirt wrote:K >> I am desperate to get a version of Apache up on my VMS 7.1 system beforecK >> the boss makes me go with a Wintel Web Server.  Does anyone know where ItJ >> can get a version of Apache that might work with UCX 4.2.  I have triedH >> the new Apache 1.3.12 beta, but it will not run on my system.  I haveH >> heard rumors that 1.3.9 ran on UCX 4.2.  I can upgrade my TCP latter,J >> but if the Web server for our intranet goes Wintel, I will never get itB >> back on VMS.  This is only needed for proof of concept, not for >> production use. > I > Why go for a beta Apache that requires latest VMS version, when you can$G > install OSU or WASD that has been running in production for many manyc- > years and will work with your VMS version ?.  M He's already running with one of those (he says he will never get the companyaP web server *back* on vms, which implies he's already running a vms web server). G The problem is that the clueless boss has never heard of those, but hash probably heard of Apache.   N Any web server running on vms, beta or not, would be better than spending your2 life trying to keep a web server up on a billybox.     -- hO ===============================================================================#M Wayne Sewell, Tachyon Software Consulting  (281)812-0738  wayne@tachysoft.xxx2: http://www.tachysoft.xxx/www/tachyon.html and wayne.html  K change .xxx to .com in addresses above, assuming you are not a spambot  :-) O ===============================================================================cO Otter, on dining with Bluto:"It's perfectly safe if you keep your arms and legsd 			away from his mouth."   ------------------------------    Date: 21 Jul 2000 08:41:01 -05009 From: Kilgallen@eisner.decus.org.nospam (Larry Kilgallen) : Subject: Re: [Fwd: FW: Wish List:  I/O vs. CPU Priorities]+ Message-ID: <Nk5dK4juj7Zj@eisner.decus.org>r  g In article <3977BABA.C068C64F@earthlink.net>, "David J. Dachtera" <djesys.nospam@earthlink.net> writes:o > Larry Kilgallen wrote:  @ >> I don't understand the tentative nature of these discussions. >> /G >> As I read EXE_STD$INSERT_IRP from module SYSQIOREQ it orders entries J >> according to IRP$B_PRI.  Is this somehow disabled on contemporary VMS ? > G > I don't know that it's disabled, but it's far less effective than onea > typically expects.  D And is this with a disk/controller that accepts multiple outstandingE IO requests at a time ?  "Smart" hardware can conspire with user code  to defeat VMS priority.k   ------------------------------  # Date: Fri, 21 Jul 2000 17:13:20 GMT-- From: "Richard D. Piccard" <piccard@ohio.edu>0: Subject: Re: [Fwd: FW: Wish List:  I/O vs. CPU Priorities]( Message-ID: <397884B0.96791A88@ohio.edu>  D Just as we can specify a maximum amount of outstanding buffer space,E might we, as system managers, be able to specify a maximum direct (oraH buffered) I/O request RATE, either for a username, for a batch queue, orH by set process/whatever?  I have situations where I don't want to stop aH job cold, but I also don't want to let it soak up the bandwidth it would: otherwise use, in order to have some left for the others.    					RDP   Larry Kilgallen wrote: > i > In article <3977BABA.C068C64F@earthlink.net>, "David J. Dachtera" <djesys.nospam@earthlink.net> writes:o > > Larry Kilgallen wrote: > B > >> I don't understand the tentative nature of these discussions. > >>I > >> As I read EXE_STD$INSERT_IRP from module SYSQIOREQ it orders entries-L > >> according to IRP$B_PRI.  Is this somehow disabled on contemporary VMS ? > >iI > > I don't know that it's disabled, but it's far less effective than one  > > typically expects. > F > And is this with a disk/controller that accepts multiple outstandingG > IO requests at a time ?  "Smart" hardware can conspire with user codee > to defeat VMS priority.0   --  B ==================================================================B Dick Piccard                           Academic Technology ManagerB piccard@ohio.edu                                 Computer ServicesB http://oak.cats.ohiou.edu/~piccard/                Ohio University   ------------------------------   End of INFO-VAX 2000.405 ************************