1 INFO-VAX	Fri, 08 Dec 2000	Volume 2000 : Issue 685       Contents:, Re: ??== Allocating only one block per file., Re: ??== Allocating only one block per file., Re: ??== Allocating only one block per file., Re: ??== Allocating only one block per file., Re: ??== Allocating only one block per file., Re: ??== Allocating only one block per file.9 Re: advice sought on transparent failover / IP clustering 9 Re: advice sought on transparent failover / IP clustering 9 Re: advice sought on transparent failover / IP clustering 9 Re: advice sought on transparent failover / IP clustering 9 Re: advice sought on transparent failover / IP clustering 9 Re: advice sought on transparent failover / IP clustering 9 Re: advice sought on transparent failover / IP clustering 9 Re: advice sought on transparent failover / IP clustering  Decnet - Proxies not working  Re: Decnet - Proxies not working  Re: Decnet - Proxies not working  Re: Decnet - Proxies not working Email alias on DEC TCP/IP ( F$GETJPI and VMS 7.1, 7.2 BYTLM = BYTCNT! Fujitsu M2488E tape drive trouble 0 Here is my Xmas-Present: cdrecord for IDE-drives+ RE: Heresy (was Re: how to sort a CSV file)  Re: New OpenVMS  Education site  Re: New OpenVMS  Education site  Re: New OpenVMS  Education site  RE: New OpenVMS  Education site  Re: One world, one processor Re: One world, one processor Re: One world, one processor Re: One world, one processor Re: One world, one processor Re: OpenVMS Items  process creation Re: process creation* Recover data from file not closed properly. Re: Recover data from file not closed properly. Re: Recover data from file not closed properly Re: Sun Cluster  Re: Sun Cluster  Re: Sun Cluster - Re: Sun CSV Constitution Election Belly Laugh - Re: Sun CSV Constitution Election Belly Laugh # Re: Testers wanted for Spell update  TZ87N compatible media. > Use of vmstar to exchange tape between AXP 2000 and Tru64 Unix Re: Very weird DCL behavior  Re: Very weird DCL behavior  Re: Write out char$(9) in DCL   F ----------------------------------------------------------------------  " Date: Fri, 8 Dec 2000 07:02:55 GMT7 From: moroney@world.std.spaamtrap.com (Michael Moroney) 5 Subject: Re: ??== Allocating only one block per file. % Message-ID: <G58M8v.Iz@world.std.com>   . aus@vim.uni-wuerzburg.de (Hans M. Aus) writes:  I >Is it possible to allocate only one block per disk file without changing  >the cluster size on the disk?  B No.  The cluster size is the smallest nonzero size a file can haveK allocated to it.  You'll have to initalize a disk with a cluster size of 1.   Q >Our OVMS 7.2-1/Alpha receives approximately 1,000 files daily via a FTP server.    J Good news.  You can initalize a disk of any reasonable size with a clusterI size of 1 under 7.2-1.  On older versions of VMS, the largest disk with a $ cluster size of 1 was about 1/2 gig.  @ >1) These text files contain less than one block of information.J >2) The files need to be directly accessible by name from a program for at >least one year.  J This gives approximately 365,000+ files for this application.  I'd suggestE you also specify /HEADERS and /MAXFILES appropiately to reflect this.   J >Is it possible define a default file allocation to just one block for all( >those files received by the FTP server?  . Again, only by INITing a disk with /CLUSTER=1.  I >What are the negative consequences of initializing one disk to a cluster  >size of one or three blocks?   J Larger files are very likely to become fragmented as will the disk itself.H I'd suggest you keep larger files off this disk.  The disk fragmentation* won't matter if it's mostly 1 block files.  ) >Is three blocks a reasonable compromise?   E This will reduce disk and file fragmentation for larger files, at the ? expense that your 1 block files will consume 3 times the space.   ( >How many files can a directory contain?  A No limit, however performance suffers for very large directories. G There used to be a hard limit at 127 block directory files, but I think I this has been eliminated.  Your disk needs to have enough contiguous file + space for the directory files as they grow.   G Note, a better solution would be some sort of archival system, or maybe D a random access file with each record one of your 'files' and a max.G record size of whatever your largest file can be.  You can retrieve the F records by a "filename" key, and you won't have to worry about 1 block cluster size disks.    -Mike    ------------------------------    Date: 08 Dec 2000 10:40:38 +0100G From: Jan Vorbrueggen <jan@mailhost.neuroinformatik.ruhr-uni-bochum.de> 5 Subject: Re: ??== Allocating only one block per file. H Message-ID: <y4n1e7l089.fsf@mailhost.neuroinformatik.ruhr-uni-bochum.de>  D What can be done to store your data as single files has already beenD described, but another complication hasn't been mentioned yet: EveryD file requires at least one block in the index file! Thus, if you go B with the method of storing things in a file each, you will need at/ least _two_ blocks (in the best case) for each.   E Alternative: Create an indexed file with what you are currently using F as a file name as the key. Write your text as a variable-length recordI to that indexed file with a "file name" as the key, and a similar routine E to read it back. Should be trivial to substitute if your code is even  remotely structured.   	Jan   ------------------------------  % Date: Fri, 08 Dec 2000 12:00:28 +0100 , From: aus@vim.uni-wuerzburg.de (Hans M. Aus)5 Subject: Re: ??== Allocating only one block per file. D Message-ID: <aus-0812001200280001@wvia26.virologie.uni-wuerzburg.de>  G Many thanks for the informative responses. This part of our application B started last week and we are very busy rethinking the application.  ? Is it common practice to change the cluster size on OVMS disks?   L Our Disk: V99$DKB400 device type COMPAQ BD009122C6, Total blocks=17,773,524.  H After the first week, there are a total of 6058 files and 12,739/115,704D blocks in the FTP directory. Appending all these files into one fileJ occupies only 2309/2322 blocks. Obviously we need to think about some typeG of size reduction scheme, especially with data that is a few weeks old.   A Should we also think about reducing the cluster size on the disk?    --  B Cheers, Hans M. Aus, Wuerzburg, Germany,  aus@vim.uni-wuerzburg.de   ------------------------------  % Date: Fri, 08 Dec 2000 08:12:24 -0500 , From: Howard S Shubs <hshubs@mindspring.com>5 Subject: Re: ??== Allocating only one block per file. > Message-ID: <hshubs-8CE3FB.08122408122000@news.mindspring.com>  E In article <aus-0812001200280001@wvia26.virologie.uni-wuerzburg.de>,  - aus@vim.uni-wuerzburg.de (Hans M. Aus) wrote:   B >Should we also think about reducing the cluster size on the disk?  M Depends on what you end up doing, and what characteristics the files on your  
 disk require.   L For instance, a 400,000 block file isn't very big.  It's only about 200MB.  G What will you do with the rest of your (multi-gigabyte) disk?  My main   concerns would be   O    1) making sure the file was not open all the time, so it could be backed up   properlyD    2) making sure you can delete old information when the time comes --   Howard S ShubsD "Run in circles, scream and shout!"  "I hope you have good backups!"   ------------------------------  % Date: Fri, 08 Dec 2000 09:53:24 -0500 - From: "Richard D. Piccard" <piccard@ohio.edu> 5 Subject: Re: ??== Allocating only one block per file. ( Message-ID: <3A30F5E0.6D335537@ohio.edu>   IOlson@dairyworld.com wrote:   > >-----Original Message----- C > >From: aus@vim.uni-wuerzburg.de [mailto:aus@vim.uni-wuerzburg.de]  >    > [snip]   > > * > >How many files can a directory contain? > M > I'm not sure what the limit is, but that isn't what you should be concerned  > with anyway.M > I have heard a "rule of thumb" that a directory should have fewer than 1000 E > files in it. If there are more than that you will start to see some C > performance degradation when/if you create or delete files in it. J > Having said that, we've got a few directories with over 10000 files thatL > don't behave "too badly", and a few that are even bigger than that that DO > behave badly. I > I would be EXTREMELY reluctant to create directories with 400000 files. H > Can you sub-divide them into (eg) one day? one week?   Think about it. >  > >-- E > >Cheers, Hans M. Aus, Wuerzburg, Germany,  aus@vim.uni-wuerzburg.de  >  > Ingemar Olson , > IT Department, Sperling     (604) 444-7367 > Dairyworld FOODS  O The limit, as others have pointed out, is the size of the directory file in all L but the most recent versions of VMS.  The 1,000 files is a nice round numberI that can be used unless the filenames are unusually long.  For example, a E directory containing 1,000 files having filenames of lnk00000000.dat, O lnk00000001.dat, etc., is 59 blocks.  If the filenames are the full legal size, E even 1,000 files would be more than a 127-block directory could list.   +                                         RDP    --B ==================================================================B Dick Piccard                           Academic Technology ManagerB piccard@ohio.edu                                 Computer ServicesB http://oak.cats.ohiou.edu/~piccard/                Ohio University   ------------------------------   Date: 8 Dec 2000 16:21:19 GMT 2 From: mathog@seqaxp.bio.caltech.edu (David Mathog)5 Subject: Re: ??== Allocating only one block per file. , Message-ID: <90r1pv$m3k@gap.cco.caltech.edu>  g In article <3A305EB7.AEE5E40C@earthlink.net>, "David J. Dachtera" <djesys.nospam@earthlink.net> writes:  >"Hans M. Aus" wrote:  > R >> Our OVMS 7.2-1/Alpha receives approximately 1,000 files daily via a FTP server. >>  B >> 1) These text files contain less than one block of information.L >> 2) The files need to be directly accessible by name from a program for at >> least one year. >   K In addition to the librarian and ZIP API David Dachtera suggested he might  D also want to sequentially append the entries as single records to anH indexed file.  Map the original file name to the record key and they canG be retrieved easily enough in random order, at will, using the standard  RMS interface to such a file.    Regards,   David Mathog mathog@seqaxp.bio.caltech.edu ? Manager, sequence analysis facility, biology division, Caltech     ------------------------------   Date: 8 Dec 2000 11:48:10 GMT ' From: david20@alpha2.mdx.ac.uk (D.Webb) B Subject: Re: advice sought on transparent failover / IP clustering0 Message-ID: <90qhpq$h7j$1@aquila.news.mdx.ac.uk>   In article <009F43FB.F652F61D@SSRL04.SLAC.STANFORD.EDU>, winston@SSRL.SLAC.STANFORD.EDU ("Alan Winston - SSRL Admin Cmptg Mgr") writes: 
 >VMSers -- > N >So I have a cluster in which the two relevant systems are Alphaserver 800 andJ >2100 attached to an HSZ70, running VMS 7.2-1 and Multinet 4.3.  They run K >the Rdb database, OSU web server (although we might change to CSWS at some ' >point), PMDF for mail processing, etc.  > K >The sitewide DNS is not under my control, although I can request that they P >put systems in there for me.  (I'm not a BIND expert; I think they're running a >current version on Solaris.)  > H >I would like to define ip names for services rather than for nodes, eg I >mail-ssrl, www-ssrl, rdb-ssrl, and have those service names be responded H >to by whichever system is least heavily loaded, but, more importantly, M >if one of those systems goes down, by the one that remains up.  (I want this E >to be true whether it goes down for planned upgrade or for a crash.)  > H >A round-robin DNS entry, which I could easily get set up, would, if oneK >system died, send users to the dead system half the time.  I think I could K >the sitewide DNS to delegate the resolution of my service names to a DNS I I >maintained, but then I run the risk of my DNS - which obviously can't be 1 >one of my two main cluster nodes - going down.     - Why can't the DNS be one of your main nodes ? H You won't have a gigantic number of entries in it so the load should be K relatively light. I don't know anything about Multinet so you might want to M see whether you could have a "clusterwide primary DNS" (ie if the node acting L as the primary goes down another node takes over that function - reading theN same DNS files) or have one of the nodes as a primary and the other node as a 
 secondary.  L If you don't want the hassle of setting up your own sub-domain and DNS then M certainly with Dec TCPIP Services 5 then you can use your main DNS so long as H it is running a relatively recent version of BIND. The TCPIP load brokerI needs to be able to update the DNS dynamically as the load on the various > nodes changes or nodes are added and removed from the cluster.M DEC TCPIP services has in the past generally been behind Multinet and TCPWARE K in the provision of TCPIP features hence I would expect that Multinet would  also provide this capability.       >The DNS would try to keepL >track of the health of my main cluster nodes, and perhaps one of my clusterH >nodes keep track of the health of the DNS and take on its IP address asG >well if it goes down, meanwhile notifying someone who can do something K >about it, which I think I know how to do in Multinet if I  have a separate  >interface for it.)  >   H With DNS's this is generally done by having a Primary DNS (where all theH changes are made to the DNS information) and a number of secondary DNS'sJ (Which contain copies of the information and are updated whenever the mainI information changes). System's using the DNS should always have their DNS K resolver set to use at least two of these addresses. I believe the relevent L RFC specifies the resolver should always be able to accept 3 such addresses.L If you wish you could have one of your systems setup as a secondary for yourL main DNS. Though I am sure the people controlling your main DNS will alreadyL have a couple of secondarys (you would of course also need to make sure thatI relevent systems then had your secondary address setup in their resolvers , otherwise your secondary would not be used).      
 David Webb VMS and Unix team leader CCSS Middlesex University  K >Similarly, I was looking at load-balancing/QoS external boxes, which would I >essentially keep track of the state of both systems and pass requests to I >the right one, but it seems to me that this introduces a single point of L >failure to the system.  If my Zeus box dies, it doesn't matter that both my >regular systems are up. > K >How do you set up what are effectively IP cluster-alias service names that H >increase apparent reliability and uptime rather than taking the risk ofL >decreasing them?  I would of course be especially interested in ways to do 6 >it that don't involve spending any additional money.  > N >We're a CSLG site, so I could theoretically change to TCP/IP Services if thatN >made it easier to do this than Multinet does (although I've started using theO >SSH server that comes with Multinet 4.3 and like it, so I'd hate to do that; I 1 >have a lot of Multinet background and no UCX.)    > ) >Thanks in advance for any useful advice.  >  >-- Alan > P >===============================================================================1 > Alan Winston --- WINSTON@SSRL.SLAC.STANFORD.EDU N > Disclaimer: I speak only for myself, not SLAC or SSRL   Phone:  650/926-3056N > Physical mail to: SSRL -- SLAC BIN 69, PO BOX 4349, STANFORD, CA  94309-0210P >=============================================================================== >    ------------------------------  # Date: Fri, 08 Dec 2000 13:28:48 GMT L From: winston@SSRL.SLAC.STANFORD.EDU ("Alan Winston - SSRL Admin Cmptg Mgr")B Subject: Re: advice sought on transparent failover / IP clustering8 Message-ID: <009F4460.344E1501@SSRL04.SLAC.STANFORD.EDU>  Z In article <90qhpq$h7j$1@aquila.news.mdx.ac.uk>, david20@alpha2.mdx.ac.uk (D.Webb) writes: >In article <009F43FB.F652F61D@SSRL04.SLAC.STANFORD.EDU>, winston@SSRL.SLAC.STANFORD.EDU ("Alan Winston - SSRL Admin Cmptg Mgr") writes: [snip]  H >I would like to define ip names for services rather than for nodes, eg J >>mail-ssrl, www-ssrl, rdb-ssrl, and have those service names be respondedI >>to by whichever system is least heavily loaded, but, more importantly,  N >>if one of those systems goes down, by the one that remains up.  (I want thisF >>to be true whether it goes down for planned upgrade or for a crash.) >>I >>A round-robin DNS entry, which I could easily get set up, would, if oneoL >>system died, send users to the dead system half the time.  I think I couldL >>the sitewide DNS to delegate the resolution of my service names to a DNS IJ >>maintained, but then I run the risk of my DNS - which obviously can't be2 >>one of my two main cluster nodes - going down.   > . >Why can't the DNS be one of your main nodes ?  N Because if I'm using the DNS to make sure I don't give out the IP address of aN dead system in my cluster, the node it's on happens to be the one that's dead," it's not going to be very helpful.  $ Otherwise, thanks for your comments.   -- Alant    O ===============================================================================r0  Alan Winston --- WINSTON@SSRL.SLAC.STANFORD.EDUM  Disclaimer: I speak only for myself, not SLAC or SSRL   Phone:  650/926-3056rM  Physical mail to: SSRL -- SLAC BIN 69, PO BOX 4349, STANFORD, CA  94309-0210eO ===============================================================================    ------------------------------  % Date: Fri, 08 Dec 2000 09:56:18 -0600e0 From: Patrick Spinler <spinler.patrick@mayo.edu>B Subject: Re: advice sought on transparent failover / IP clustering( Message-ID: <3A3104A2.CBD8F66D@mayo.edu>   Alan:r  D I might consider your load balancing QoS solution, but with a twist:  B Disclaimer: it's awkward compared to decnet/lat, but it does work.  C Have 2 boxes for the load balancing gateway, with a heartbeat going D between them, and one box owning an "active" ip.  If that box fails,@ have the other do an automatic takover of that IP on one of it's interfaces.d  # Here's my attempt at bad ascii art:t  %                        load balancingi%                         sw runs here h%                        +------------+ %       incoming line    |box 1       |-,    +-------------------- eth0   eth1-------+,    | 10.0.0.1          |    eth2    |      |,    |                   +-----|------+      |,    |                         | dedicated   |D  --+                         | heartbeat   +--- to your real cluster,    |                         | circuit     |,    |                   +-----|------+      |,    |  (incoming line)  |    eth2    |      |,    + - - - - - - - - - - eth0   eth1-------+%      (inactive unless) |box 2       |'%      (heartbeat fails) +------------+t    E For an example of how to do this with something cheap, specifically ad linux box, check out:   D   http://www.LinuxVirtualServer.org/     for the load balancing, andA   http://www.linux-ha.org/               for the address takeovers	 functions   E AFAIK these projects are made to interact gracefully with each other.l  D You should be able to serve your cluster with cheap/surplus pc class	 hardware.    -- Pat    * Alan Winston - SSRL Admin Cmptg Mgr wrote: > \ > In article <90qhpq$h7j$1@aquila.news.mdx.ac.uk>, david20@alpha2.mdx.ac.uk (D.Webb) writes: > >In article <009F43FB.F652F61D@SSRL04.SLAC.STANFORD.EDU>, winston@SSRL.SLAC.STANFORD.EDU ("Alan Winston - SSRL Admin Cmptg Mgr") writes: > [snip] > I > >I would like to define ip names for services rather than for nodes, egaL > >>mail-ssrl, www-ssrl, rdb-ssrl, and have those service names be respondedJ > >>to by whichever system is least heavily loaded, but, more importantly,P > >>if one of those systems goes down, by the one that remains up.  (I want thisH > >>to be true whether it goes down for planned upgrade or for a crash.) > >>K > >>A round-robin DNS entry, which I could easily get set up, would, if oneeN > >>system died, send users to the dead system half the time.  I think I couldN > >>the sitewide DNS to delegate the resolution of my service names to a DNS IL > >>maintained, but then I run the risk of my DNS - which obviously can't be2 > >>one of my two main cluster nodes - going down. > >u0 > >Why can't the DNS be one of your main nodes ? > P > Because if I'm using the DNS to make sure I don't give out the IP address of aP > dead system in my cluster, the node it's on happens to be the one that's dead,$ > it's not going to be very helpful. > & > Otherwise, thanks for your comments. > 	 > -- Alano > Q > =============================================================================== 2 >  Alan Winston --- WINSTON@SSRL.SLAC.STANFORD.EDUO >  Disclaimer: I speak only for myself, not SLAC or SSRL   Phone:  650/926-3056.O >  Physical mail to: SSRL -- SLAC BIN 69, PO BOX 4349, STANFORD, CA  94309-0210sQ > ===============================================================================    --  ?       This message does not represent the policies or positionsa1 	     of the Mayo Foundation or its subsidiaries.y3   Patrick Spinler			email:	Spinler.Patrick@Mayo.EDU '   Mayo Foundation			phone:	507/284-9485i   ------------------------------   Date: 8 Dec 2000 11:40:36 -0500e* From: young_r@eisner.decus.org (Rob Young)B Subject: Re: advice sought on transparent failover / IP clustering+ Message-ID: <JqmVG5f30ttn@eisner.decus.org>n  = 	I was in a very similar situation.  The only cooperation youn= 	need from the owners of DNS is to have them add an entry fors; 	you.  Suppose you could call it cluster.slac.stanford.edu. B 	This entry would be a NS entry pointing to BOTH alphas.  Multinet? 	Cluster Service Name is easy to setup and is detailed on pages?= 	6-35,6-36 in the Multinet Administrator's Guide version 4.3.0  D 	You also mention cluster alias.  Unfortunately (in my opinion) thatA 	was a bad choice of a name.  The alias is "grabbed" by the firste= 	box that boots.  If that box crashes and another box is alsom@ 	setup for the cluster alias, it will then answer to that IP.  I> 	want to establish a cluster alias for data gateways that talk? 	to a port.  Currently, one box is hardwired as the talker.  In < 	theory (and practice) it would be better for the gateway toA 	talk to an IP that is an alias and if that box crashed, we woulde> 	simply restart the app on the node that grabbed the alias andA 	wouldn't be scrambling to make changes.  So the alias, does haveo 	nice uses.h  A 	Suggestion for a different name other than cluster alias?  Maybe & 	"permanent IP" or "floating address".  B 	Here is what use of cluster service name looks like for me, note:  K Cluster Service Names:          clusname1.sys.site.com,cluster.sys.site.coml     NAME1_YOUNGR.WORK> multi configh0 MultiNet Network Configuration Utility V4.1(103)= [Reading in MAXIMUM configuration from MULTINET:MULTINET.EXE]rF [Reading in configuration from MULTINET:NETWORK_DEVICES.CONFIGURATION] NET-CONFIG>showeC Interface                                Adapter     CSR Address   i Flags/VectorC ---------                                -------     -----------     ------------K se0      (Shared VMS Ethernet/FDDI)       -NONE-        -NONE-       -NONE-t:           [TCP/IP: 199.99.99.24, IP-SubNet: 255.255.255.0]2           [VMS Device: EXA0, Link Level: Ethernet]    2 Official Host Name:             name1.sys.site.com- Default IP Route:               199.99.99.254 C IP Cluster Aliases:             199.99.99.25  !Setup but not in use K Cluster Service Names:          clusname1.sys.site.com,cluster.sys.site.comr) Domain Nameservers:             127.0.0.1a+                                 199.99.99.4R+                                 199.99.99.2 , Local Domain:                   sys.site.com# Timezone:                       EST-* Timezone Rules:                 US/EASTERN9 Anonymous FTP Directory:        DISK$SITE_PAG:[ANONYMOUS]s$ Load UCX $QIO driver:           TRUE$ Load PWIP (Pathworks) driver:   TRUE/ WHOIS Default Server:           RS.INTERNIC.NETd   ---c  B 	Telnet to clusname1 or cluster puts me into the least loaded box:    & NAME1_YOUNGR.WORK> mu netc domain show- Connected to NETCONTROL server on "LOCALHOST" K < name1.sys.site.com Network Control V4.1(10) at Fri 8-Dec-2000 11:31AM-EST  < Service CLUSTER.SYS.SITE.COM:o% <   Nodename      Address      Ratinge% <   --------  ---------------  ------ $ <   NAME2      199.99.99.22      228$ <   NAME3      199.99.99.23      178$ <   NAME1      199.99.99.24      105! < Service CLUSNAME1.SYS.SITE.COM: % <   Nodename      Address      Ratingw% <   --------  ---------------  ------C$ <   NAME2      199.99.99.22      228$ <   NAME3      199.99.99.23      178$ <   NAME1      199.99.99.24      105
 < End of line-  M In this case, NAME2 answers to CLUSTER or CLUSNAME1.  Here is what an actual   nslookup looks like:  	 > cluster  Server:  LOCALHOST Address:  127.0.0.1i   Name:    CLUSTER.SYS.SITE.COMc4 Addresses:  199.99.99.23, 199.99.99.22, 199.99.99.24  I Note in the 2 minutes that went by, NAME3 is now least loaded.  Hope this0 helps.   				Rob0    8 In article <009F43FB.F652F61D@SSRL04.SLAC.STANFORD.EDU>,N winston@SSRL.SLAC.STANFORD.EDU ("Alan Winston - SSRL Admin Cmptg Mgr") writes: > VMSers --  > O > So I have a cluster in which the two relevant systems are Alphaserver 800 and=K > 2100 attached to an HSZ70, running VMS 7.2-1 and Multinet 4.3.  They run oL > the Rdb database, OSU web server (although we might change to CSWS at some( > point), PMDF for mail processing, etc. > L > The sitewide DNS is not under my control, although I can request that theyQ > put systems in there for me.  (I'm not a BIND expert; I think they're running a= > current version on Solaris.) > I > I would like to define ip names for services rather than for nodes, eg EJ > mail-ssrl, www-ssrl, rdb-ssrl, and have those service names be respondedI > to by whichever system is least heavily loaded, but, more importantly, aN > if one of those systems goes down, by the one that remains up.  (I want thisF > to be true whether it goes down for planned upgrade or for a crash.) > I > A round-robin DNS entry, which I could easily get set up, would, if one L > system died, send users to the dead system half the time.  I think I couldL > the sitewide DNS to delegate the resolution of my service names to a DNS IJ > maintained, but then I run the risk of my DNS - which obviously can't beK > one of my two main cluster nodes - going down.  The DNS would try to keepiM > track of the health of my main cluster nodes, and perhaps one of my clustereI > nodes keep track of the health of the DNS and take on its IP address as-H > well if it goes down, meanwhile notifying someone who can do somethingL > about it, which I think I know how to do in Multinet if I  have a separate > interface for it.) > L > Similarly, I was looking at load-balancing/QoS external boxes, which wouldJ > essentially keep track of the state of both systems and pass requests toJ > the right one, but it seems to me that this introduces a single point ofM > failure to the system.  If my Zeus box dies, it doesn't matter that both myn > regular systems are up.  > L > How do you set up what are effectively IP cluster-alias service names thatI > increase apparent reliability and uptime rather than taking the risk of M > decreasing them?  I would of course be especially interested in ways to do  7 > it that don't involve spending any additional money. r > O > We're a CSLG site, so I could theoretically change to TCP/IP Services if that O > made it easier to do this than Multinet does (although I've started using thenP > SSH server that comes with Multinet 4.3 and like it, so I'd hate to do that; I2 > have a lot of Multinet background and no UCX.)   > * > Thanks in advance for any useful advice. > 	 > -- Alanp > Q > ===============================================================================t2 >  Alan Winston --- WINSTON@SSRL.SLAC.STANFORD.EDUO >  Disclaimer: I speak only for myself, not SLAC or SSRL   Phone:  650/926-3056hO >  Physical mail to: SSRL -- SLAC BIN 69, PO BOX 4349, STANFORD, CA  94309-0210kQ > ===============================================================================o >    ------------------------------   Date: 8 DEC 2000 16:52:09 GMTe4 From: karcher@thuria.waisman.wisc.edu (Carl Karcher)B Subject: Re: advice sought on transparent failover / IP clustering5 Message-ID: <8DEC00.16520901@thuria.waisman.wisc.edu>d  d In a previous article, winston@SSRL.SLAC.STANFORD.EDU ("Alan Winston - SSRL Admin Cmptg Mgr") wrote:  0 ->>Why can't the DNS be one of your main nodes ? ->  P ->Because if I'm using the DNS to make sure I don't give out the IP address of aP ->dead system in my cluster, the node it's on happens to be the one that's dead,$ ->it's not going to be very helpful.  D But nothing is stoping you from running more than one primary in the@ cluster. We have a two node cluster and each node runs BIND as a; primary, sharing the same configuration file (pointed to byd TCPIP$BIND_SERVER_DATA).   ------------------------------   Date: 8 Dec 2000 18:08:25 GMTW' From: david20@alpha2.mdx.ac.uk (D.Webb)hB Subject: Re: advice sought on transparent failover / IP clustering0 Message-ID: <90r82p$nfa$1@aquila.news.mdx.ac.uk>   In article <009F4460.344E1501@SSRL04.SLAC.STANFORD.EDU>, winston@SSRL.SLAC.STANFORD.EDU ("Alan Winston - SSRL Admin Cmptg Mgr") writes:s[ >In article <90qhpq$h7j$1@aquila.news.mdx.ac.uk>, david20@alpha2.mdx.ac.uk (D.Webb) writes:I >>In article <009F43FB.F652F61D@SSRL04.SLAC.STANFORD.EDU>, winston@SSRL.SLAC.STANFORD.EDU ("Alan Winston - SSRL Admin Cmptg Mgr") writes:r >[snip]e >aI >>I would like to define ip names for services rather than for nodes, eg DK >>>mail-ssrl, www-ssrl, rdb-ssrl, and have those service names be respondedaJ >>>to by whichever system is least heavily loaded, but, more importantly, O >>>if one of those systems goes down, by the one that remains up.  (I want this G >>>to be true whether it goes down for planned upgrade or for a crash.)e >>>fJ >>>A round-robin DNS entry, which I could easily get set up, would, if oneM >>>system died, send users to the dead system half the time.  I think I couldaM >>>the sitewide DNS to delegate the resolution of my service names to a DNS IeK >>>maintained, but then I run the risk of my DNS - which obviously can't bew3 >>>one of my two main cluster nodes - going down.  e >>/ >>Why can't the DNS be one of your main nodes ?i > O >Because if I'm using the DNS to make sure I don't give out the IP address of aeO >dead system in my cluster, the node it's on happens to be the one that's dead,g# >it's not going to be very helpful.i >o% >Otherwise, thanks for your comments.h >y  I As I said before either have a cluster-aware primary DNS (If such a thinga5 exists) or setup a secondary on another cluster node.c  I However it is probably simpler to do dynamic updates to your existing DNSa servers.  
 David Webb VMS and Unix team leader CCSS Middlesex University   ------------------------------    Date: 09 Dec 2000 01:13:03 +0800/ From: Paul Repacholi <prep@k9.prep.synonet.com>dB Subject: Re: advice sought on transparent failover / IP clustering0 Message-ID: <87u28eltuo.fsf@k9.prep.synonet.com>  N winston@SSRL.SLAC.STANFORD.EDU ("Alan Winston - SSRL Admin Cmptg Mgr") writes:  0 > >Why can't the DNS be one of your main nodes ? > P > Because if I'm using the DNS to make sure I don't give out the IP address of aP > dead system in my cluster, the node it's on happens to be the one that's dead,$ > it's not going to be very helpful. > & > Otherwise, thanks for your comments. > 	 > -- AlanR  3 So run DNS on the CLUSTER, and ship the CLUSTER IP.L Not 'suported' but it works.   ~Paul4   ------------------------------   Date: 8 Dec 2000 18:31:16 GMT=' From: david20@alpha2.mdx.ac.uk (D.Webb)-B Subject: Re: advice sought on transparent failover / IP clustering0 Message-ID: <90r9dk$ns2$1@aquila.news.mdx.ac.uk>  l In article <8DEC00.16520901@thuria.waisman.wisc.edu>, karcher@thuria.waisman.wisc.edu (Carl Karcher) writes:e >In a previous article, winston@SSRL.SLAC.STANFORD.EDU ("Alan Winston - SSRL Admin Cmptg Mgr") wrote:. >u1 >->>Why can't the DNS be one of your main nodes ?S >-> Q >->Because if I'm using the DNS to make sure I don't give out the IP address of agQ >->dead system in my cluster, the node it's on happens to be the one that's dead, % >->it's not going to be very helpful.r >sE >But nothing is stoping you from running more than one primary in themA >cluster. We have a two node cluster and each node runs BIND as as< >primary, sharing the same configuration file (pointed to by >TCPIP$BIND_SERVER_DATA).o  I Great. I know this had been talked about a lot but wasn't sure whether itw= had actually been implemented on any of the VMS TCPIP stacks.s  
 David Webb VMS and Unix team leader CCSS Middlesex University   ------------------------------  $ Date: Fri, 8 Dec 2000 15:30:04 +01001 From: "Hugh D. R. Evans" <hevans@wm.estec.esa.nl>c% Subject: Decnet - Proxies not workinga* Message-ID: <90qr9t$k7j$1@news1.xs4all.nl>  
 Hello all,  L We've configured several DECNET proxies on our cluster with a cluster alias.F They were working fine up until the system disk in the ancient VAX hasI become flakey and the machine regularly fails to a boot prompt. As we are F not likely to repair the system disk (RA90), the hardware fix is not a	 solution.=  J The problem is that the cluster alias fail to be used when this old VAX isI in a down state (all machines should use it for outgoing requests, and doSD when this machine is running). The rest of the cluster appears to beJ functioning normally. Is there an NCP command that allows me to remove theH dependency of the cluster alias from this node and assign it to another?   Thanks,    Hugh   ------------------------------  $ Date: Fri, 8 Dec 2000 15:50:48 +01001 From: "Hugh D. R. Evans" <hevans@wm.estec.esa.nl>r) Subject: Re: Decnet - Proxies not working8* Message-ID: <90qsgo$oh1$1@news1.xs4all.nl>  + Sorry, forgot some of the relevant details:w   OS: OpenVms 7.1  DECNET Ph IV (NASAXP v5.0)  C The "MC NCP set exec alias node <CLUSTER_ALIAS>" has been executed.aK FAL has ALIAS incoming/outgoing ENABLED and correct OBJECT userid/password.   J It is a mixed Alpha/VAX cluster, the Old VAX used to be the boot node, but all machines no longer> use it - it is the only machine to boot from it's system disk.  K As previously stated, the problem is the cluster nodes only use the cluster+* alias when this old vax is up and running.   Regards, Hugh      < "Hugh D. R. Evans" <hevans@wm.estec.esa.nl> wrote in message$ news:90qr9t$k7j$1@news1.xs4all.nl... > Hello all, > G > We've configured several DECNET proxies on our cluster with a cluster  alias.H > They were working fine up until the system disk in the ancient VAX hasK > become flakey and the machine regularly fails to a boot prompt. As we areiH > not likely to repair the system disk (RA90), the hardware fix is not a > solution.e >lL > The problem is that the cluster alias fail to be used when this old VAX isK > in a down state (all machines should use it for outgoing requests, and dooF > when this machine is running). The rest of the cluster appears to beL > functioning normally. Is there an NCP command that allows me to remove theJ > dependency of the cluster alias from this node and assign it to another? >l	 > Thanks,u >b > Hugh >e >    ------------------------------  $ Date: Fri, 8 Dec 2000 12:08:30 -0500( From: "David M Smith" <dsmit115@csc.com>) Subject: Re: Decnet - Proxies not workinga' Message-ID: <90r4ik$31g$1@lore.csc.com>w  K In order for a cluster alias to work (that is, for a node in the cluster toLK respond to an incoming connect request which specifies the alias node name) L at least one of the running nodes must be configured as a full routing node.L I suspect, from your description, that the VAX was setup for routing and the' remaining nodes are setup as end nodes.v  I You can check this in NCP by issuing a SHOW EXEC CHAR command and looking L for the TYPE = value. If this says nonrouting IV for all nodes, then that is
 your problem.d  I If that is the problem, then you must change one or more of the remainingeE nodes to be routing nodes. This is done with NCP DEFINE EXECUTOR TYPEmL ROUTING IV, after which DECnet must be shutdown and restarted, or the systemI rebooted (it cannot be changed on the fly with DECnet up and running). In L order for this to work, you must be licensed for routing, with a PAK such as DVNETRTG or whatever.y  < "Hugh D. R. Evans" <hevans@wm.estec.esa.nl> wrote in message$ news:90qr9t$k7j$1@news1.xs4all.nl... > Hello all,L > The problem is that the cluster alias fail to be used when this old VAX isK > in a down state (all machines should use it for outgoing requests, and dokF > when this machine is running). The rest of the cluster appears to beL > functioning normally. Is there an NCP command that allows me to remove theJ > dependency of the cluster alias from this node and assign it to another?   ------------------------------   Date: 8 Dec 2000 18:06:11 GMT 2 From: hoffman@xdelta.zko.dec.nospam (Hoff Hoffman)) Subject: Re: Decnet - Proxies not working=6 Message-ID: <90r7uj$q4u$1@mailint03.im.hou.compaq.com>  ^ In article <90qr9t$k7j$1@news1.xs4all.nl>, "Hugh D. R. Evans" <hevans@wm.estec.esa.nl> writes: ..K :The problem is that the cluster alias fail to be used when this old VAX is	J :in a down state (all machines should use it for outgoing requests, and doE :when this machine is running). The rest of the cluster appears to be:K :functioning normally. Is there an NCP command that allows me to remove the0I :dependency of the cluster alias from this node and assign it to another?p  D   Follow the steps for setting up the cluster alias and the proxies?  F   I'd guess that this OpenVMS VAX node was the DECnet router or DECnetB   extended-function node, you'll need to add another router or an E   extended-function license to the cluster, or retarget the existing iF   routing license to another node and configure that node as a router.  N  --------------------------- pure personal opinion ---------------------------L    Hoff (Stephen) Hoffman   OpenVMS Engineering   hoffman#xdelta.zko.dec.com   ------------------------------  % Date: Fri, 08 Dec 2000 11:32:54 +0200o From: mauf@uhu.unizar.es" Subject: Email alias on DEC TCP/IP* Message-ID: <00120811325414@uhu.unizar.es>  > Set an account for ALIAS@HOST.IT, and then, from this account,   MAIL> SET forward ...e  
 Works for me!a   Regardss   Miguel Angel   ------------------------------  # Date: Fri, 08 Dec 2000 18:37:01 GMTa* From: Doran Werling <rwscsinc@my-deja.com>1 Subject: F$GETJPI and VMS 7.1, 7.2 BYTLM = BYTCNTs) Message-ID: <90r9od$trp$1@nnrp1.deja.com>h  ; Has anyone noticed that F$GETJPI(pid,"BYTLM") always equalsdE F$GETJPI(pid,"BYTCNT) under VMS 7.1 and 7.2 ? I was troubleshooting an@ process quota problem involving bytlm and observed these 2 value decrement together.   
 Doran WerlingR RW/SCS Inc.i    & Sent via Deja.com http://www.deja.com/ Before you buy.g   ------------------------------  % Date: Fri, 08 Dec 2000 09:43:23 -0600r From: canellis@al.brooks.af.milE* Subject: Fujitsu M2488E tape drive trouble- Message-ID: <00120809432369@al.brooks.af.mil>R   All,  C 	I am attempting to attache a Fujitsu M2488E tape drive on an Alpha I running OpenVms 6.2.  When I boot the system I see the tape drive.  As I NM monitor the drive, it sees if I put a tape in the drive.  The trouble starts 5N when I attempt to mount the drive.  I get the "Volume is not software enabled"N error.  When I run SYS$ETC:SCSI_INFO against the drive, it errors at the first3 QIOW, with a timeout error.  Any ideas?  Thank you,5   Alan J. Canellis   ------------------------------  % Date: Fri, 08 Dec 2000 08:03:03 +0100tA From: Eberhard Heuser-Hofmann <vaxinf@dg3.chemie.uni-konstanz.de> 9 Subject: Here is my Xmas-Present: cdrecord for IDE-drives 0 Message-ID: <00c601c060e4$e9c53780$1d072286@ws1>   Hi,c   >iA > You all know that SCSI-CDR/CD-WR-drives are more expensive than  IDE-drives.nF > My Alphastation 500au has a 32x-CDROM-IDE-drive and uses the DQDRIV= ER.r > F > I have decided to spend some money and bought a very recent IDE-CDR= W-drive F > with so called "burnproof"-option build in ("Forget the buffer unde= rrun
 problem"), > a TEAC CD-W512EB.N >WC > After one week of working hard everything now looks great! I haveT installed the newestF > dqdriver.c after a small modification (The timeout-parameter is 15 = sec in the original). >o3 > Get the following files if you are interested in:- >-D > ftp://v36.chemie.uni-konstanz.de/cdrecord_vms/CDRECORD-1_10A07.ZIP( >  and DQDRIVER.C in the same directory. >9% > For further instructions read this:I >C > C =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=-C =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=9C =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=- =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D =3D=3D=3D=3D=3D=3D=3D=3D8 > This README was written by Dr. Eberhard Heuser-Hofmann% > vaxinf@chclu.chemie.uni-konstanz.de. >iA > If you are looking for a *.zip file instead of the tar archive,9 > check this address:  >e0 > ftp://v36.chemie.uni-konstanz.de/cdrecord_vms/ >aF > You'll find several versions of cdrecord. I recommend to use the la= test> > version. Please keep in mind that i.e. CDRECORD_1_8A40 meansF > Version 1.8 Alpha 40 whereas CDRECORD_1_8_1 is the final release 1.= 8.1. >w >h >eF /*-------------------------------------------------------------------= -------i */< > cdrecord is been tested by using the following CDR/CD-RWs: >eF > HP6020, Philips CDD200, Yahaha CRW 4260tx, TEAC 58S and a TEAC CD-W= 512EB. >iF > There is a command-procedure called build_all.com in the [.cdrecord= ]q subdir, F > that compiles and links the code. The cdrecord-program is been prod= uced inn > [.cdrecord.vms] subdirectory.r >)% > Define this program as a symbol via  > * > $ cdrecord:=3D=3D$disk:[dir]cdrecord.exe >rF > $ cdrecord -help gives an idea about the options, that are possible= :h >e > cdrecord -help6 > $75$dka0:[sys_tools.cdrecord]cdrecord.exe;31: Usage:F > $75$dka0:[sys_tools.cdrecord]cdrecord.exe;31 [options] track1...tra= cknr
 > Options:< >         -version        print version information and exit@ >         -v              increment general verbose level by oneF >         -V              increment SCSI command transport verbose le= vel by one 9 >         -debug          print additional debug messages A >         dev=3Dtarget      SCSI target to use as CD/DVD-Recorder F >         timeout=3D#       set the default SCSI command timeout to #= .NF >         driver=3Dname     user supplied driver name, use with extre= me care F >         driveropts=3Dopt  a comma separated list of driver specific=  optionsD >         -checkdrive     check if a driver for the drive is presentF >         -prcap          print drive capabilities for MMC compliant = drives> >         -inq            do an inquiry for the drive and exit4 >         -scanbus        scan the SCSI bus and exitD >         -reset          reset the SCSI bus with the cdrecorder (if	 possible)hF >         -ignsize        ignore the known size of a medium (may caus= ea	 problems)eE >         -useinfo        use *.inf files to overwrite audio options.a. >         speed=3D#         set speed of driveA >         blank=3Dtype      blank a CD-RW disc (see blank=3Dhelp)pF >         -load           load the disk and exit (works only with tra= yt loader)o= >         -eject          eject the disk after doing the worka= >         -dummy          do everything with laser turned off D >         -msinfo         retrieve multi-session info for mkisofs >= =3D 1.109 >         -toc            retrieve and print TOC/PMA datae6 >         -atip           retrieve and print ATIP dataB >         -multi          generate a TOC that allows multi sessionF >                         In this case default track type is CD-ROM X= A2F >         -fix            fixate a corrupt or unfixated disk (generat= e a= TOC)A >         -nofix          do not fixate disk after writing tracks F >         -waiti          wait until input is available before openin= g SCSIF >         -force          force to continue on some errors to allow b= lankingk	 bad disks=E >         -dao            Write disk in DAO mode. This option will beo replaced in the-	 > future.-> >         tsize=3D#         Length of valid data in next track< >         padsize=3D#       Amount of padding for next trackF >         pregap=3D#        Amount of pre-gap sectors before next tra= ckF >         defpregap=3D#     Amount of pre-gap sectors for all but tra= ck #1lF >         mcn=3Dtext        Set the media catalog number for this CD =	 to 'text'aF >         isrc=3Dtext       Set the ISRC number for the next track to=  'text'eF >         index=3Dlist      Set the index list for the next track to = 'list'B >         -audio          Subsequent tracks are CD-DA audio tracksF >         -data           Subsequent tracks are CD-ROM data mode 1 (d= efault)tB >         -mode2          Subsequent tracks are CD-ROM data mode 2@ >         -xa1            Subsequent tracks are CD-ROM XA mode 1@ >         -xa2            Subsequent tracks are CD-ROM XA mode 2: >         -cdi            Subsequent tracks are CDI tracksF >         -isosize        Use iso9660 file system size for next data = trackqC >         -preemp         Audio tracks are mastered with 50/15 =B5sN preemphasisnF >         -nopreemp       Audio tracks are mastered with no preemphas= is	 (default)l@ >         -pad            Pad data tracks with 15 zeroed sectorsF >                         Pad audio tracks to a multiple of 2352 byte= s,: >         -nopad          Do not pad data tracks (default)F >         -shorttrack     Subsequent tracks may be non Red Book < 4 s= econds if in DAO modeB >         -noshorttrack   Subsequent tracks must be >=3D 4 seconds; >         -swab           Audio data source is byte-swappedh (little-endian/Intel)e7 > The type of the first track is used for the toc type.t- > Currently only form 1 tracks are supported.e > $ cdrecord -versionoF > Cdrecord 1.10a07 (Alpha/VAX-CPQ-VMS/OpenVMS) Copyright (C) 1995-200= 0 J=F6rg	 Schillingi >g> > One important point is how the CDR-device is being selected: >s
 > dev=3Di,j,km >  > where:3 > i: 0   1   2   3   4   5   6   7   8   9   10  11n4 >    dka dkb dkc dkd gka gkb gkc gkd dqa dqb dqc dqd >  > j: scsi-id > k: sub-lun (usually 0) >c > As an example: >y > dev=3D1,5,2 selects dkb502:  >nF > An typical example how to put virtual disk data onto a cd-rom is gi= ven' here:  > : > $ cdrecord speed=3D2 dev=3D0,4,0 dka200:[000000]vda0.dsk >yF > If you use an IDE-CD-Recorder that has the burnproof feature build = in, youo > should activate this:l > ; > $ cdrecord speed=3D12 -driveropts=3Dburnproof dev=3D9,0,0e dka200:[000000]vda0.dskr >a8 > Additionally I recommend to change the rms-parameters: >y& > "$ set rms/block=3D127/buff=3D255" , > B > in order to prevent buffer underun for non burnproof CDR/CD-RWs. > F > The original IDE-driver-source dqdriver.c "X-21", that Compaq distr= ibutes,- > needs to be modified:e >9# > #define TIMEOUT_TIME        16000R >c > instead of >p  > #define TIMEOUT_TIME        15 >o- > Be aware that cdrecord has copyright on it.e >g
 > Eberhard >uF > PS: Upto now I have earned no money for that work. If I read postin= gs thatL showA > that other people use "my" program and sell it I'm a little bitd
 disappointed.a   ------------------------------  % Date: Fri, 08 Dec 2000 09:28:36 -0500 1 From: "Farrell, Michael" <MFarrell@voltdelta.com>n4 Subject: RE: Heresy (was Re: how to sort a CSV file)C Message-ID: <7DF45F22D904D31192EE00805F578DF2018BC5ED@NY_EXCHANGE1>a  G Just curious to see if you got this message.  I had replied as it showsbL below, but never saw it appear in "Info-VAX".  So I wondered if it ever made it out of here.m   Mike   > -----Original Message----- > From:	Farrell, Michael  , > Sent:	Wednesday, December 06, 2000 8:12 AM, > To:	'Chris Sharman'; Info-VAX@Mvb.Saic.Com6 > Subject:	RE: Heresy (was Re: how to sort a CSV file) >  > Chris, > H > If the file being loaded into Excel is considered a text file (*.txt),J > generally the "Wizard" will run and start asking questions about how theI > data is delimited etc..  In step 3, there is a dialog box that lets youhL > specify the data format by column.  You can select the relevant columns ofL > data and click on the appropriate radio button to have the data treated as > text, dates, numerics, etc.B > K > This is true in Excel '97 and later versions.  I don't know about earliere > ones.w >  > Hope this helps. >  > Mike Farrell >  > 	-----Original Message-----u: > 	From:	Chris Sharman [SMTP:Chris.Sharman@CCAgroup.co.uk]- > 	Sent:	Wednesday, December 06, 2000 4:59 AMt > 	To:	Info-VAX@Mvb.Saic.Com3 > 	Subject:	Heresy (was Re: how to sort a CSV file)c >  > 	> [Begin heresy mode]E > 	> If you have Excel and a convenient way to transfer text files to- > and B > 	> from it, it's not hard to pull in a CSV file, sort it with or	 > without.B > 	> headings, then write it back out in CSV format. Caveat: Excel > accepts a.; > 	> limited number of lines (64k in one or more versions).r > 	> [End heresy mode] > F > 	Anyone know how to exert any control over Excel's processing of CSV
 > 	files ? > 	In particular:nC > 	1) how to hand it text fields containing numbers: eg credit cardyE > 	numbers, product codes starting with a digit, etc. I find it has a= > niceE > 	security feature when fed credit card numbers: it renders them for E > 	example as 4.55E+15, even when they're double-quoted. Save them asd > CSVn" > 	again & they're really trashed.F > 	2) Similar problems with anything it can make a date from: 1/2 etc.@ > 	3) Any other useful tricks with CSV files, pointers to useful > websites,e > 	experts, etc. > 
 > 	Thanks, > 	Chris Sharman   ------------------------------  * Date: Fri, 8 Dec 2000 09:27:27 +0000 (UTC)' From: Osmo Kujala <kujala@tukki.jyu.fi> ( Subject: Re: New OpenVMS  Education site, Message-ID: <90q9hv$g53$1@mordred.cc.jyu.fi>  3 David Mathog <mathog@seqaxp.bio.caltech.edu> wrote:p   > A different section says:P  4 >   "However, the Software may be copied and used onC >    multiple computers that are under your control in a classroom oM >    environment, including use by multiple students who have agreed to theseuD                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^6 >    License Terms solely for instructional purposes."  M > Oh right.  I defy the people at Compaq to identify a single college or K-12nH > school ANYWHERE on the face of the Earth that has a room with multipleD > single user VMS workstations set up for "instructional purposes".   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^  F IMHO text "including use by multiple students ..." clearly states thatE that multiple users are allowed on multiple computers and there is notE restriction to single user computers in the classroom environment andmD those students need not be the individuals who ordered the PAKs. TheH License Grant text is meaningful here. The PAKs of course allow for thatC (and maybe more). I see it that the person who ordered Licences foriF classroom is responsible for the use and agreement of License Terms by	 students.T   I may be wrong of course.s&                            Osmo Kujala   ------------------------------   Date: 8 Dec 2000 16:24:25 GMTl2 From: mathog@seqaxp.bio.caltech.edu (David Mathog)( Subject: Re: New OpenVMS  Education site, Message-ID: <90r1vp$m3k@gap.cco.caltech.edu>  V In article <90q9hv$g53$1@mordred.cc.jyu.fi>, Osmo Kujala <kujala@tukki.jyu.fi> writes:4 >David Mathog <mathog@seqaxp.bio.caltech.edu> wrote: >m >> A different section says: >s5 >>   "However, the Software may be copied and used onnD >>    multiple computers that are under your control in a classroom N >>    environment, including use by multiple students who have agreed to theseE >                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^d7 >>    License Terms solely for instructional purposes."  >cN >> Oh right.  I defy the people at Compaq to identify a single college or K-12I >> school ANYWHERE on the face of the Earth that has a room with multipledE >> single user VMS workstations set up for "instructional purposes". a > ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^. >iG >IMHO text "including use by multiple students ..." clearly states thatmF >that multiple users are allowed on multiple computers and there is noB >restriction to single user computers in the classroom environment  K Yes there is such a restriction.  The PAKs themselves only deliver a singlerH user license and you cannot load multiple licenses one one machine.  See4 the FAQ at the web site which says this explicitly.    >I may be wrong of course.   Sorry, you are.h   David Mathog mathog@seqaxp.bio.caltech.edui? Manager, sequence analysis facility, biology division, Caltech h   ------------------------------   Date: 8 Dec 2000 08:40:27 -0700 1 From: nothome@spammers.are.scum (Malcolm Dunnett) ( Subject: Re: New OpenVMS  Education site, Message-ID: <+bOk7IV021je@malvm1.mala.bc.ca>  - In article <90q9hv$g53$1@mordred.cc.jyu.fi>, 0-     Osmo Kujala <kujala@tukki.jyu.fi> writes:n  I >> school ANYWHERE on the face of the Earth that has a room with multiple E >> single user VMS workstations set up for "instructional purposes". n  >  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ > H > IMHO text "including use by multiple students ..." clearly states thatG > that multiple users are allowed on multiple computers and there is nooG > restriction to single user computers in the classroom environment and C > those students need not be the individuals who ordered the PAKs. r  C    If you go to the Compaq website and look at the FAQ posted theretI you'll see they explicitly state you MAY NOT combine PAKS under this planrJ to permit multiple users on a single system. They're pretty clear that anyH license granted under this program ( and consequently any computer whichG loads that license ) can only be used by a single user at a time. GiveneC that context it seems clear that their intent is that an instructordH may order one set of licenses and install them on a bunch of single userD workstations - but those workstations may only be used by one personL at a time and only when under his direct supervision in a classroom setting.  G    This is itself not a bad thing ( other than the "direct supervision".D clause rendering it worthless ). If the Tru64 enthusiast program hadH permitted this usage a year ago we probably would have purchased a bunchF of PWS systems rather than a bunch of X terms. However it does nothingG for those institutions which want to offer VMS in a multi-user setting.    ------------------------------  $ Date: Fri, 8 Dec 2000 12:53:41 -0500) From: "Ebinger . Eric" <EEbinger@drc.com>r( Subject: RE: New OpenVMS  Education siteB Message-ID: <7162F87E9EF4D311BA9900805FC1D3AE7A625B@and02.drc.com>   > -----Original Message-----% > From: mathog@seqaxp.bio.caltech.edub> > >IMHO text "including use by multiple students ..." clearly 
 > states thateH > >that multiple users are allowed on multiple computers and there is noD > >restriction to single user computers in the classroom environment > = > Yes there is such a restriction.  The PAKs themselves only M > deliver a single= > user license and you cannot load multiple licenses one one   > machine.  Seet6 > the FAQ at the web site which says this explicitly.  >   6 Of course, the FAQ says that "Base OpenVMS PAKs cannot7 be combined".  Isn't it the OpenVMS-Alpha-User PAK thath controls the number of users?        Eric Ebinger   ------------------------------  % Date: Fri, 08 Dec 2000 10:23:44 +0000r% From: Alan Greig <agreig@my-deja.com> % Subject: Re: One world, one processorm8 Message-ID: <01d13t0m7m82l6ootdgb0nol1vak9t9dj0@4ax.com>  E On 08 Dec 2000 02:06:58 GMT, dashw459@aol.comeatspam (Doug W.) wrote:t  J >The following quote was taken from an article ONE WORLD, ONE PROCESSOR by. >Linley Gwennap in the December LINUX JOURNAL. > K >..."Compaq, HP and SGI have slowed the pace of their architectures (Alpha,wO >PA-RISC and MIPS, respectively) to the point that they have fallen behind XeontM >in performance in many key applications.  This decline has caused a downwardc& >spiral in the sales for RISC vendors. >rF >  As a result, HP and SGI have already announced they will eventuallyD >discontinue their RISC lines, and Compaq is likely to follow suit."  C And if Compaq don't jump on statements like this then quite bluntlyn; they are probably true - whatever Compaq employees are toldTC internally. We keep hearing about all the Alpha development work ino? the pipeline but the hard facts are that Alpha should really beu? shipping at 1.6Ghz EV68 by now. It is difficullt to see all the.B promised enhancements materialising when the pace has been so slowC over the last two years. There hasn't really been any *significant*DF upggrades since the 500Mhz EV6 over two years ago. I do not consider a? less than 2:1 performance hike in over two years good progress.=  D If EV8 is not to be the last Alpha chip then Compaq had better start taking its gloves off.  E And VMS engineering had better keep their eyes on IA-64 just in case.i   --
 Alan Greig   ------------------------------  % Date: Fri, 08 Dec 2000 09:49:24 -0300 ) From: fabio_compaq@ep-bc.petrobras.com.brc% Subject: Re: One world, one processorsL Message-ID: <OF93FD230C.0C99E54E-ON032569AF.00457D51@ep-bc.petrobras.com.br>  C I believe in the creation of the new wave: PROCESSOR CONSOLIDATION.R) Stop with this myriad of processors . . .   H To the end-customer it=B4s important to stay with Alpha architecture. L= ast= year we changed oureH Alphaservers 4100 running Windows NT + Lotus Notes to Proliant 8500 + N= T.H What mess... transfer 7500 users from one 2 machine cluster to another.=  * Just becase Compaq decided stop WNT/Alpha.  H Imagine porting all the Alpha applications to IA-64 ? Wil be the end of=   OpenVMS, because4 a lot of customers will decide to port to Unix......  H The market should have the max of 4, 5 processor lines.....  Alpha, Spa= rc,o Intel, MIPS, Power PC.   It=B4s enough....a   Regardst   FC                6 Alan Greig <agreig@my-deja.com> em 08/12/2000 08:23:44H                                                                        =     =20nH                                                                        =     =20tH                                                                        =     =20     @                                                              =20@                                                              =20@                                                              =20@  Para:    Info-VAX@Mvb.Saic.Com                              =20@                                                              =20@  cc:      (bcc: Fabio dos Santos Cardoso/E-P-BC/Contratada)  =20@                                                              =20@                                                              =20@                                                              =20@  Assunto: Re: One world, one processor                       =20@                                                              =20           =     E On 08 Dec 2000 02:06:58 GMT, dashw459@aol.comeatspam (Doug W.) wrote:   H >The following quote was taken from an article ONE WORLD, ONE PROCESSOR=  bye. >Linley Gwennap in the December LINUX JOURNAL. >oH >..."Compaq, HP and SGI have slowed the pace of their architectures (Al= pha,H >PA-RISC and MIPS, respectively) to the point that they have fallen beh= inds XeonD >in performance in many key applications.  This decline has caused a downward& >spiral in the sales for RISC vendors. > F >  As a result, HP and SGI have already announced they will eventuallyD >discontinue their RISC lines, and Compaq is likely to follow suit."  C And if Compaq don't jump on statements like this then quite bluntlye; they are probably true - whatever Compaq employees are toldeC internally. We keep hearing about all the Alpha development work ina? the pipeline but the hard facts are that Alpha should really be ? shipping at 1.6Ghz EV68 by now. It is difficullt to see all the B promised enhancements materialising when the pace has been so slowC over the last two years. There hasn't really been any *significant*IF upggrades since the 500Mhz EV6 over two years ago. I do not consider a? less than 2:1 performance hike in over two years good progress.A  D If EV8 is not to be the last Alpha chip then Compaq had better start taking its gloves off.  E And VMS engineering had better keep their eyes on IA-64 just in case.    --
 Alan Greig     =    ------------------------------  % Date: Fri, 08 Dec 2000 13:41:51 +0100   From: Paul Sture <paul@sture.ch>% Subject: Re: One world, one processor + Message-ID: <VA.000001c0.047aabc3@sture.ch>   I In article <20001207210658.08136.00002406@ng-mi1.aol.com>, Doug W. wrote:m) > From: dashw459@aol.comeatspam (Doug W.)  > Newsgroups: comp.os.vmsh  > Date: 08 Dec 2000 02:06:58 GMT# > Subject: One world, one processor  > K > The following quote was taken from an article ONE WORLD, ONE PROCESSOR bya/ > Linley Gwennap in the December LINUX JOURNAL.  >mN Bah! Two or three years since I heard his name, and then he was writing about F IA64 taking the world by storm. Well, it hasn't done that yet, has it?  M > ...."Compaq, HP and SGI have slowed the pace of their architectures (Alpha,rP > PA-RISC and MIPS, respectively) to the point that they have fallen behind XeonN > in performance in many key applications.  This decline has caused a downward' > spiral in the sales for RISC vendors.  >iG >   As a result, HP and SGI have already announced they will eventually E > discontinue their RISC lines, and Compaq is likely to follow suit."  >   Q From what I remember of his comments about IA64, I wouldn't have expected him to d say anything different.    > N > I hope someone from Compaq replies to this article.  I think it would have a; > good chance of being published in a future LINUX JOURNAL.= >4   Excellent idea!  ___ 
 Paul Sture Switzerlandc   ------------------------------  % Date: Fri, 08 Dec 2000 13:14:12 +0000s  From: steven.reece@quintiles.com% Subject: Re: One world, one processor H Message-ID: <OFB63FD77E.17024B28-ON802569AF.00484245@qedi.quintiles.com>  J So why not make it two - Alpha and Intel?  But which Intel - IA32 or IA64?  H But then why not make it just one, Alpha.  It's the fastest and the best< and still has more than 50% of its 25 year life ahead of it.  F Bear in mind that most users and system managers are keen to keep whatG they're running, even if it does have holes in it in our opinion.  PlusnG each processor may be suited to a particular job or type of processing.   1 Standards are great.  Lets have hundreds of them.:      + fabio_compaq@ep-bc.petrobras.com.br wrote :e  F >>>The market should have the max of 4, 5 processor lines.....  Alpha, Sparc, Intel, MIPS, Power PC.<<<i   ------------------------------   Date: 8 Dec 2000 13:25:28 GMTd' From: david20@alpha2.mdx.ac.uk (D.Webb)p% Subject: Re: One world, one processor70 Message-ID: <90qng8$ito$1@aquila.news.mdx.ac.uk>  x In article <OF93FD230C.0C99E54E-ON032569AF.00457D51@ep-bc.petrobras.com.br>, fabio_compaq@ep-bc.petrobras.com.br writes:D >I believe in the creation of the new wave: PROCESSOR CONSOLIDATION.* >Stop with this myriad of processors . . . >   ) Rubbish. That is the last thing we need. gJ Consolidation closes down the exploration of alternative paths to improvedE performance. If you have only one type of chip you are constrained by"? the historic design decisions which went into making that chip.-H Why is Intel making the IA-64 chip ? Answer because other companies had F moved to 64 bit. If there were no other chips Intel would have little N incentive to do anything but keep on marginally improving their current chips.K There would be no reason to invest lots of money in any radical new design.   N (True at some point they would have to - but only because some startup companyI had produced or was about to produce a new chip design which offered the 	N improved performance customers wanted. And we might be waiting a long time for> that to happen because the startup costs would be horrendous).  I Now it might be argued that in a few years time we would have reached theeE limits of what can be done with silicon and all the tricks to improve K performance will have been tried. At that point it could be argued that the3K final silcon chip design will be created and everything will consolidate on K that. However research is already underway into optical computing and otheraK designs which means that I'd doubt that such a consolidation would actually M happen. Instead you might well see a number of radically different approachesnL competing in the market place which would make the architectural differences4 between IA-64, Alpha and SPARC chips seem miniscule.F And what drives this research. Answer : The perceived need to increase performance + competition.    
 David Webb VMS and Unix team leader CCSS Middlesex University  I >To the end-customer it=B4s important to stay with Alpha architecture. L=  >ast >year we changed ourI >Alphaservers 4100 running Windows NT + Lotus Notes to Proliant 8500 + N=t >T.rI >What mess... transfer 7500 users from one 2 machine cluster to another.=m > + >Just becase Compaq decided stop WNT/Alpha.> >nI >Imagine porting all the Alpha applications to IA-64 ? Wil be the end of=o >s >OpenVMS, becauses5 >a lot of customers will decide to port to Unix......s >nI >The market should have the max of 4, 5 processor lines.....  Alpha, Spa=  >rc, >Intel, MIPS, Power PC.  >e >It=B4s enough.... >h >Regards >  >FCn >e >  >r >e >t >  >  > 7 >Alan Greig <agreig@my-deja.com> em 08/12/2000 08:23:44vI >                                                                       =e >    =20I >                                                                       =h >    =20I >                                                                       =r >    =20 >  >-A >                                                             =20(A >                                                             =20eA >                                                             =20 A > Para:    Info-VAX@Mvb.Saic.Com                              =20rA >                                                             =20uA > cc:      (bcc: Fabio dos Santos Cardoso/E-P-BC/Contratada)  =20aA >                                                             =20hA >                                                             =20^A >                                                             =20 A > Assunto: Re: One world, one processor                       =20aA >                                                             =20h >a >a >  >m >t >= >p > F >On 08 Dec 2000 02:06:58 GMT, dashw459@aol.comeatspam (Doug W.) wrote: >^I >>The following quote was taken from an article ONE WORLD, ONE PROCESSOR=. > by/ >>Linley Gwennap in the December LINUX JOURNAL.  >>I >>..."Compaq, HP and SGI have slowed the pace of their architectures (Al=  >pha,sI >>PA-RISC and MIPS, respectively) to the point that they have fallen beh=r >ind >XeonTE >>in performance in many key applications.  This decline has caused a 	 >downwardm' >>spiral in the sales for RISC vendors.o >>G >>  As a result, HP and SGI have already announced they will eventuallyiE >>discontinue their RISC lines, and Compaq is likely to follow suit."  > D >And if Compaq don't jump on statements like this then quite bluntly< >they are probably true - whatever Compaq employees are toldD >internally. We keep hearing about all the Alpha development work in@ >the pipeline but the hard facts are that Alpha should really be@ >shipping at 1.6Ghz EV68 by now. It is difficullt to see all theC >promised enhancements materialising when the pace has been so slowyD >over the last two years. There hasn't really been any *significant*G >upggrades since the 500Mhz EV6 over two years ago. I do not consider ai@ >less than 2:1 performance hike in over two years good progress. > E >If EV8 is not to be the last Alpha chip then Compaq had better start  >taking its gloves off.y >rF >And VMS engineering had better keep their eyes on IA-64 just in case. >o >--t >Alan Greigc >e >o >= >  >s   ------------------------------  % Date: Fri, 08 Dec 2000 12:50:45 -0500s  From: norm.raphael@jamesbury.com Subject: Re: OpenVMS Items4 Message-ID: <C22569AF.00614C43.00@jklh21.valmet.com>  M I just tried the link to the eco mailing lists from the OpenVMS FAQ and it ise
       broken.o  M I got there from the openvms.compaq.com eco page and find that the ".html" ine	       thee  & FAQ should be ".shtml" for it to work.   ------------------------------  # Date: Fri, 08 Dec 2000 13:08:50 GMTn8 From: "Alwin Dieperink" <alwin.dieperink@elca-matrix.ch> Subject: process creationt& Message-ID: <C35Y5.120$_V6.46373@news>   Hi all,o  B We have a lot of process creations (up to 2500 creprc/hour) on ourL overloaded system (alpha 8400-5/625, VMS 7.1-H1, 8 CPUs, 8GB). What negativeF influence has this on our system ? Would we have better performance by$ reducing it ? How much will we win ?   Thanks for any advicet -- Alwin   ------------------------------   Date: 8 Dec 2000 16:21:31 GMTg2 From: hoffman@xdelta.zko.dec.nospam (Hoff Hoffman) Subject: Re: process creationh6 Message-ID: <90r1qb$pcm$1@mailint03.im.hou.compaq.com>  a In article <C35Y5.120$_V6.46373@news>, "Alwin Dieperink" <alwin.dieperink@elca-matrix.ch> writes:^C :We have a lot of process creations (up to 2500 creprc/hour) on our M :overloaded system (alpha 8400-5/625, VMS 7.1-H1, 8 CPUs, 8GB). What negativerG :influence has this on our system ? Would we have better performance byo% :reducing it ? How much will we win ?e  E   You are creating somewhat less than one process per second, peak.  o  M   For an AlphaServer 8400 system, this process creation rate does not appear yM   to be particularly extreme.  Depending on how these creations are performedsK   (and on other system activity), this can certainly trigger either I/O or dN   file-level contention issues.  Approaches such as faster I/O and RMS global N   buffers and such can be useful here.  If these are process spawn operations,K   then you may want to disable the logical name and symbol copy operations fI   (or reduce the size of the context getting copied into the subprocess.) H   Process creation can also be reflected by increased I/O and inner-mode   system activity.  K   I will assume the release in use is V7.1-1H1, and I would thus encourage eI   upgrading to OpenVMS Alpha V7.1-2 or V7.2-1, with current ECOs applied.a  H   I would encourage loading AMDS or Availability Manager for monitoring.  K   I would encourage identifying the underlying performance-limiting factor.h  J   I would encourage identifying the specific image(s) involved.  LOGINOUT?  J   I would encourage identifying improvements (caching, image installation,H   increased memory, correct working set sizing, I/O improvements, use of4   fast path, etc) based on current system behaviour.  I   I would encourage establishing a performance baseline, to identify and nE   to evaluate the relative success of various tuning-related changes.   I   Please follow the instructions in the performance management manual foriG   identifying the source of the current performance bottleneck.  What I I   have posited above is based solely on your process creation rate and onsF   some general performance-related topics, and an assumption that the F   reported process creation rate is the performance bottleneck -- and I   making any assumptions during performance tuning is precarious at best.   N  --------------------------- pure personal opinion ---------------------------L    Hoff (Stephen) Hoffman   OpenVMS Engineering   hoffman#xdelta.zko.dec.com   ------------------------------  # Date: Fri, 08 Dec 2000 12:27:53 GMTc2 From: usenetfilter_bjblueplumb@hotmail.com (Bryce)3 Subject: Recover data from file not closed properly@1 Message-ID: <3a30d2ca.268434554@news.btc-bci.com>q  E We have a utility that creates a file and logs data to the file whileaA running.  The utility runs as a detached process.  We stopped thesF process with $ stop/id=111 which apparently caused the log file to notE be closed properly.  When we do a $ dir/siz=all we see 0/804 so therec4 should be 804 blocks of data but we can't get to it.  6 Is there a way to fix the file so we can see the data?  2 +-+- remove "usenetfilter_" to send me email -+-+- Brycen   ------------------------------  % Date: Fri, 08 Dec 2000 13:53:45 +0100"  From: Paul Sture <paul@sture.ch>7 Subject: Re: Recover data from file not closed properly + Message-ID: <VA.000001c3.04859328@sture.ch>n  > In article <3a30d2ca.268434554@news.btc-bci.com>, Bryce wrote:4 > From: usenetfilter_bjblueplumb@hotmail.com (Bryce) > Newsgroups: comp.os.vms.5 > Subject: Recover data from file not closed properlym% > Date: Fri, 08 Dec 2000 12:27:53 GMTe > G > We have a utility that creates a file and logs data to the file while C > running.  The utility runs as a detached process.  We stopped thedH > process with $ stop/id=111 which apparently caused the log file to notG > be closed properly.  When we do a $ dir/siz=all we see 0/804 so therer6 > should be 804 blocks of data but we can't get to it. > 8 > Is there a way to fix the file so we can see the data? > &    $ SET FILE logfilename /END_OF_FILE    rI You might see garbage in the last few blocks of the file, but you'll get   most of it.O ___r
 Paul Sture Switzerlandg   ------------------------------  % Date: Fri, 08 Dec 2000 09:09:17 -0500 2 From: norm lastovica <norman.lastovica@oracle.com>7 Subject: Re: Recover data from file not closed properly ) Message-ID: <3A30EB8D.6A4FEB3@oracle.com>    $ HELP SET FILE /END   Bryce wrote: > G > We have a utility that creates a file and logs data to the file whilerC > running.  The utility runs as a detached process.  We stopped thehH > process with $ stop/id=111 which apparently caused the log file to notG > be closed properly.  When we do a $ dir/siz=all we see 0/804 so there 6 > should be 804 blocks of data but we can't get to it. > 8 > Is there a way to fix the file so we can see the data? > 4 > +-+- remove "usenetfilter_" to send me email -+-+- > BryceB   -- g> norman lastovica / oracle rdb engineering / usa / 610.696.4685   ------------------------------  % Date: Fri, 08 Dec 2000 08:07:22 -0300 ) From: fabio_compaq@ep-bc.petrobras.com.br  Subject: Re: Sun ClusterL Message-ID: <OFD456E8FB.C38029F7-ON032569AF.003D1319@ep-bc.petrobras.com.br>  
 I AGREE . . .                   A andrew harrison <andrew.nospam@uk.sun.com> em 07/12/2000 16:54:57 L                                                                             L                                                                             L                                                                                 >                                                               >                                                               >                                                               >  Para:    Info-VAX@Mvb.Saic.Com                               >                                                               >  cc:      (bcc: Fabio dos Santos Cardoso/E-P-BC/Contratada)   >                                                               >                                                               >                                                               >  Assunto: Re: Sun Cluster                                     >                                                                             ! steven.reece@quintiles.com wrote:  >-4 > Peter Langstoeger (eplan at kapsch dot net) asked:E > >>>Then why do they (Compaq) still destroy the brand "Digital" ?<<<T > H > Because it's not Digital any more.  It is Compaq.  One big Compaq thatK > offers the whole range of products from PCs through Wintel servers, Linuxs  > boxes, Tru64, NSK and OpenVMS.G > Compaq are not a PC box shifter, they are a company offering the fulll ranges0 > of product for the full range of applications. >e  D The problem with this is that Compaq are a PC box shifter, its theirF culture and for example its one of the reasons why so many people leftF Digital when it was bought by Compaq because the pervasive culture was so alien to them.m  > Compaq also show little or no sign of extending out beyond the
 boundariesF of their PC space. Not suprising when you consider that Compaq is used toE having nearly all the details of their product strategies dictated tor themG by other people (Intel and Microsoft). Even the basic specifications ofp; PC's are dictated to Compaq in Microsofts standard platform  specification.  G Compaq became a sucess by being very very close to Intel and Microsoft,  being C good at manufacturing and distribution but not by being a source of0 radicalm ideas or technical inovations.  D In other words Compaq does not historically have any thought leadersF or people who go and inovate. The Digital purchase gave them access to thisF kind of people resource but that is now long gone, lost in the fallout from
 the purchase.8   Regards2 Andrew Harrisonj Enterprise IT Architects   ------------------------------  $ Date: Fri, 8 Dec 2000 15:10:37 -00001 From: "Michael Shield" <Michael.Shield@tesco.net>X Subject: Re: Sun Cluster, Message-ID: <90qtdv$m7k$1@barcode.tesco.net>   Well,r  E If you want to show just how little you apreciate what goes on in theeH combined company that was compaq/digital/tandem, go ahead agree with old6 bone head. He knows about as much as you do obviously.  J Sorry if this seems a little direct, but I haven't time to type and delete as my daughter needs me.   Mike Shieldn   Usual AH slush removed   ------------------------------  % Date: Fri, 08 Dec 2000 16:54:20 +0000d  From: steven.reece@quintiles.com Subject: Re: Sun ClusterH Message-ID: <OFC9D52AEF.AFEFA7B6-ON802569AF.00573466@qedi.quintiles.com>  8 I would suggest that you're wrong on most counts Andrew.K Yes there have been significant departures both before and after the Compaq E purchase.  However, one of the things that you should realise is thatLH employees get demoralized when they are uncertain of their future.  WhenH the new employer comes along or when their conditions change it can takeI some time for the people to become adjusted to the new environment.  Somee never do and leave the company.8  C Compaq are showing significant signs of not being a PC box shifter,aK although it doesn't help very much when other IT professionals or the pressnF still label them as box shifters.  One could accuse Sun of being a boxJ shifter too, although that would again be untrue.  Both companies are much more.a  H As far as product strategy being dictated to them, this is the case withK most if not all of the companies who sell PCs.  Dell, for example, would beoH unlikely to go  out and develop their own range of processor chips sinceJ it's just too expensive to do.  They (the PC vendors) don't generally sellJ their own operating systems for PCs because Mr. and Mrs. X just want to be7 running Windows because that's what they're exposed to.0  I Compaq are different in that they have products to go from the desktop torJ the high-end datacentre, they can group them together in clusters or leaveJ them standalone, they can support web sites and search engines that handleJ hundreds of thousands of transactions a day and they can offer support andJ professional services to make sure that the complete solution doesn't fallG over because someone forgot to put an error checking mechanism on theirs cache. Steve.   Our beloved Andrew wrote: $ >>>steven.reece@quintiles.com wrote: >n4 > Peter Langstoeger (eplan at kapsch dot net) asked:E > >>>Then why do they (Compaq) still destroy the brand "Digital" ?<<<  > H > Because it's not Digital any more.  It is Compaq.  One big Compaq thatK > offers the whole range of products from PCs through Wintel servers, Linuxk  > boxes, Tru64, NSK and OpenVMS.G > Compaq are not a PC box shifter, they are a company offering the fulla rangee0 > of product for the full range of applications. >-  D The problem with this is that Compaq are a PC box shifter, its theirF culture and for example its one of the reasons why so many people leftF Digital when it was bought by Compaq because the pervasive culture was so alien to them.s    > Compaq also show little or no sign of extending out beyond theJ boundaries of their PC space. Not surprising when you consider that Compaq is used H to having nearly all the details of their product strategies dictated toI them by other people (Intel and Microsoft). Even the basic specificationsh of; PC's are dictated to Compaq in Microsofts standard platformw specification.  G Compaq became a sucess by being very very close to Intel and Microsoft,iI being good at manufacturing and distribution but not by being a source ofc' radical ideas or technical innovations.t  D In other words Compaq does not historically have any thought leadersF or people who go and inovate. The Digital purchase gave them access toK this kind of people resource but that is now long gone, lost in the fallout  from the purchase.<<<x   ------------------------------  % Date: Fri, 08 Dec 2000 13:41:48 +0100o  From: Paul Sture <paul@sture.ch>6 Subject: Re: Sun CSV Constitution Election Belly Laugh+ Message-ID: <VA.000001bf.047aa29c@sture.ch>   E In article <MLXX5.13$Ci5.1314@news1.primary.net>, Jack Peacock wrote:u, > From: "Jack Peacock" <peacock@simconv.com> > Newsgroups: comp.os.vms 8 > Subject: Re: Sun CSV Constitution Election Belly Laugh& > Date: Thu, 7 Dec 2000 18:31:48 -0800 > 8 > "richard n. frank" <rnfrank@llnl.gov> wrote in message: > news:4.2.0.58.20001207080047.00b06db0@poptop.llnl.gov... > > Good morning,,H > > I thought I'd run this up the flagpole and see if anybody salutes... > > J > > Has anyone ever considered restructuring this group along the model of9 > > tru64-unix-managers@ornl.gov?  That is a great group. ) > > People submit questions to the group, 1 > > responses are directed to the submitter only, E > > the submitter is obliged to summarize responses back to the list.  > > H > > There is a lot of good information in this group too, but the volume > of > > mail is staggering.2 > > B > Why fix what isn't broken.  I access comp.os.vms via USENET eachG > morning.  The volume is easily managed.  I like seeing the varied and H > even off-topic questions and answers.  I do not want someone filteringJ > my information.  I prefer to rely on my own abilities to discern what is! > relevant to me and what is not.d > I Although I can see the logic of the original submitter posting a summary  I back to the group, in order to cut down traffic, I think it's far better  J for all of us to see all answers. Not only does it give others the chance J to jump in when they see a better way (and even to correct something that @ is plain wrong!), but the silent lurkers get the benefit of the  discussion.8  C > Even Andrew serves a useful purpose in reminding all of those VMS>G > fanatics that VMS has no privileged right to exist.  I suffer through I > the Solaris-VMS arguments and glean a few gems from the comments.  I doP@ > not want those gems in the rough sifted out for me in advance. > F > If you are getting too much mail, consider shifting to the newsgroupJ > format instead.  In the meantime, no salute from me.  I like it just the > way it is now, no changes.  J There's another alternative to the newsgroup, which I use at work*, which D is to subscribe to the digest. That gives typically 2-3 mails a day K containing all the posts. This not only keeps my inbox tidy, but keeps the e "email police" off my back :-)  K * I prefer to use Netscape to browse the newsgroup by message threads, but  J when our news server is "sick" or unaccessible, I've got the mail digests  as a fall back.a ___h
 Paul Sture Switzerlandp   ------------------------------   Date: 8 Dec 2000 08:22:33 -0500 , From: koehler@eisner.decus.org (Bob Koehler)6 Subject: Re: Sun CSV Constitution Election Belly Laugh+ Message-ID: <HLSkmCQoEkC$@eisner.decus.org>t  l In article <4.2.0.58.20001207080047.00b06db0@poptop.llnl.gov>, "richard n. frank" <rnfrank@llnl.gov> writes: > Good morning, F > I thought I'd run this up the flagpole and see if anybody salutes... > H > Has anyone ever considered restructuring this group along the model of7 > tru64-unix-managers@ornl.gov?  That is a great group. ' > People submit questions to the group, / > responses are directed to the submitter only,mC > the submitter is obliged to summarize responses back to the list.6  B No thanks, I like it the way it is.  I don't trust a VMS newbie toC properly summarize a set of responses to a technical question.  Theo/ volume is much lower than so many other groups.f  F ----------------------------------------------------------------------? Bob Koehler                     | Computer Sciences Corporation1= NASA GSFC Flight Software       | Federal Sector, Civil GrouprE                                 | please remove ".aspm" when replyingw   ------------------------------  % Date: Fri, 08 Dec 2000 14:05:57 +0100r  From: Paul Sture <paul@sture.ch>, Subject: Re: Testers wanted for Spell update+ Message-ID: <VA.000001c4.0490bbf6@sture.ch>u  F In article <874s0gb23v.fsf@k9.prep.synonet.com>, Paul Repacholi wrote: > Newsgroups: comp.os.vmsA* > Subject: Testers wanted for Spell update1 > From: Paul Repacholi <prep@k9.prep.synonet.com>z" > Date: 08 Dec 2000 00:57:56 +0800 >  > 2 > Hi folks, I am looking for a few people to trial3 > an update to Spell before it goes to the Freeware7 > CD. Alpha and Vax, 7.2.< > 6 T'would help if your email address wasn't bouncing :-)  F By which you can deduce that yes, I wouldn't mind doing some testing,  on both platforms. ___1
 Paul Sture Switzerlandt   ------------------------------  $ Date: Fri, 8 Dec 2000 14:35:25 -0000+ From: "Tim Gray" <tim.gray@nospam_rl.ac.uk>   Subject: TZ87N compatible media.+ Message-ID: <90qrmr$s1i@newton.cc.rl.ac.uk>o  I I'm buying some new blank media for my TZ87N tape drive. I know I can use K DLT Tape III 10GB/20GB media, but notice there's a DLT Tape IIIXT 15GB/30GB K that's hardly different in price considering the quoted capacity. Is the XT H version just a longer length tape version, will it work ok in my TZ87N??   Cheers for any info -c   Tim.    / --- To use my email address, remove the nospam_    ------------------------------  # Date: Fri, 08 Dec 2000 08:58:40 GMTf From: "jme" <jme@elca.ch>tG Subject: Use of vmstar to exchange tape between AXP 2000 and Tru64 Unixa& Message-ID: <4p1Y5.118$_V6.45949@news>   Hello,  H To exchange data on tape between an AXP 2000 machine with V7.1 operatingJ system and an Unix Tru64 OSF V5.0 1094 alpha machine I chose vmstar 3.3.8.J I am having problems to read the tape under Tru64 when it is created under AXP.  5 I'm  using the following DCL script to write the tapen   $!- $! Backup a directory on tape with tar formatn $! $! JME 20.10.2000  $!% $ on error then goto error_processing  $! $ allocate/nolog MK: Tapem, $ mount/foreign/record=512/block=10240 Tape: $!
 $ sho time- $ tar :== $user4:[jme.tar_backup]vmstar.exe;1s $!1 $ tar/verbose/create Tape: user4:[.backup_dir...]u $!
 $ sho time $ dismount/nounload Tape:r $ deallocate Tape: $ deassign Tape  $! $ exit $! $ error_processing: ( $ write sys$output "An error occurs ..." $ exit $!  + Has anybody experienced the same problems ?n   ------------------------------  % Date: Fri, 08 Dec 2000 13:41:46 +0100W  From: Paul Sture <paul@sture.ch>$ Subject: Re: Very weird DCL behavior+ Message-ID: <VA.000001be.047a9ab5@sture.ch>d  < In article <yJHEn8xpzLK3@eisner.decus.org>, Rob Young wrote:, > From: young_r@eisner.decus.org (Rob Young) > Newsgroups: comp.os.vmsr& > Subject: Re: Very weird DCL behavior! > Date: 7 Dec 2000 11:39:10 -0500- >  >  > J >  Submit your batch jobs to EXECUTION queues with a JOB LIMIT of 1.  Only9 >  1 will be active at a time and the others wait, right?- > / >  Create a seperate queue for each tape drive:M >  >  DRIVE1_BATCHr >  DRIVE2_BATCH , etc. > B >  These are fed from a generic queue so as one job completes, theF >  next spools off to an execution queue.  The DCL to do the backup is> >  generic enough as it knows which tape drive to backup to by >  parsing the queue name. > C >  Assumption is it doesn't matter which backup goes to which tape t$ >  (i.e. you want to get them done). > B >  You can use a .DAT file to have all the volumes to be backed upA >  in there.  Or you may just be backing up files or somewhere ine= >  between.  We had our most critical volumes in there first.  > G >  And yes, I did a DISMOUNT/NOUNLOAD .. regarding rewinding... I don't G >  think so.  If it was rewinding, it didn't matter to me as my backupsvA >  were done in the morning :).  However, looking into this I did8E >  put a SET MAGTAPE command to make sure it was EOF (peeling throughrB >  a few DSN articles now I am unclear as to my reasoning but I doD >  know we did test restores and did do BACKUP/LIST , etc. from timeB >  to time).  Actually.. re-reading some of the DSN articles shows >  my reasoning was faulty:n > I > When a new saveset is added to a tape containing multiple savesets, thee9 > new saveset will have consecutive TM's, indicating EOV.p > A >  So my SET MAGTAPE command was overkill.  The question then is, 3 >  does a DISMOUNT/NOUNLOAD go to BOT?  Don't know.  >c` Yes, DISMOUNT/NOUNLOAD _does_ rewind the tape. This command can take some time, as it's waiting > for the rewind to complete before returning to the DCL prompt.  _ DISMOUNT/UNLOAD (the default) is instantaneous in returning to the DCL prompt, but carries the eW much higher penalty, when issued by mistake, that you have to use the robot utility to  _ physically unload and reload the tape. If you don't have a tape robot that means a trip to the e computer room :(  U Not quite the issue it used to be with reel to reel tapes (think _slow_ rewind), the t[ BACKUP/RELEASE_TAPE qualifier can save quite a bit of time in conjunction with /RECORD (or  L /DELETE), as it starts the rewind immediately it has finished with the tape.  Z Hmm. I don't think I've tried /RELEASE_TAPE with tape robots. Maybe time for some testing. >  > $ !  > $       gosub build_commandb > $ !a > $ ! ) > $       proc_name = "BCK ''disk_to_do'"e) > $       set process/name="''proc_name'"  > $ !f > $ !     Perform the Backup > $ !c > $       write sys$output ""oR > $       write sys$output "Beginning backup of ''disk_to_do' to ''tape_drive' at  > ''f$time()'"1 > $       on severe_error then call error_handlerrG > $       backup_command 'disk_path' 'tape_drive':'disk_to_do'.bck/save0M > $       write sys$output "Completed backup of ''disk_to_do' at ''f$time()'"n > $       write sys$output "". > $ !> > $ !1/ > $       set magtape/end_of_file 'tape_drive': ) > $       dismount/nounload 'tape_drive': 9 > $       open/write mailtmp 'this_dir'mailtmp.'this_pid'  > $       write mailtmp ""0 > $       write mailtmp " Backup command used: " > $       write mailtmp ""% > $       write mailtmp " ''piece1' " % > $       write mailtmp " ''piece2' " % > $       write mailtmp " ''piece3' " % > $       write mailtmp " ''piece4' " $ > $       write mailtmp " ''quals' " > $       write mailtmp "" > $       close/nolog mailtmp M > $       mail/sub="''DISK_TO_DO' ''TAPE_DRIVE' ''LABEL_NAME' ''DISK_PATH'" - 3 >         'this_dir'mailtmp.'this_pid'  backup_logs  > $       delete = "delete" . > $       delete 'this_dir'mailtmp.'this_pid'; > $       exit >  >  > $ build_command: > $       quals = "" > $ !  > $ ! 7 > $       if backup_method - "INCR" .eqs. backup_method  > $       then* > $               quals = quals + "/IMAGE". > $               disk_path = "''disk_to_do':" > $       else6 > $               quals = quals + "/SINCE=BACKUP/FAST"9 > $               disk_path = "''disk_to_do':[*...]*.*;*"r > $       endifi > $ !< > $ ! 7 > $       if backup_method - "RECO" .nes. backup_methodb > $       then+ > $               quals = quals + "/RECORD"  > $       endifr > $ !i > $ ! 8 > $       piece1 = "BACKUP/NOREWIND/LABEL=''label_name'"3 > $       piece2 = "/IGNORE=(INTERLOCK,LABEL_PROC)"sJ > $       piece3 = "/JOURNAL=TS$BUL:''YYYYMMDD'_''DISK_TO_DO'_''TAPE'.BJL"& > $       piece4 = "/BLOCK_SIZE=65535"0 > $       piece5 = "/MEDIA_FORMAT=''COMPACTION'"M > $       backup_command = piece1 + piece2 + piece3 + piece4 + piece5 + qualst > $       return >  > H No, I can't see what the SET MAGTAPE/END_OF_FILE is getting you, either. > ` > In article <heOX5.783$He.10598@wagner.videotron.net>, "Syltrem" <syltrem@videotron.ca> writes:M > > I know I use subproprocesses (hey! I'm not that dumb! :-) ). The the realoI > > question about using one job with subprocesses or many jobs, is: DoesaJ > > DISMOUNT/NOUNLOAD rewind the tape? - don't get me wrong, rewinding andO > > unloading are 2 separate functions :-) I thought it did but I may be wrong.o > > O > > If it does not, then I could use SUBMIT with SYNCHRONIZE but... If I submittO > > 4 backup jobs, I can only synchronize with ONE of them. If this one takes 2gN > > hours to finish and the other 3 run in 5 minutes, I will loose 3 times oneL > > hour and 55 minutes of tape drive availability because my main procedureN > > will "unfreeze" only after that one very long backp is finished.. No good.= > > That makes me loose almost 6 hours of usable backup time.1 > > J > > Unless I can implement some sort of AST mechanism in DCL and watch forM > > completion of any of  the 4 batch jobs (that's exactly what my mailbox isoK > > doing, by the way) then SYNCH can't do the job. I could use one job pera4 > > magtape but that's not what what I want, either. > > L > > I think I will end up writing a 15 lines Basic program to handle the I/OK > > from this damn mailbox. DCL like it or not. It'll be the fastest way of  > > fixing everything. > > J > > Thanks for all replies (and if someone could answer the question about5 > > no-unloading/rewinding above, I will appreciate).: > >  > > Syltremg > >  >C   ___.
 Paul Sture Switzerland4   ------------------------------  $ Date: Fri, 8 Dec 2000 13:40:02 -0500& From: "Syltrem" <syltrem@videotron.ca>$ Subject: Re: Very weird DCL behavior7 Message-ID: <NR9Y5.2146$He.115852@wagner.videotron.net>s  J That's a good way of doing it. The main job would submit a bunch of backup9 jobs to the generic queue. No need for SYNCH or anything.e  @ I will do a test for the rewind question when I have a minute...  K OK. The minutes has finally come. It does rewind the tape, but getting backoE to EOT is pretty fast. EOT follows a save set that took 45 minutes toh4 record, but getting back to EOT take about 1 minute.   Syltrem   7 "Rob Young" <young_r@eisner.decus.org> wrote in message % news:yJHEn8xpzLK3@eisner.decus.org...g >h >wI > Submit your batch jobs to EXECUTION queues with a JOB LIMIT of 1.  Only 8 > 1 will be active at a time and the others wait, right? >i. > Create a seperate queue for each tape drive: >  > DRIVE1_BATCH > DRIVE2_BATCH , etc.  > A > These are fed from a generic queue so as one job completes, theeE > next spools off to an execution queue.  The DCL to do the backup isr= > generic enough as it knows which tape drive to backup to byu > parsing the queue name.s >aA > Assumption is it doesn't matter which backup goes to which tapeo# > (i.e. you want to get them done).  >hA > You can use a .DAT file to have all the volumes to be backed up @ > in there.  Or you may just be backing up files or somewhere in< > between.  We had our most critical volumes in there first. >nF > And yes, I did a DISMOUNT/NOUNLOAD .. regarding rewinding... I don'tF > think so.  If it was rewinding, it didn't matter to me as my backups@ > were done in the morning :).  However, looking into this I didD > put a SET MAGTAPE command to make sure it was EOF (peeling throughA > a few DSN articles now I am unclear as to my reasoning but I dolC > know we did test restores and did do BACKUP/LIST , etc. from timeyA > to time).  Actually.. re-reading some of the DSN articles showse > my reasoning was faulty: >iI > When a new saveset is added to a tape containing multiple savesets, theP9 > new saveset will have consecutive TM's, indicating EOV.e > @ > So my SET MAGTAPE command was overkill.  The question then is,2 > does a DISMOUNT/NOUNLOAD go to BOT?  Don't know. >r >u > $ !n > $       gosub build_commanda > $ !y > $ !y) > $       proc_name = "BCK ''disk_to_do'" ) > $       set process/name="''proc_name'"o > $ !t > $ !     Perform the Backup > $ !  > $       write sys$output ""o@ > $       write sys$output "Beginning backup of ''disk_to_do' to ''tape_drive' at > ''f$time()'"1 > $       on severe_error then call error_handleriG > $       backup_command 'disk_path' 'tape_drive':'disk_to_do'.bck/savec@ > $       write sys$output "Completed backup of ''disk_to_do' at ''f$time()'" > $       write sys$output ""c > $ !u > $ ! / > $       set magtape/end_of_file 'tape_drive':e) > $       dismount/nounload 'tape_drive':b9 > $       open/write mailtmp 'this_dir'mailtmp.'this_pid'n > $       write mailtmp ""0 > $       write mailtmp " Backup command used: " > $       write mailtmp ""% > $       write mailtmp " ''piece1' "a% > $       write mailtmp " ''piece2' "b% > $       write mailtmp " ''piece3' "a% > $       write mailtmp " ''piece4' "-$ > $       write mailtmp " ''quals' " > $       write mailtmp "" > $       close/nolog mailtmpc= > $       mail/sub="''DISK_TO_DO' ''TAPE_DRIVE' ''LABEL_NAME's ''DISK_PATH'" -13 >         'this_dir'mailtmp.'this_pid'  backup_logs1 > $       delete = "delete"a. > $       delete 'this_dir'mailtmp.'this_pid'; > $       exit >r >s > $ build_command: > $       quals = "" > $ !o > $ !i7 > $       if backup_method - "INCR" .eqs. backup_method  > $       then* > $               quals = quals + "/IMAGE". > $               disk_path = "''disk_to_do':" > $       else6 > $               quals = quals + "/SINCE=BACKUP/FAST"9 > $               disk_path = "''disk_to_do':[*...]*.*;*"  > $       endifc > $ !t > $ !u7 > $       if backup_method - "RECO" .nes. backup_methodT > $       then+ > $               quals = quals + "/RECORD"o > $       endife > $ !i > $ !t8 > $       piece1 = "BACKUP/NOREWIND/LABEL=''label_name'"3 > $       piece2 = "/IGNORE=(INTERLOCK,LABEL_PROC)" J > $       piece3 = "/JOURNAL=TS$BUL:''YYYYMMDD'_''DISK_TO_DO'_''TAPE'.BJL"& > $       piece4 = "/BLOCK_SIZE=65535"0 > $       piece5 = "/MEDIA_FORMAT=''COMPACTION'"G > $       backup_command = piece1 + piece2 + piece3 + piece4 + piece5 +n qualsv > $       return >s >d > Robs >I > A > In article <heOX5.783$He.10598@wagner.videotron.net>, "Syltrem"m <syltrem@videotron.ca> writes:H > > I know I use subproprocesses (hey! I'm not that dumb! :-) ). The the realI > > question about using one job with subprocesses or many jobs, is: DoesdJ > > DISMOUNT/NOUNLOAD rewind the tape? - don't get me wrong, rewinding andH > > unloading are 2 separate functions :-) I thought it did but I may be wrong. > > H > > If it does not, then I could use SUBMIT with SYNCHRONIZE but... If I submitG > > 4 backup jobs, I can only synchronize with ONE of them. If this onen takes 2oJ > > hours to finish and the other 3 run in 5 minutes, I will loose 3 times oneoL > > hour and 55 minutes of tape drive availability because my main procedureH > > will "unfreeze" only after that one very long backp is finished.. No good.u= > > That makes me loose almost 6 hours of usable backup time.  > >mJ > > Unless I can implement some sort of AST mechanism in DCL and watch forJ > > completion of any of  the 4 batch jobs (that's exactly what my mailbox isK > > doing, by the way) then SYNCH can't do the job. I could use one job per 4 > > magtape but that's not what what I want, either. > > L > > I think I will end up writing a 15 lines Basic program to handle the I/OK > > from this damn mailbox. DCL like it or not. It'll be the fastest way ofw > > fixing everything. > >sJ > > Thanks for all replies (and if someone could answer the question about5 > > no-unloading/rewinding above, I will appreciate).  > >0 > > Syltrem  > >m   ------------------------------  # Date: Fri, 08 Dec 2000 07:40:09 GMTo From: b_con@my-deja.come& Subject: Re: Write out char$(9) in DCL) Message-ID: <90q38p$uv5$1@nnrp1.deja.com>0  7 In article <200012020743_MC2-BCDF-569F@compuserve.com>,o5   "Richard B. Gilbert" <DRAGON@compuserve.com> wrote:uE >         How about F$FAO("!_").   This is an under utilized functions that = > is* > capable of all sorts of nice formatting! >n+ > Message text written by "Mark-Simon Pope" F > >    I need to write out a tab delimited file, I believe I'd need to do i=m > t  > byE > writing out an ASCII character 9, however, I can't seem to recall a  metho= > dI > to do this in DCL. > <d >t= > You can define a symbol TAB with the ascii value 9 this wayo TAB[0,7]=9  (or ==).5 then you can use the symbol in your WRITE command, exf( $ WRITE OUT TAB, "abcdefg", TAB, "12345"   Bendix Riis-    & Sent via Deja.com http://www.deja.com/ Before you buy.S   ------------------------------   End of INFO-VAX 2000.685 ************************