1 INFO-VAX	Thu, 29 Jun 2000	Volume 2000 : Issue 361       Contents: Re: Affordable Debate Summary  Re: Amazon? # Re: college experience with OpenVMS # Re: college experience with OpenVMS # Re: college experience with OpenVMS  Re: Compaq advertizes & Re: Compaq paying for software ports ?! DCPS F1.8 expires; G1.8 available  DE500 on Alpha VMS 6.2% Re: Does VMS require a SYSTEM account % Re: Does VMS require a SYSTEM account % Re: Does VMS require a SYSTEM account % Re: Does VMS require a SYSTEM account % Re: Does VMS require a SYSTEM account  Re: Email quotas& Re: good news (for me,  I think) . . .A How do you Use LCKPAG on ALPHA to lock a full process in memory ? E Re: How do you Use LCKPAG on ALPHA to lock a full process in memory ? < Re: How to create a native VMS command - and a few questions$ how to register a node in decnet osi( Re: how to register a node in decnet osi( Re: how to register a node in decnet osi% Re: LPR from VMS v6.2 to Linux Redhat % Re: LPR from VMS v6.2 to Linux Redhat % Re: LPR from VMS v6.2 to Linux Redhat 7 Re: Northern Light vs. Google (and the winner is . . .) 7 Re: Northern Light vs. Google (and the winner is . . .) 7 Re: Northern Light vs. Google (and the winner is . . .) 7 Re: Northern Light vs. Google (and the winner is . . .) 7 RE: Northern Light vs. Google (and the winner is . . .) 7 Re: Northern Light vs. Google (and the winner is . . .) 7 Re: Northern Light vs. Google (and the winner is . . .) 7 Re: Northern Light vs. Google (and the winner is . . .) 7 Re: Northern Light vs. Google (and the winner is . . .) 7 Re: Northern Light vs. Google (and the winner is . . .) 7 Re: Northern Light vs. Google (and the winner is . . .) 7 Re: Northern Light vs. Google (and the winner is . . .) 1 re: Northern Light vs. Google (power consumption) 1 Re: Northern Light vs. Google (power consumption) 1 Re: Northern Light vs. Google (power consumption) . Re: OpenVMS clusters vs other systems clusters. Re: OpenVMS clusters vs other systems clusters. re: OpenVMS clusters vs other systems clusters. Re: OpenVMS clusters vs other systems clusters. Re: OpenVMS clusters vs other systems clusters. Re: OpenVMS clusters vs other systems clusters. Re: OpenVMS clusters vs other systems clusters. Re: OpenVMS clusters vs other systems clusters. RE: OpenVMS clusters vs other systems clusters2 Re: OpenVMS loses big, was:  RE: Compaq advertises2 Re: OpenVMS loses big, was:  RE: Compaq advertises2 RE: OpenVMS loses big, was:  RE: Compaq advertises2 Re: OpenVMS loses big, was:  RE: Compaq advertises2 Re: OpenVMS loses big, was:  RE: Compaq advertises2 Re: OpenVMS loses big, was:  RE: Compaq advertises1 Re: OpenVMS loses big, was: RE: Compaq advertises  Re: OVMS 7.2 updates Re: PARITY ERROR! Perl suddenly hangs on VMS system ! RS232 CTS/RTS flow control on VAX  Satallite not booting  Re: Satallite not booting  Re: Satellite not booting  Re: See how a device is spooled 
 Re: SMTP Mail  Re: SMTP Mail - followupJ Re: Sun Porting Clearinghouse for Staroffice, anybody going to port toVMS? Re: VAX on Intel?   F ----------------------------------------------------------------------  + Date: Thu, 29 Jun 2000 13:16:39 +0000 (UTC) ' From: Osmo Kujala <kujala@tukki.jyu.fi> & Subject: Re: Affordable Debate Summary, Message-ID: <8jfi7n$o5c$1@mordred.cc.jyu.fi>  6 David J. Dachtera <djesys.nospam@earthlink.net> wrote: > Bill Todd wrote:
 >> ... you" >> aren't getting more support ...  J > Gee, Bill I looked at the posts from Roger Barnett, Richard Jordan, Hank= > Van der Waal, ... those posts sure look like support to me!     Me too. I agree more with David.J Bill has has been trying to teach us how the reasoning at Compaq might go. Thanks to him too for that.   D I see the situation very simple. VMS has potential to be popular allF purpose operating system. Simply because it's best. (Well there may be3 better which I don't know much: Plan9?, Xanadu?...)   B It's a shame that DEC and Compaq haven't really tried to sell VMS.G If they continue like now VMS will wither away. To mee VMS is dead when E it's only used in small niche. (And soon there will be others in that  niche too.)   D First thing to get VMS popular is making it affordable at low level.@ Price of DS10 with VMS may be fair but what we need is PC level.D To be exact I'd like to see that Specfp/dollar should be better thanC what Wintel has. That has been possible with old Alpha motherboards B and Hobbyist license. So I don't think it's completely impossible.  E Good help to get popularity and commodity prices would be VMS port to F PC. That might be IA64, IA32, Crusoe or something else, but it must be> popular hardware. To get fast effect it should be X86 I guess.G The most important thing here is that VMS must run in popular hardware! I Actually it does now in CharonVAX, but will that be affordable and usable F I don't know. The popular hardware may well contain Alpha processor orB emulate Alpha or VAX, but VMS must be the real thing. Licenses may= then be some new PC category, and support may be lower level.   C CharonVAX kind of proves that VMS in X86 is possible. Could they at H Compaq do something innovative for example port parts of VMS and emulate& something; or use code morphing or ...  E I don't believe it's big problem for Compaq to make very low cost VMS H PC model if they will. So far they (DEC mostly) have done many tricks to* prevent VMS running in too cheap hardware.  F To get VMS popular, popular software should run in it. I'd like to seeF that everything needed will be in Linux (and Windows will decline), soK it might be enough to get Linux software to run in VMS easily. When number  F of VMS systems increases there will be more software (ports and native builds).  I What is the real reason that Compaq people seem to be reluctant to spread  VMS?E Security by obscurity might be one. Customers with big money may feel  safer like now. J Maybe they at DEQ really know how bad VMS is to port, so they desided thatF after Alpha there will be no VMS. Maybe VMS isn't so wonderful we haveJ thought? Other companies doesn't seem to have so much problem with porting to new hardware.  ! just some of my personal thoughts .                                    Osmo Kujala   ------------------------------  % Date: Wed, 28 Jun 2000 23:14:23 -0400 * From: David A Froble <davef@tsoft-inc.com> Subject: Re: Amazon?- Message-ID: <395ABF0F.224714C5@tsoft-inc.com>   ( Andrew Harrison SUNUK Consultancy wrote:  K > But don't fool yourself that the GS boxes performing faster will make any 
 > difference, K > if performance had been a real buying criteria then Amazon would not have  > bought$ > the HP V-Series in the first case.  5 Once again, the 'blind dog' finally finds a bone. :-)   O (Sorry Andrew, but if I have to 'kick' someone, guess it should be someone from  Sun.)   P The above appears to be the reality.  Just that simple.  Do the people at AmazonP really care what systems they are running, as long as the task gets done?  Not aO chance!  Are they ready to promise anything to get into what they see as a good O business?  Damn right!  If the HP systems don't get the job done, will there be M another change in course?  Damn right!  It's a bit like the movies, where the O mafia hit man says to his soon to be ex-drinking buddy, "Hey, nothing personal,  it's strickly business".   Dave   --  4 David Froble                       Tel: 724-529-04504 Dave Froble Enterprises, Inc.      Fax: 724-529-0596> DFE Ultralights, Inc.              E-Mail: davef@tsoft-inc.com6 T-Soft, Inc.  170 Grimplin Road  Vanderbilt, PA  15486   ------------------------------  % Date: Wed, 28 Jun 2000 23:49:15 -0400 * From: David A Froble <davef@tsoft-inc.com>, Subject: Re: college experience with OpenVMS- Message-ID: <395AC73B.ABB2E35C@tsoft-inc.com>    David Mathog wrote:  > K > Moreover, there are no signs whatsoever of a "renaissance" for OpenVMS in C > academia.  "Last dying twitches" is a closer description.  Compaq G > may be selling some monster servers to corporations, but it's selling K > precious little OpenVMS to colleges.  In fact, the only sales that I know G > of are to groups that are looking for faster machines to run existing G > OpenVMS software.  I'd be amazed if there are any "new" sales of this D > platform in the academic market at all, with the remotely possiblyM > exception that there may be sales of large servers for business purposes at  > the larger state schools.  > M > Compaq has given academics no reason whatsoever to stay with this platform: M > the OS itself and the hardware it runs on cost too much, there is virtually E > no native software, only limited quantities of ported software, the G > academic program stinks in comparison to the competition, an end user E > requires incredible determination and must be nearly clairvoyant to J > actually complete a purchase successfully, and now, in what is certainlyL > the final straw, it turns out that on the same hardware, Linux outperformsK > OpenVMS by huge factors under typical work loads for these sorts of small  > machines.   7 Sorry David, gotta set the record straight on this one.   L Yes, the past was dismal.  Today is the first day of the rest of your life. 7 Live the future, or live in the past, it's your choice.   P The VMS people are in the process of giving VMS licenses 'free' to academia, andM realize that such a move is not enough, and are also addressing the issues of D hardware and faculty.  This is real!  This is today, and the future.  E Don't know where it will go, specifically in academia, but there is a L renaissance that's starting to happen with VMS.  It'll go as far as everyoneH involved will let it go.  So, will you continue to live in the past, andO complain about regretable past actions, or will you embrace the future, and the = efforts the VMS people are and will make to reach out to you?   O So, free VMS licensing, who knows what on hardware, and efforts to help you use  VMS in education.  Interested?   Dave   --  4 David Froble                       Tel: 724-529-04504 Dave Froble Enterprises, Inc.      Fax: 724-529-0596> DFE Ultralights, Inc.              E-Mail: davef@tsoft-inc.com6 T-Soft, Inc.  170 Grimplin Road  Vanderbilt, PA  15486   ------------------------------   Date: 29 Jun 2000 15:37:35 GMT1 From: bill@triangle.cs.uofs.edu (Bill Gunshannon) , Subject: Re: college experience with OpenVMS, Message-ID: <8jfqfv$1343$1@info.cs.uofs.edu>  - In article <395AC73B.ABB2E35C@tsoft-inc.com>, -  David A Froble <davef@tsoft-inc.com> writes:  |>  : |> Sorry David, gotta set the record straight on this one. |>  O |> Yes, the past was dismal.  Today is the first day of the rest of your life.  : |> Live the future, or live in the past, it's your choice. |>  S |> The VMS people are in the process of giving VMS licenses 'free' to academia, and P |> realize that such a move is not enough, and are also addressing the issues ofG |> hardware and faculty.  This is real!  This is today, and the future.   K Sorry Dave, way too little, way too late.  We have moved yet another course J off of VMS.  And this was the first, introductory CS1 course.  It has leftL VMS in favor of UNIX and/or Windows platforms.  That leaves only two coursesJ in the entire CS/CIS curriculum that will be done on VMS.  And long beforeP a student gets to this stage of his coursework he will have become very familiarK with other OSes.  Thus making learning VMS a mere anoyance and guaranteeing M that they will learn the bear minimum necessary to get through these courses.    |>  H |> Don't know where it will go, specifically in academia, but there is aO |> renaissance that's starting to happen with VMS.  It'll go as far as everyone K |> involved will let it go.  So, will you continue to live in the past, and R |> complain about regretable past actions, or will you embrace the future, and the@ |> efforts the VMS people are and will make to reach out to you?  J If Compaq came in here today and gave every faculty member a desktop AlphaJ with VMS and CDE I am not sure how many would actually use it and how manyI would let it sit ont he edge of their desk while they continued to do all  their real work on PC's.  H If Compaq gave me enough Alphas with VMS and CDE to set up an entire labH I expect they would end out being used as little more than Xterminals to& get to the other servers running UNIX.  H Sad but true, but I would be more than glad to have Compaq try and prove me wrong.  ;-)   |>  R |> So, free VMS licensing, who knows what on hardware, and efforts to help you use! |> VMS in education.  Interested?   G I'm interested, the question is; Is there anyone interested who is in a + position to actually do anything about it??    bill   --  J Bill Gunshannon          |  de-moc-ra-cy (di mok' ra see) n.  Three wolvesD bill@cs.scranton.edu     |  and a sheep voting on what's for dinner. University of Scranton   |A Scranton, Pennsylvania   |         #include <std.disclaimer.h>       ------------------------------   Date: 29 Jun 2000 16:53:51 GMT1 From: bill@triangle.cs.uofs.edu (Bill Gunshannon) , Subject: Re: college experience with OpenVMS, Message-ID: <8jfuuv$159c$1@info.cs.uofs.edu>  , In article <8jfqfv$1343$1@info.cs.uofs.edu>,4  bill@triangle.cs.uofs.edu (Bill Gunshannon) writes:0 |> In article <395AC73B.ABB2E35C@tsoft-inc.com>,0 |>  David A Froble <davef@tsoft-inc.com> writes: |>M |> If Compaq came in here today and gave every faculty member a desktop Alpha M |> with VMS and CDE I am not sure how many would actually use it and how many L |> would let it sit ont he edge of their desk while they continued to do all |> their real work on PC's.  |>  K |> If Compaq gave me enough Alphas with VMS and CDE to set up an entire lab K |> I expect they would end out being used as little more than Xterminals to ) |> get to the other servers running UNIX.  |>  K |> Sad but true, but I would be more than glad to have Compaq try and prove  |> me wrong.  ;-)  |>   |> |> U |> |> So, free VMS licensing, who knows what on hardware, and efforts to help you use $ |> |> VMS in education.  Interested? |>  J |> I'm interested, the question is; Is there anyone interested who is in a. |> position to actually do anything about it??  D I know it's unusual to respond to one's self, but this actually made me think about this somemore.   B There probably is a way to get a program like this going.  Porting@ some of the Open Source stuff loating around the UNIX world that@ provide the desktop apps that VMS seems to lack (like StarOfficeB for example!!) would make useful and interesting student projects.@ But it would take a lot of support from Compaq especially in the> form of hardware/sfotware donations.  While I can see a strong; potential in something like this, the people at the funding > approval level have been pretty much brainwashed by MS at this@ point and until they can be won back they are unlikely to commit@ any funds at all to something they have probably never heard of,	 like VMS.   D If it sounds like I'm begging, I am.  Most colleges and universitiesB are not profit making organizations.  That is usually reflected inC tight budgets that are controlled by pointy-hairs and not by people A who actually know anything about technology.  I'm used to begging A for equipment to try and support our students.  That's why I have A three samll VAXen and am frequently seen here and elswhere trying C to talk people out obsolete hardware (like TK50's, M7546's, KDA50's E and even RLV21's, hint, hint).  I am sure that there are more schools ? like us than the Harvards and the MITs with virtually unlimited  budgets.  @ So, it's been said before, what can be done about it??  Is there@ any likelyhood that Compaq will look far enough into the future > to see the potential of seeding colleges and universities like us??  ) I guess we'll have to just wait and see!!   
 All the best.    bill   --  J Bill Gunshannon          |  de-moc-ra-cy (di mok' ra see) n.  Three wolvesD bill@cs.scranton.edu     |  and a sheep voting on what's for dinner. University of Scranton   |A Scranton, Pennsylvania   |         #include <std.disclaimer.h>       ------------------------------  # Date: Thu, 29 Jun 2000 12:55:31 GMT 0 From: "Terry C. Shannon" <shannon@world.std.com> Subject: Re: Compaq advertizes& Message-ID: <Fwx2E0.EwL@world.std.com>  J "Brian Schenkenberger, VAXman-" <system@SendSpamHere.ORG> wrote in message* news:009EC494.5BEAB368@SendSpamHere.ORG...L > In article <910612C07BCAD1119AF40000F86AF0D80528449A@kaoexc4.kao.dec.com>,- "Main, Kerry" <Kerry.Main@compaq.com> writes:S0 > >Wait until tomorrows (June 28) announcements. >2L > June 28 is here and more than 1/2 gone.  If there's an announcement, where > was it made?  0 Must have been the Compaq/Oracle announcement...  E Compaq and Oracle Align to Offer Customers of All Sizes an Integrated  Missionn   Critical E-Business Platform  ; HOUSTON and REDWOOD SHORES, Calif., June 28 /PRNewswire/ -- F (http://www.oracle.com/tellmemore/?228177) Compaq Computer CorporationF (NYSE: CPQ) and Oracle Corporation (Nasdaq: ORCL) today announced thatF Compaq will adopt the Oracle Internet Platform across its full line ofE server and storage offerings to provide customers of all sizes with axK reliable, scalable, and complete mission critical e-business platform. WithnL this agreement, the companies will jointly optimize and test this Compaq andH Oracle E-Business Platform, and will deliver this offering through their$ professional services organizations.  D The Oracle Internet Platform will be optimized for Compaq servers --J including ProLiant(TM) running Windows NT(R)/Windows 2000(TM), AlphaServerI running Tru64 UNIX(TM) or OpenVMS(TM), and the companies plan to optimize.I for Himalaya(TM) running NonStop(TM) Kernel (NSK) -- and StorageWorks(TM)LE storage offerings. This Compaq and Oracle E-Business Platform will belL available from Compaq during calendar 4Q00. With this agreement, Compaq will6 also adopt internally the Oracle Internet Platform....   ------------------------------  % Date: Thu, 29 Jun 2000 01:13:56 -0400n* From: David A Froble <davef@tsoft-inc.com>/ Subject: Re: Compaq paying for software ports ?e- Message-ID: <395ADB14.D8E7D851@tsoft-inc.com>e  ( Andrew Harrison SUNUK Consultancy wrote: > " > Arrrrrrrggggggggggggghhhhhhhhhh.  N Hey, pretty good Andrew.  But, could you hold that growl just a bit longer?  I do like blue. :-)e  E > Where are you going to get the apps from if you don't get them fromv? > UNIX. What would you prefer applications that require a Win32l( > library hosted on OpenVMS to run ????? > H > They don't grow on trees and people don't just go off an knock off say< > an Office suite in Fortran or Pascal at the drop of a hat. > E > Hint, hint, hint, most of the commercial apps that currently run ont< > OpenVMS are developed on UNIX and key bits of OpenVMS were > also developed on UNIX.:  I First, before I make a statement, I'd ask just what is your definition ofVP 'commercial apps'.  Then, I'd probably choose to disagree with you regardless ofL the answer.  David Mathog also claimed that 95% of VMS apps come from Unix. N Maybe true for his biology type apps.  Not everybody is a biologist.  Most VMSC apps probably are custom developed, on VMS, to meet specific needs.n  G > If you have an IP stack on OpenVMS it probably came from one or other-F > flavour of UNIX, are you suggesting that people should not use IP onE > OpenVMS because the IP stack will pollute the purity of the OpenVMS: > code base.  K Actually, I prefer DECnet.  May not be the world's standard.  But it's moretN secure, robust (in a VMS environment), and easy to use.  Ok, just gave you theP lead in for some comment about small fish in a smaller pond, or such.  Go ahead, I know you can do it.k  I > If your sentiments are shared by even a small percentage of the OpenVMS D > advocates out there then sorry but OpenVMS is doomed. Nothing doesG > for an OS quicker than the impression held by senior IT managers thatiE > the only people who know anything about it are unsocialised weirdostM > with extreme opinions that could get them sectioned if they arn't carefull.k > :-)h  M Sectioned?  Is that fun?  Try it and let me know.  Unsocialized?  Yeah, I cano handle that.   Dave :-)   --  4 David Froble                       Tel: 724-529-04504 Dave Froble Enterprises, Inc.      Fax: 724-529-0596> DFE Ultralights, Inc.              E-Mail: davef@tsoft-inc.com6 T-Soft, Inc.  170 Grimplin Road  Vanderbilt, PA  15486   ------------------------------  # Date: Thu, 29 Jun 2000 15:00:36 GMTe+ From: Paul Anderson <panderson@genicom.com>h* Subject: DCPS F1.8 expires; G1.8 availableC Message-ID: <panderson-D676CB.11004129062000@news.bellatlantic.net>o   Dear friends of DCPS,o  B We are pleased to announce that an updated field test of DECprint ; Supervisor (DCPS) V1.8, called DCPS G1.8, is now available.t  H Anyone running the previous field test version (F1.8) should upgrade to G the new version immediately, as F1.8 expires on July 1.  Although DCPS cD F1.8 print queues will not fail after July 1, execution queues once  stopped will not restart.a  D This is the final field test version before the release of V1.8, so D please report any problems with the software or documentation to me.  I Updated kits and documentation are available on a password-protected FTP XH site.  If you do not know the location and would like to run DCPS G1.8,  please let me know.d   Paul   -- h"    Paul Anderson, DCPS Engineering"    GENICOM Corporation, Gardner MA   ------------------------------  # Date: Thu, 29 Jun 2000 13:39:31 GMTE/ From: "John Nixon" <jorlnixon@worldnet.att.net>u Subject: DE500 on Alpha VMS 6.2oF Message-ID: <nkI65.14213$kK.434722@bgtnsc05-news.ops.worldnet.att.net>  G In preparing for our Alpha VMS upgrade, we copied a system disk from angK Alpha 2100 VMS 6.2 system to another Alpha 2100 that in a previous life wasSJ running Windows NT.   We had already loaded VMS 7.2-1 on this 2100 with noJ problem.  We plan to run through a mock upgrade from VMS 6.2 to 7.2-1 withC all patches and layered products etc... on the test sytsem prior tom  upgrading the production system.  I The problem is that when we attempt to boot the new 2100 under VMS 6.2 itlI crashes,  (I hate to give incomplete error messages, but the last messagehI was something about a setting on the ethernet port.   I am repeating thiswD message from someone else that actually saw it, but is not currently available).   G The 2100 has a DE500 fast ethernet card.  Will this work under VMS 6.2? H Could it cause the crash?   Are there any settings that will force it to work under VMS 6.2?.  K We have a couple of DEFPA-DAs (fddi) that we plan to install so we can loadoE 6.2 and proceed with the upgrade.  Is that a viable workaround?  (our L network has all sorts of hubs and switches so we can use either fddi or fast or slow ethernet).   ------------------------------  % Date: Thu, 29 Jun 2000 08:38:38 +0100a< From: John Macallister <j.macallister1@physics.oxford.ac.uk>. Subject: Re: Does VMS require a SYSTEM account6 Message-ID: <000629083838.670@nplvms.physics.ox.ac.uk>  > > What happens to a file owned by a UIC which does not exist ?  @ Nothing. We have many files here belonging to non-existent UIC's=  for one reason or another (usually usernames which have been   removed from the UAF).f   John   ------------------------------  % Date: Thu, 29 Jun 2000 08:33:54 +0100D< From: John Macallister <j.macallister1@physics.oxford.ac.uk>. Subject: Re: Does VMS require a SYSTEM account6 Message-ID: <000629083354.670@nplvms.physics.ox.ac.uk>  E VMS requires a system account. There has to be at least one username.lD Otherwise how would one login? You cannot delete the system account. John   ------------------------------    Date: 29 Jun 2000 09:16:51 -0500 From: briggs@eisner.decus.org-. Subject: Re: Does VMS require a SYSTEM account+ Message-ID: <ftybDCxO7IEW@eisner.decus.org>f  u In article <000629083354.670@nplvms.physics.ox.ac.uk>, John Macallister <j.macallister1@physics.oxford.ac.uk> writes:wG > VMS requires a system account. There has to be at least one username.fF > Otherwise how would one login? You cannot delete the system account.  ' You seem to be making three statements.   " 1.  VMS requires a system account.   No.  It doesn't.  A 2.  There has to be at least one username else you cannot log in.a  = No.  There doesn't.  There are ways and ways of "logging in".   ) 3.  You cannot delete the system account.   B Yes, you can.  Although I've never tried to do so using AUTHORIZE.    G Have you ever booted your system with a missing SYSUAF file?  It works.r  & 	John Briggs			briggs@eisner.decus.org   ------------------------------    Date: 29 Jun 2000 09:50:49 -0500, From: koehler@eisner.decus.org (Bob Koehler). Subject: Re: Does VMS require a SYSTEM account+ Message-ID: <PR5fDQrPdAHG@eisner.decus.org>y  ^ In article <slkbod98jev84@corp.supernews.com>, "Peter Weaver" <peter.weaver@stelco.ca> writes:E > I am on the phone with a support person from the OpenVMS ManagementeA > Station support group and she is telling me that the ManagementaB > Station software *HAS* to run under the SYSTEM account. I am 99%H > confident that VMS does not require a SYSTEM account. I would considerE > any product that *has* to run from the SYSTEM account to be broken.,F > Does anyone know (I am still on the phone so it is difficult to lookF > up right now) if running VMS without a SYSTEM account is a supported > configuration.  F   I don't know about supported, but we've run a system with the SYSTEM   account disuser'ed for years.e  H   Boot time startups don't check the UAF before starting processes underH   SYSTEM.  We set up other accounts with readall to do backups and such,D   and we found that you can still log in as SYSTEM but only from the&   console, which was in a locked room.   ------------------------------  % Date: Thu, 29 Jun 2000 01:31:01 -0400y* From: David A Froble <davef@tsoft-inc.com>. Subject: Re: Does VMS require a SYSTEM account- Message-ID: <395ADF15.33DEAC14@tsoft-inc.com>p   JF Mezei wrote:e >  > Peter Weaver wrote:oH > > Does anyone know (I am still on the phone so it is difficult to lookH > > up right now) if running VMS without a SYSTEM account is a supported > > configuration. > O > For a long time, I had a DISUSER floag on SYSTEM and experienced no problems.IN > I r-enabled it so that I could submit  batch jobs during startup (although I5 > could have used SUBMIT/USER in the systartup file).  > J > However, I had another cryptic account name with the same UIC as SYSTEM. > > > What happens to a file owned by a UIC which does not exist ?  J Uh.. could you define a 'UIC that doesn't exist'?  If you're talking aboutP SYSUAF records, that's a totally different entity.  Just as a directory does notP have to be tied to a SYSUAF record.  If there's a file with a specific UIC, then0 that UIC exists on that disk, and on the system.   Dave   -- v4 David Froble                       Tel: 724-529-04504 Dave Froble Enterprises, Inc.      Fax: 724-529-0596> DFE Ultralights, Inc.              E-Mail: davef@tsoft-inc.com6 T-Soft, Inc.  170 Grimplin Road  Vanderbilt, PA  15486   ------------------------------  % Date: Thu, 29 Jun 2000 14:42:38 +0200 = From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>f Subject: Re: Email quotass) Message-ID: <395B443E.56B5162B@gtech.com>M   Horse Nuts wrote:dB > ho do I go about turning off email quotas for myself, on 7.1....   What email quota ?   Arne   ------------------------------  % Date: Thu, 29 Jun 2000 01:49:44 -0400o* From: David A Froble <davef@tsoft-inc.com>/ Subject: Re: good news (for me,  I think) . . .n- Message-ID: <395AE378.46993CCE@tsoft-inc.com>b   John Nixon wrote:  > L > After several false starts, and millions of dollars spent, we have finally > decided that it is better toI > migrate our VAX applications to Alpha VMS than to rewrite them on Unix.y > I > Well, at least we have decided that we need to move the applications ton > Alpha VMS for now N > because we don't know how long it will take us to move to unix, and we don't > want to be > stuck on VAX indefinitely. > G > SOooooo,  I am probably going to be asking a lot of VAX to Alpha typep > questions. >  > Here is a starter question:o > D > I am intimately aware of the performance problems related to large > directories on pre VMS7.2.M > I am also aware of the improvements to large directory performance with VMSC > 7.2, > - > But,  HOW MUCH    improvement can I expect?n > D > On our VAX VMS 6.X systems, when we put 3 or 4 thousand files in a > directory, we would haveN > all sorts of problems.   Now,  on hearing that VMS 7.2 fixed these problems, > I have heard aD > proposal to allow the application to put over  60,000 files in one$ > directory.   I think this is nuts,> > but I have not had the opportunity to test the improvements. >  > Can anyone comment on this?i > 8 > Thanks for all your help in the past (and the future).  L Sorry, but I gotta ask.  Just what is in these files?  Sounds to me like theN directory is being used in a manner better served by a file.  Is this just badL applicaation design?  I really cannot envision (though I'm sure someone willL enlighten me) any reason to have so many files in one directory.  Except forO systems supporting thousands of students, I have a hard time with a disk having$
 60,000 files.u   Dave, who's truly curious.   -- @4 David Froble                       Tel: 724-529-04504 Dave Froble Enterprises, Inc.      Fax: 724-529-0596> DFE Ultralights, Inc.              E-Mail: davef@tsoft-inc.com6 T-Soft, Inc.  170 Grimplin Road  Vanderbilt, PA  15486   ------------------------------  # Date: Thu, 29 Jun 2000 12:58:45 GMTw From: paul_hallam@my-deja.comnJ Subject: How do you Use LCKPAG on ALPHA to lock a full process in memory ?) Message-ID: <8jfh5r$6ac$1@nnrp1.deja.com>>  G After all the help on the VOLATILE/UNALIGNED problem I posted earlier I D must say thanks and I hope someone can answer this question as well.  G I have searched the ask-the-wizard section in decs pages but either I'mbF using the wrong search strings or this question is so simple it hasn't been asked of them before.   ANYWAY :  F In Pascal on a VMS V7.1 ALPHA system I need to lock a complete program
 in memory.D I know I need to use SYS$LCKPAG with the address ranges, called onceC for the data and once for the code BUT after reading the linker and 1 programming manuals I am still a little confused.'   A)1 In the linker options file do I need to include ao. "symbol_vector = (my_routine_name_1=procedure,.                   my_routine_name_2=procedure,/                   my_function_name_1=procedure,v                   etc ...M                   ) $ psect=$DATA$,$Literal,$local,$code$"  D and then ensure the program calls SYS$LCKPAG twice, firstly with theF address of one of the procedures named above and then with the address' of any variable used in the program ???s   B)D Do I need to list ALL the psect types, such as $ABS$, $BSS$, $LINK$,D $LITERALS$, etc in the 'psect=' line or am I being a little over the top with this.   Many thanks in advance   Paul    & Sent via Deja.com http://www.deja.com/ Before you buy.    ------------------------------  # Date: Thu, 29 Jun 2000 14:18:08 GMT = From: system@SendSpamHere.ORG (Brian Schenkenberger, VAXman-) N Subject: Re: How do you Use LCKPAG on ALPHA to lock a full process in memory ?0 Message-ID: <009EC53B.C8AEFB96@SendSpamHere.ORG>  I In article <8jfh5r$6ac$1@nnrp1.deja.com>, paul_hallam@my-deja.com writes:hH >After all the help on the VOLATILE/UNALIGNED problem I posted earlier IE >must say thanks and I hope someone can answer this question as well.R >1H >I have searched the ask-the-wizard section in decs pages but either I'mG >using the wrong search strings or this question is so simple it hasn'te >been asked of them before.n >b	 >ANYWAY :s >eG >In Pascal on a VMS V7.1 ALPHA system I need to lock a complete programs >in memory.iE >I know I need to use SYS$LCKPAG with the address ranges, called onceeD >for the data and once for the code BUT after reading the linker and2 >programming manuals I am still a little confused.  < Are you sure you need $LCKPAG?  ... or will $LKWSET suffice?   >A)f2 >In the linker options file do I need to include a/ >"symbol_vector = (my_routine_name_1=procedure,>/ >                  my_routine_name_2=procedure,h0 >                  my_function_name_1=procedure, >                  etc ... >                  )% >psect=$DATA$,$Literal,$local,$code$"l >tE >and then ensure the program calls SYS$LCKPAG twice, firstly with thetG >address of one of the procedures named above and then with the address1( >of any variable used in the program ??? >4 >B)dE >Do I need to list ALL the psect types, such as $ABS$, $BSS$, $LINK$, E >$LITERALS$, etc in the 'psect=' line or am I being a little over the  >top with this.r >  >Many thanks in advance  >  >Paul? >  >t' >Sent via Deja.com http://www.deja.com/. >Before you buy.  F All depends upon what (and why) you need to lock down any memory.  NotF knowing what your application requires, how it's sections fit togetherG and where (... the .MAP is help full here), and how and what data is top+ be referenced; this is difficult to answer.o  H Generally, $LCKPAG and/or $LKWSET are used to lock pages when code is toH execute at an elevated IPL where paging causes rather undesired behaviorH (ie. the system crashes).  On the Alpha, you typically need to lock downG not only the procedure's code but its linkage section as well.  Data toi; be reference at the elevated IPL must also be locked down. s  H If you need to lock down ALL of the code, linkage and data, one easy wayH is to create two separate psects (ie. $$$BEGIN and ___END) and place twoJ reference lables (ie. BEGIN_LOCK_ADDRESS and END_LOCK_ADDRESS respectivelyG in each of the two aforementioned psects) into these psects.  Then yoursH code will create a vector to pass to $LCKPAG/$LKWSET with the address ofH BEGIN_LOCK_ADDRESS and END_LOCK_ADDRESS.  Now all you need to do is pro-H perly link all of this together (an OPTION link deck can get you there).  : Still, I'd have to ask why you believe you need $LCKPAG.   --O VAXman- OpenVMS APE certification number: AAA-0001     VAXman(at)TMESIS(dot)COM    ------------------------------  # Date: Thu, 29 Jun 2000 04:39:13 GMTa2 From: "Zane H. Healy" <healyzh@shell1.aracnet.com>E Subject: Re: How to create a native VMS command - and a few questions 1 Message-ID: <RpA65.10$ma.881@typhoon.aracnet.com>o  : Larry Kilgallen <Kilgallen@eisner.decus.org.nospam> wrote:D > It is mostly the base VMS documentation which has been put online.E > Sometimes I think it is a mistake, as people go searching "the net"tC > and ignoring that CD-ROM full of documentation that came with the  > software.o  K Actually following the links from the following page will lead you to a lotF of the other documentation:o  * http://www.openvms.digital.com/commercial/  L Unfortunatly DECset is one of the products that doesn't have any doc's up atJ this time.  What would be really nice is if the normal documentation page,+ would include links to these other manuals.n  E One nice thing about having the documention on line is that if you'rerJ somewhere that doesn't have the doc CD's, or doesn't have them mounted you' can still access the data.  Very handy.y   			Zane    ------------------------------  % Date: Thu, 29 Jun 2000 10:30:42 +0200d From: "Wim" <wim@wanadoo.nl>- Subject: how to register a node in decnet osi * Message-ID: <8jf1g3$6v8$1@news1.xs4all.nl>   Hello,  K I'm pretty new with decnet osi phase V. I could always add a node with thisa commandoE mcr decnet_register and then I would get a menu what looked like thisC  I DECNET_REGISTER - Manage node registrations in network directory servicesV=   Use Return, Ctrl-N, and Ctrl-P to move between input fieldsi*   Use "?" to obtain help, Ctrl-Z to cancel  2   1 - Show information about registered node names#   2 - Register or modify node namesyD   3 - Update registered node towers using information from the nodes#   4 - Rename a registered node nameoD   5 - Repair the synonym and address links for registered node names   6 - Deregister node names   &   7 - Export node names to a data file(   8 - Import node names from a data file  (   9 - Set preferences and network values  "  10 - Manage the directory service  11 - Spawn to DCL  I But when I do exactly the same on an other node I get a prompt???    likeb this  DECNET_REGISTER>T  F How can I get a menu like above instead of the decnet_register prompt.= The machines are the same and the software is the same aswell-   Regards Wim    ------------------------------  % Date: Thu, 29 Jun 2000 09:41:48 +0000a- From: ezzaoudi med <m.ezzaoudi@digitem.co.ma>A1 Subject: Re: how to register a node in decnet osi8- Message-ID: <395B19DC.19A2E0F2@digitem.co.ma>n   HI9 Your device termninal is probably not defined or unknown. ) When I meet this probleme , I just type :n $SET TERM /DEV=3DVT100/ And I configure correctly my emulator as VT100.8 By   Wim a =E9crit :    > Hello, >tF > I'm pretty new with decnet osi phase V. I could always add a node w= ith this	 > command8F > mcr decnet_register and then I would get a menu what looked like th= is >yF > DECNET_REGISTER - Manage node registrations in network directory se= rvices? >   Use Return, Ctrl-N, and Ctrl-P to move between input fields , >   Use "?" to obtain help, Ctrl-Z to cancel >.4 >   1 - Show information about registered node names% >   2 - Register or modify node names F >   3 - Update registered node towers using information from the node= sn% >   4 - Rename a registered node name F >   5 - Repair the synonym and address links for registered node name= s- >   6 - Deregister node names  >0( >   7 - Export node names to a data file* >   8 - Import node names from a data file > * >   9 - Set preferences and network values >t$ >  10 - Manage the directory service >  11 - Spawn to DCL >VF > But when I do exactly the same on an other node I get a prompt???  =   like > this >  DECNET_REGISTER>o >lF > How can I get a menu like above instead of the decnet_register prom= pt.w? > The machines are the same and the software is the same aswell6 >t
 > Regards Wiml   ------------------------------  % Date: Thu, 29 Jun 2000 11:58:49 +0200n From: "Wim" <wim@wanadoo.nl>1 Subject: Re: how to register a node in decnet osi1* Message-ID: <8jf6l7$rho$1@news1.xs4all.nl>  : "ezzaoudi med" <m.ezzaoudi@digitem.co.ma> wrote in message' news:395B19DC.19A2E0F2@digitem.co.ma...t HI9 Your device termninal is probably not defined or unknown.l) When I meet this probleme , I just type :  $SET TERM /DEV=VT100   CUT>>t   Thanks that was the solution   ------------------------------  % Date: Thu, 29 Jun 2000 08:42:27 +0200f2 From: martin@radiogaga.harz.de (Martin Vorlaender). Subject: Re: LPR from VMS v6.2 to Linux Redhat; Message-ID: <395aefd3.524144494f47414741@radiogaga.harz.de>a   sboulay@my-deja.com wrote:H : We are desperately trying to print from VMS to a Linux box via LPR. We7 : send the job but all we get is a "Retained on error".e   That's not quite correct.0   : Here's the full message: :a& : BASE> show queue UCX$LPD_QUEUE /full ...:6 :   Entry  Jobname         Username     Blocks  Status6 :   -----  -------         --------     ------  ------A :      34  LOGIN           SBOULAY           3  Retained on errorn. :        %UCX-F-NOMSG, Message number 030AEB74   There you are:   $ @find_message %X30AEB74c Found in UCX$MSG- %UCX-F-LPD_LONGPRCAP, Printcap entry too long-  $ :  and here's our printcap.dat file: : #3 : # LOCAL PRINTERS : #c : UCX$LPD_QUEUE:\s :         :lp=UCX$LPD_QUEUE:\j :         :sd=UCX$LPD_SPOOL: : #a : LOCAL1|local1:\ 0 :         :lf=/SYS$SPECIFIC/UCX_LPD/LOCAL1.LOG:\ :         :lp=LOCAL1:\ :         :rm=10.0.0.98:\  :         :rp=lp0:\w+ :         :sd=/SYS$SPECIFIC/UCX_LPD/LOCAL1:e : #e  < My guess would be to remove the trailing ':' on every entry.   cu,    Martin  6 P.S.: The find_message tool used above is nothing but:  	   $ loop:Y'   $   f = F$SEARCH("SYS$MESSAGE:*.EXE")i&   $   IF f .EQS. "" THEN GOTO end_loop   $   SET MESSAGE 'f'.   $   m = F$MESSAGE('p1') <   $   IF F$LOCATE("NOMSG",m) .LT. F$LENGTH(m) THEN GOTO loop7   $   WRITE SYS$OUTPUT "Found in ", F$PARSE(f,,,"NAME")f   $   WRITE SYS$OUTPUT m   $   GOTO fin   $ endloop:(   $ WRITE SYS$OUTPUT "Message not found"   $ fin:&   $ SET MESSAGE SYS$MESSAGE:SYSMSG.EXE --J One OS to rule them all       | Martin Vorlaender  |  VMS & WNT programmer7 One OS to find them           | work: mv@pdv-systeme.denN One OS to bring them all      |       http://www.pdv-systeme.de/users/martinv/> And in the Darkness bind them.| home: martin@radiogaga.harz.de   ------------------------------  % Date: Thu, 29 Jun 2000 03:24:57 -0700 , From: nmike <mikeNOmiSPAM@marsat.ru.invalid>. Subject: Re: LPR from VMS v6.2 to Linux Redhat9 Message-ID: <34d7f918.0015d637@usw-ex0106-046.remarq.com>e  % I had this problem with printing fromM& VMS to BSDI box running LPRNG package.$ Red Hat Linux uses this package too,$ AFAIK. LPRNG is checking all inbound) message headers for potentially maliciousm& characters, which have special meaning* on UNIX. One of them is "$". You can force' LPRNG to accept these jobs by includingg+ ":fix_bad_job" parameter into /etc/printcap ) file on RedHat for each queue descriptionh+ you want to access from VMS. This way LPRNGo* will replace all "bad" symbols in job name+ by underscores and then happily process it.o Hope this will help.   With best regards,   Mikhail Nosov, Morsviazsputnik IT Manager.i E-Mail: mike@marsat.ru http://www.marsat.ru  - P.S. I tried to reply to your message from myt. mail client, but seems something went wrong...      ; -----------------------------------------------------------e  7 Got questions?  Get answers over the phone at Keen.com.o Up to 100 minutes free!l http://www.keen.comc   ------------------------------  # Date: Thu, 29 Jun 2000 12:47:40 GMT 4 From: LESLIE@209-16-45-102.insync.net (Jerry Leslie). Subject: Re: LPR from VMS v6.2 to Linux Redhat' Message-ID: <MzH65.892$5L3.4698@insync>@   sboulay@my-deja.com wrote: : Help!e :rH : We are desperately trying to print from VMS to a Linux box via LPR. WeG : send the job but all we get is a "Retained on error". Here's the fulll
 : message: :o& : BASE> show queue UCX$LPD_QUEUE /fullB : Server queue UCX$LPD_QUEUE, idle, on XYZ::, mounted form DEFAULTC :   /BASE_PRIORITY=4 /DEFAULT=(FEED,FORM=DEFAULT) /NOENABLE_GENERICdH :   /OWNER=[SYSTEM] /PROCESSOR=UCX$LPD_SMB /PROTECTION=(S:M,O:D,G:R,W:S) :   /RETAIN=ERRORp :l6 :   Entry  Jobname         Username     Blocks  Status6 :   -----  -------         --------     ------  ------A :      34  LOGIN           SBOULAY           3  Retained on errord. :        %UCX-F-NOMSG, Message number 030AEB74H :          Submitted 28-JUN-2000 18:53:08.77 /FORM=DEFAULT /PRIORITY=1009 :          File: _XYZ$DKA0:[VMS$COMMON.SYSMGR]LOGIN.COM;3rC :          Completed 28-JUN-2000 18:53:08.85 on queue UCX$LPD_QUEUEt : $ :  and here's our printcap.dat file: : #  : # LOCAL PRINTERS : #  : UCX$LPD_QUEUE:\i :         :lp=UCX$LPD_QUEUE:\T :         :sd=UCX$LPD_SPOOL: : #p : LOCAL1|local1:\x0 :         :lf=/SYS$SPECIFIC/UCX_LPD/LOCAL1.LOG:\ :         :lp=LOCAL1:\ :         :rm=10.0.0.98:\m :         :rp=lp0:\,+ :         :sd=/SYS$SPECIFIC/UCX_LPD/LOCAL1:  : #e :hH : What are we doing wrong? The Linux box at 10.0.0.98 is quite alive andI : pinging back. We also tried printing via lpr from Win2000 and it works.5 :B : Please respond! Thanks!w :e :    Sylvain    , Using LOOK_MSG.COM, written by Ferry Bolhr:      $ ucx show versiona  @     Digital TCP/IP Services for OpenVMS VAX Version V4.2 - ECO 30     on a VAXstation 4000-96 running OpenVMS V7.1      $ @look_msg 030AEB74g       Message found in UCX$MSG.EXE:5    030AEB74 -> %UCX-F-LPD_UNPRTR, Unknown printer !ASN  > I think UCX's LPR server wants a name, not an IP address; i.e. this line in PRINTCAP.DAT:             :rm=10.0.0.98:\    should contain a node name:o             :rm=linux_printer:\h  2 Try defining a name in the local host table; e.g.:  1    $ ucx set host linux_printer/address=10.0.0.98o  . and redefine the printer queue using the name.  / --Jerry Leslie   leslie@209-16-45-97.insync.net ;                  leslie@209-16-45-102.insync.net is invalido2                  (my opinions are strictly my own)   ------------------------------  % Date: Thu, 29 Jun 2000 01:26:02 -0400t' From: "Bill Todd" <billtodd@foo.mv.com> @ Subject: Re: Northern Light vs. Google (and the winner is . . .)( Message-ID: <8jemjh$dkk$1@pyrite.mv.net>  4 Main, Kerry <Kerry.Main@compaq.com> wrote in messageD news:910612C07BCAD1119AF40000F86AF0D8052844A4@kaoexc4.kao.dec.com...   ...a  H > In this mission critical environment, data replication is certainly noE > replacement for backups. A file created yesterday, but accidentally  deletedcF > or corrupted today typically must be restorable from the last nights backup.u >  >eK > Data replication in most mission critical applications is usually definedo asJ > an availability feature - not a data backup feature. It is why shadowing (orIK > mirroring) is recommended as an availability feature that complements thes" > data integrity backup processes. >yA > Yes, big storage arrays are pushing the backup windows and tapenL > technologies, but this does not negate the need for backups, it just meansH > that additional creative technologies need to be adopted to handle it.  G Which is exactly what I was describing:  the replication (at a separate D site) handles the issue of media failure (or site disaster) that hasK traditionally been one of the reasons for backups, and the snapshots handle,K the restoration of deleted or corrupted data, the other traditional role ofa backups.  B What I said was that *traditional backup*, not the requirements it! satisfies, may be on the way out.a  K As for your other comments about the areas that may be exceptions to normaliI backup requirements, if they indicate that you've reassessed your initialpE suggestion that Google's approach was inadvisable *for its particular  application*, then we agree.   - bill   >s
 > Regards, >e > Kerry Main > Senior Consultant, > Compaq Canadat > Professional Servicesr > Voice : 613-592-4660 > FAX   : 819-772-7036 > Email : kerry.main@compaq.comf >. >  >W > -----Original Message-----B > From: young_r@eisner.decus.org [mailto:young_r@eisner.decus.org]) > Sent: Wednesday, June 28, 2000 11:29 PMr > To: Info-VAX@Mvb.Saic.ComaB > Subject: Re: Northern Light vs. Google (and the winner is . . .) >  >u: > In article <395A9B97.1EFFFEE5@usfamily.net>, Keith Brown" > <kbrown780@usfamily.net> writes: >o > >t> > > I don't believe that replication is a good replacement forB > > backup. Suppose your data is deleted or corrupted, should this > > be replicated too? > >  > ; > There must be nice journalling techniques here if you are < > committed to it.  i.e. critical stuff is journalled.  Hard9 > to corrupt flat files.  Be very careful doing copies ori@ > synchronizations.  A program/interface that won't allow you to	 > destroyl > things would be convenient.s >B? > Besides, see the other thread.  How does one go about backingb? > up 20, 30, 80 Terabytes over 100 MBit pipes?  Very carefully,  > I suppose. >vA > If the data gets stomped, crawl the web from scratch and suffer-; > the consequences?  Maybe.  Or take a few weeks to restoremA > the 12 Terabytes?  Maybe you only lost or corrupted 5000 files?: >: > Crawl the web for them.L >  > Rob  >u   ------------------------------  % Date: Thu, 29 Jun 2000 01:43:31 -0400 ' From: "Bill Todd" <billtodd@foo.mv.com> @ Subject: Re: Northern Light vs. Google (and the winner is . . .)( Message-ID: <8jenkb$efg$1@pyrite.mv.net>  5 Rob Young <young_r@eisner.decus.org> wrote in messagei% news:ZOWe2atIF5jK@eisner.decus.org...aL > In article <8je5k5$r9p$1@pyrite.mv.net>, "Bill Todd" <billtodd@foo.mv.com> writes:t   ...s  J > > I'm not sure you understood what Nigel was saying.  If the servers areJ > > indeed mirroring each other (and index modifications aren't constant), thenL > > there will certainly be plenty of times when one server of a mirror pair canrL > > easily handle the load and the other can spin down its disks and put its CPUaB > > into stand-by mode without noticeably affecting response time. > >r >l >L= > Oh... seems that isn't the way it works though... the indexo: > is split amongst chunks of 40 servers and load balancers@ > fire to an idle server.  I may be misinterpreting this though: > / > http://www.internetwk.com/lead/lead060100.htm   H Hard to say exactly what they're doing within that 'cluster'.  The otherJ article seemed to suggest server-level replication, which I interpreted asK Novell SFT-style duplexing, but there could be considerably more redundancy 
 than that.  L 4000 servers in 3 (totally replicated, by my reading) locations means aroundH 1300 in each location (maybe a bit less, given that I suspect they use aL bunch for front-end operations).  But even 1300, at 80 GB each, works out toH around 100 TB of storage - or 50 TB mirrored (or at the very least 30 TBK mirrored, given that the other article indicated that the older servers hadnK 22 GB rather than 40 GB drives).  The article seems to state that Google is.I indexing about 1 TB of data per month, but that leaves enough leeway that;E the data might or might not be being multiply-mirrored - but it seems J unlikely that all 40 servers in a cluster replicate each other fully (thatI would use up the entire capacity in a single month, which would only makeyK sense if they start from scratch each month so as to retire defunct links).    >oF > "Within the data centers, Google uses its own traffic management andL > load-balancing software to direct traffic to the best server. The index of thewK > Web is broken down into parts, with each section of the index distributed  to a; > cluster of about 40 servers for redundancy and failover."c >e= > If they have many more servers than they need, that is very.= > wasteful.  So the 4000 must come into play (certainly isn'tS> > for the storage needs).  Maybe they have 2 extra servers per= > nodule?  In other words, those 40 per grouping are probablye0 > tickled periodically during peak search times. >I. > Perhaps you can give us your interpretation? > 8 > I'm not a network guy, but maybe that 40 server number8 > wasn't just arbitrary.  How many ports in a large hub? > @ > So far, that is the best overview of what they have that I canD > find.  The actual thesis itself was particularly troubling to find( > and I have already lost the link :-( . >5 >$L > > Unless host-mirrored large servers are also able to monitor activity andE > > spin down unneeded disks in a similar manner, this could make thep low-end.I > > server farm *more* energy efficient than the centralized alternative.  > >g >i= > Don't know about that.  Each server has 2 disks... year endt< > they have 2 * 10000 disks or 20000 disks.  I want to say a@ > large search service disk farm has a 1000 18 gig drives, maybe > more but how much more?e >a >  > >   Couple that with the@ > >> fact they have to spin up the disk drives and you see whereC > >> their edge of a very fast search suddenly goes in the oppositenA > >> direction and they become one of the slower search services.  > >o > > ...  > >nE > >> But what if Google is only 5% efficient?  Sure, they are gettingn? > >> the job done and done well.. but they are wasting valuabley@ > >> resources and only a matter of time before the tree huggers > >> catch on. > >dJ > > I'm not sure what you're suggesting.  Since Google is using reasonably largesL > > disk drives, their disk power consumption seems likely to about the same asK > > any alternative.  Since they're likely using less than state-of-the-art I > > close-to-1-GHz CPUs, that portion of the load shouldn't be *too* mucha worse.- > > So where is the comparative inefficiency?  > >e >n: > From an earlier post... if they are now up to 18 millionD > searches a day and within an 8 hour period they perform 12 millionE > of them, they are doing 417 searches per second.  Since each searchr@ > appears to take .5 second (actually less), those 4000 CPUs are5 > mostly idle.  Even less than 5% loaded I would say.d  L Didn't realize you were singling out CPU loading.  But that just allows themI to use slow (cheap) processors that consume only a few watts apiece, thus < contributing only a small amount to the overall power drain.   >cC > Each server has 80 Gigs of storage from the above article... from ? > another article (in May I believe) they talked about using 80TE > Terabytes of storage out of the 320 Terabytes in the servers.  OverpG > 20% in use there.  But now they are up to 500 million articles, maybeaA > 12 Terabytes for actual data.  I'll have to dig to see if I can0D > estimate that or if that info is readily available.  If 12 or evenD > 20 .. 20 Terabytes means they are actually using 6% of the storageC > available for data with over 1 Terabyte for index (from the above H > article).  The rest replicated hither and yon.  But storage efficiencyC > is one of those things hard to pin down.. wasn't really moaning a E > whole lot about that.. but using less than 5% of the CPU horsepoweraD > available does seem to be quite a waste of resources.  But they doJ > point out their index phase is very CPU intensive.  Whatever that means. >GG > Let's talk next about the 20000 fans they will have spinning year-endsF > (2 per server, right?).  That seems like an interesting topic.  ThenH > off to the waste heat a power supply uses -- 10000 of those, year-end.  H Power supplies and fans are a fact of life in any installation, and theyB scale at least somewhat proportionally to the power being activelyK dissipated doing useful work.  So if the disks and processors don't consumerJ noticeably more power than a more centralized approach, neither should the supplies and fans.   >i > Rob  >    ------------------------------  % Date: Thu, 29 Jun 2000 06:38:20 -0400 + From: Tim Shoppa <shoppa@trailing-edge.com>i@ Subject: Re: Northern Light vs. Google (and the winner is . . .)1 Message-ID: <395AEEDC.4D68D52A@trailing-edge.com>    Bill Todd wrote:F > Indeed, for this large and volatile an environment, backup is prettyM > unattractive.  For that matter, it's convincingly arguable that traditionaltG > backup is in general on the way out, given the decreasing gap in cost0J > between on-line and off-line storage and the increasing relative cost of? > creating, organizing, and storing said off-line storage [...]n  > I agree with Keith: This argument is total crap.  Just because> disk drives are so ridiculously cheap these days does not mean9 that the data on them is worth any less.  And the cost ofpA off-line tape storage is *not* increasing - it's getting cheaper,t? just not as quickly as low-end disk drives are getting cheaper.h  E I will agree that many modern platforms lack effective back-up tools,rE but that's a fault of the platform's implementers.  They've basically3> said "backup is hard so we won't do it", and they got you hook+ line and sinker, Bill.  They don't have me.p   Tim.   ------------------------------  % Date: Thu, 29 Jun 2000 15:02:00 +0200n= From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>l@ Subject: Re: Northern Light vs. Google (and the winner is . . .)) Message-ID: <395B48C8.72E4B17A@gtech.com>    Rob Young wrote:F >         Even though David J. thinks Bill needs a vacation, I believeE >         he is spot on here.  As Bill points out, this is replicatedn4 >         hither and yon, backups are not necessary.   ????  @ Replication, shadowing etc. of data is absolutey not an argument for dropping backup !n  C Such features does cover most hardware failures, but does not cover C all software faiures. You are not better off having 2 copies of theA wrong data than just 1 ! :-)  5 And in many cases access to history may be important.    Arne   ------------------------------  % Date: Thu, 29 Jun 2000 11:11:05 -0400P+ From: "Main, Kerry" <Kerry.Main@compaq.com>v@ Subject: RE: Northern Light vs. Google (and the winner is . . .)J Message-ID: <910612C07BCAD1119AF40000F86AF0D8052844AE@kaoexc4.kao.dec.com>   Bill,E  H >>> As for your other comments about the areas that may be exceptions toH normal backup requirements, if they indicate that you've reassessed yourB initial suggestion that Google's approach was inadvisable *for its+ particular application*, then we agree. >>>O  K Well, all I am saying is that the technical massively distributed computing K environment requires much more software customization and tuning and higherrL numbers of people to support it. Perhaps no backups are fine for them now...  F However - One of the challenges Google will face in the future is whenJ Customers start asking for further customizations ie. "I want to establishH my own profile, so this particular query will run everyday at this time,1 with the results sent as an email update to me.."i  J This is one of the real bennies NorthernLight already offers. Online stockI folks can get automatic daily information, breaking information on stocks 
 they own..  B This type of user specific profile information does need backups.   L Anyway, at a higher level, this is a religious computing strategy debate ie.H would 50 low end computers with 2Gb of memory, 50 system disks be betterI suited to address a problem vs. 1 (or 2 for availability) high end systemiK with 100Gb of memory with one system disk pair (shadowed) and whatever userc storage was required?   D Unless all of the associated costs are included ie. people (how manyK developers and operations staff to write, maintain and support this heavilyiC customized environment?), processes and technology, then a good and/& reasonable comparison is not possible.  H If indeed, all of these costs were considered in the Google environment,( then perhaps it was the right approach.   G My issue is that many of those in the massively distributed environmentnL often (not always) only look at the hardware and "free" OS costs and justify* their whole project on these costs only...   Regards,  
 Kerry Main Senior Consultant,
 Compaq Canada  Professional ServicesM Voice : 613-592-4660 FAX   : 819-772-7036 Email : kerry.main@compaq.com@       -----Original Message-----, From: Bill Todd [mailto:billtodd@foo.mv.com]% Sent: Thursday, June 29, 2000 1:26 AMj To: Info-VAX@Mvb.Saic.Coma@ Subject: Re: Northern Light vs. Google (and the winner is . . .)      4 Main, Kerry <Kerry.Main@compaq.com> wrote in messageD news:910612C07BCAD1119AF40000F86AF0D8052844A4@kaoexc4.kao.dec.com...   ...U  H > In this mission critical environment, data replication is certainly noE > replacement for backups. A file created yesterday, but accidentally  deleted F > or corrupted today typically must be restorable from the last nights backup.  >  >GK > Data replication in most mission critical applications is usually definedM asJ > an availability feature - not a data backup feature. It is why shadowing (orRK > mirroring) is recommended as an availability feature that complements the " > data integrity backup processes. >oA > Yes, big storage arrays are pushing the backup windows and tape L > technologies, but this does not negate the need for backups, it just meansH > that additional creative technologies need to be adopted to handle it.  G Which is exactly what I was describing:  the replication (at a separate0D site) handles the issue of media failure (or site disaster) that hasK traditionally been one of the reasons for backups, and the snapshots handleqK the restoration of deleted or corrupted data, the other traditional role ofn backups.  B What I said was that *traditional backup*, not the requirements it! satisfies, may be on the way out.   K As for your other comments about the areas that may be exceptions to normal4I backup requirements, if they indicate that you've reassessed your initialmE suggestion that Google's approach was inadvisable *for its particular- application*, then we agree.   - bill   >C
 > Regards, >w > Kerry Main > Senior Consultant, > Compaq CanadaI > Professional Servicesr > Voice : 613-592-4660 > FAX   : 819-772-7036 > Email : kerry.main@compaq.com  >r >e >i > -----Original Message-----B > From: young_r@eisner.decus.org [mailto:young_r@eisner.decus.org]) > Sent: Wednesday, June 28, 2000 11:29 PMm > To: Info-VAX@Mvb.Saic.Com@B > Subject: Re: Northern Light vs. Google (and the winner is . . .) >n >i: > In article <395A9B97.1EFFFEE5@usfamily.net>, Keith Brown" > <kbrown780@usfamily.net> writes: >  > > > > > I don't believe that replication is a good replacement forB > > backup. Suppose your data is deleted or corrupted, should this > > be replicated too? > >$ >$; > There must be nice journalling techniques here if you are < > committed to it.  i.e. critical stuff is journalled.  Hard9 > to corrupt flat files.  Be very careful doing copies orv@ > synchronizations.  A program/interface that won't allow you to	 > destroy  > things would be convenient.a > ? > Besides, see the other thread.  How does one go about backingm? > up 20, 30, 80 Terabytes over 100 MBit pipes?  Very carefully,s > I suppose. >lA > If the data gets stomped, crawl the web from scratch and suffer ; > the consequences?  Maybe.  Or take a few weeks to restoreiA > the 12 Terabytes?  Maybe you only lost or corrupted 5000 files?a >i > Crawl the web for them.  >  > Robb >s   ------------------------------  % Date: Thu, 29 Jun 2000 12:18:05 -0400 ' From: "Bill Todd" <billtodd@foo.mv.com>,@ Subject: Re: Northern Light vs. Google (and the winner is . . .)( Message-ID: <8jfsq2$rm3$1@pyrite.mv.net>  5 Keith Brown <kbrown780@usfamily.net> wrote in message & news:395AB748.1C62A193@usfamily.net...   ...t  H > > Indeed, for this large and volatile an environment, backup is prettyC > > unattractive.  For that matter, it's convincingly arguable that  traditionaldI > > backup is in general on the way out, given the decreasing gap in costtL > > between on-line and off-line storage and the increasing relative cost ofE > > creating, organizing, and storing said off-line storage:  keepingiJ > > differential (to minimize space requirements; perhaps partial as well, sincegF > > a lot of data is easily recoverable by other means) redundant (andC > > separated, to guard against both media and whole-site failures)u snapshots onJ > > line is one promising alternative, and particularly attractive in that itK > > can allow users to recover from their own mistakes without invoking the- Gods > > of the Data Center.s > >[
 > > - bill > >s > > >u > > > -- > > > Keith Brownn > > > kbrown780@usfamily.net >V? > I agree with you that snapshots will probably be way to go ashB > the advantages are readily apparent. Your statement that backups  > will soon be obsolete is crap.  H Rather, your ability to read is what appears to be 'crap':  my statementI (which you included above, but apparently didn't bother to read) was thatsL "it's convincingly arguable that traditional backup is in general on the wayK out", a view which I happen to share with such diverse people as the CEO ofnI EMC (a company rather more interested in the future of storage than most) L and Linus Torvalds - both of whose industry credibility likely exceeds yours5 (though I wouldn't consider either to be infallible).h   - bill    This is the usual spin vendors = > put out when they don't have the functionality so lets tell & > everyone they don't need it anymore. >d > --
 > Keith Browns > kbrown780@usfamily.net   ------------------------------  % Date: Thu, 29 Jun 2000 12:36:18 -0400 ' From: "Bill Todd" <billtodd@foo.mv.com>l@ Subject: Re: Northern Light vs. Google (and the winner is . . .)( Message-ID: <8jfts9$smn$1@pyrite.mv.net>  6 Tim Shoppa <shoppa@trailing-edge.com> wrote in message+ news:395AEEDC.4D68D52A@trailing-edge.com...( > Bill Todd wrote:H > > Indeed, for this large and volatile an environment, backup is prettyC > > unattractive.  For that matter, it's convincingly arguable thata traditionaleI > > backup is in general on the way out, given the decreasing gap in cost L > > between on-line and off-line storage and the increasing relative cost ofA > > creating, organizing, and storing said off-line storage [...]e > @ > I agree with Keith: This argument is total crap.  Just because@ > disk drives are so ridiculously cheap these days does not mean; > that the data on them is worth any less.  And the cost ofoC > off-line tape storage is *not* increasing - it's getting cheaper, A > just not as quickly as low-end disk drives are getting cheaper.   L I'm beginning to believe that a noticeable percentage of comp.os.vms readers are functionally illiterate.  G When one states that the *gap* between two items is decreasing, it saysdK *nothing* about the absolute direction either one is taking:  the statement F you make above is *exactly* the same as the statement I made about the0 relative media cost of on- and off-line storage.  G And surely you're not suggesting that the cost of *managing* backups is,I decreasing at as great a rate as the media cost:  that would be in markedlF contrast to the generally (and rapidly) increasing percentage of totalK expenditure devoted to managing storage (rather than simply purchasing it).T  I And finally, the alternative to traditional backup I suggested, combininge6 replication and snapshot technology to create multipleF geographically-distributed copies of both current and historical data,L protects the disk data in the same ways traditional backup does (at least ifL you trust your file system and disk drivers not to suffer Byzantine failuresH that go out of their way to corrupt areas on multiple disks unrelated toJ current operations:  I suspect most VMS users do have such trust, but onceF you have a snapshot, the truly paranoid can still arrange for it to be7 copied elsewhere at leisure if they feel it necessary).e   >aG > I will agree that many modern platforms lack effective back-up tools,nG > but that's a fault of the platform's implementers.  They've basicallyn@ > said "backup is hard so we won't do it", and they got you hook- > line and sinker, Bill.  They don't have me.e  H And wouldn't want to, if all your days are as bad as this one.  Get some8 sleep and take a fresh look at this discussion tomorrow.   - bill   >e > Tim.   ------------------------------  % Date: Thu, 29 Jun 2000 12:38:41 -0400 ' From: "Bill Todd" <billtodd@foo.mv.com>i@ Subject: Re: Northern Light vs. Google (and the winner is . . .)( Message-ID: <8jfu0m$sms$1@pyrite.mv.net>  L I'm getting tired of responding to people who haven't taken the time to readI and understand what's already been written on this subject.  So I'll justt suggest that you do so.t   - bill  5 Arne Vajhj <arne.vajhoej@gtech.com> wrote in message.# news:395B48C8.72E4B17A@gtech.com...r > Rob Young wrote:H > >         Even though David J. thinks Bill needs a vacation, I believeG > >         he is spot on here.  As Bill points out, this is replicatedi6 > >         hither and yon, backups are not necessary. >n > ???? >aB > Replication, shadowing etc. of data is absolutey not an argument > for dropping backup !  >sE > Such features does cover most hardware failures, but does not cover E > all software faiures. You are not better off having 2 copies of thec > wrong data than just 1 ! :-) >v7 > And in many cases access to history may be important.c >u > Arne   ------------------------------  % Date: Thu, 29 Jun 2000 12:53:46 -0400c' From: "Bill Todd" <billtodd@foo.mv.com>B@ Subject: Re: Northern Light vs. Google (and the winner is . . .)' Message-ID: <8jfut4$ii$1@pyrite.mv.net>y  4 Main, Kerry <Kerry.Main@compaq.com> wrote in messageD news:910612C07BCAD1119AF40000F86AF0D8052844AE@kaoexc4.kao.dec.com... > Bill,o >eJ > >>> As for your other comments about the areas that may be exceptions toJ > normal backup requirements, if they indicate that you've reassessed yourD > initial suggestion that Google's approach was inadvisable *for its- > particular application*, then we agree. >>>u >'C > Well, all I am saying is that the technical massively distributedy	 computingnF > environment requires much more software customization and tuning and higherG > numbers of people to support it. Perhaps no backups are fine for themo now... >rH > However - One of the challenges Google will face in the future is whenL > Customers start asking for further customizations ie. "I want to establishJ > my own profile, so this particular query will run everyday at this time,3 > with the results sent as an email update to me.."m  L And such facilites shouldn't be difficult to add as new front-end processingJ that uses the back-end configuration already in place:  the new front endsJ will indeed require more in the way of backup, but that doesn't affect the
 back ends.   > L > This is one of the real bennies NorthernLight already offers. Online stockK > folks can get automatic daily information, breaking information on stockst > they own.. >aC > This type of user specific profile information does need backups.y >aJ > Anyway, at a higher level, this is a religious computing strategy debate ie.tJ > would 50 low end computers with 2Gb of memory, 50 system disks be betterK > suited to address a problem vs. 1 (or 2 for availability) high end systemyH > with 100Gb of memory with one system disk pair (shadowed) and whatever user > storage was required?g >sF > Unless all of the associated costs are included ie. people (how manyE > developers and operations staff to write, maintain and support thisP heavilyvE > customized environment?), processes and technology, then a good ande( > reasonable comparison is not possible. >eJ > If indeed, all of these costs were considered in the Google environment,) > then perhaps it was the right approach., >aI > My issue is that many of those in the massively distributed environmentnF > often (not always) only look at the hardware and "free" OS costs and justifyp, > their whole project on these costs only...  I An entirely legitmate observation.  However, a start-up can also face theaH reality that taking a short-term view may be the only viable option:  if? they just don't have the $1 million or so up front to pay for arI configuration that will likely turn out to be optimal well down the road, B then an incremental approach that *can* grow, even if not quite as* optimally, is the right approach for them.  J As long as mainframe-level solutions are low-volume, high-margin products,G there will always be situations where high-volume, low-margin commoditykH distributed systems are enough less expensive to pay for rather a lot ofF fussing around.  And they give the customer greater flexibility in theL future than when they're locked into a proprietary mainframe-style solution. But turn-key they're not.t   - bill   >b
 > Regards, >i > Kerry Main > Senior Consultant, > Compaq Canadan > Professional Servicess > Voice : 613-592-4660 > FAX   : 819-772-7036 > Email : kerry.main@compaq.comt >g >i >c > -----Original Message-----. > From: Bill Todd [mailto:billtodd@foo.mv.com]' > Sent: Thursday, June 29, 2000 1:26 AMa > To: Info-VAX@Mvb.Saic.Com B > Subject: Re: Northern Light vs. Google (and the winner is . . .) >  >d >a6 > Main, Kerry <Kerry.Main@compaq.com> wrote in messageF > news:910612C07BCAD1119AF40000F86AF0D8052844A4@kaoexc4.kao.dec.com... >- > ...- > J > > In this mission critical environment, data replication is certainly noG > > replacement for backups. A file created yesterday, but accidentally 	 > deletediH > > or corrupted today typically must be restorable from the last nights	 > backup.v > >o > >aE > > Data replication in most mission critical applications is usually  defined  > asL > > an availability feature - not a data backup feature. It is why shadowing > (or I > > mirroring) is recommended as an availability feature that complements  ther$ > > data integrity backup processes. > >rC > > Yes, big storage arrays are pushing the backup windows and tapeoH > > technologies, but this does not negate the need for backups, it just means J > > that additional creative technologies need to be adopted to handle it. >-I > Which is exactly what I was describing:  the replication (at a separate F > site) handles the issue of media failure (or site disaster) that hasF > traditionally been one of the reasons for backups, and the snapshots handleJ > the restoration of deleted or corrupted data, the other traditional role of
 > backups. >nD > What I said was that *traditional backup*, not the requirements it# > satisfies, may be on the way out.g >sF > As for your other comments about the areas that may be exceptions to normalK > backup requirements, if they indicate that you've reassessed your initialnG > suggestion that Google's approach was inadvisable *for its particularh > application*, then we agree. >t > - bill >n > >k > > Regards, > >e > > Kerry Main > > Senior Consultant, > > Compaq Canadae > > Professional Servicest > > Voice : 613-592-4660 > > FAX   : 819-772-7036! > > Email : kerry.main@compaq.com  > >i > >t > >r > > -----Original Message-----D > > From: young_r@eisner.decus.org [mailto:young_r@eisner.decus.org]+ > > Sent: Wednesday, June 28, 2000 11:29 PM  > > To: Info-VAX@Mvb.Saic.Com D > > Subject: Re: Northern Light vs. Google (and the winner is . . .) > >. > >h< > > In article <395A9B97.1EFFFEE5@usfamily.net>, Keith Brown$ > > <kbrown780@usfamily.net> writes: > >r > > >s@ > > > I don't believe that replication is a good replacement forD > > > backup. Suppose your data is deleted or corrupted, should this > > > be replicated too? > > >e > >s= > > There must be nice journalling techniques here if you are > > > committed to it.  i.e. critical stuff is journalled.  Hard; > > to corrupt flat files.  Be very careful doing copies orsB > > synchronizations.  A program/interface that won't allow you to > > destroyv > > things would be convenient.o > >hA > > Besides, see the other thread.  How does one go about backing,A > > up 20, 30, 80 Terabytes over 100 MBit pipes?  Very carefully,n > > I suppose. > >iC > > If the data gets stomped, crawl the web from scratch and suffert= > > the consequences?  Maybe.  Or take a few weeks to restoreaC > > the 12 Terabytes?  Maybe you only lost or corrupted 5000 files?s > >. > > Crawl the web for them.  > >r > > Robn > >o >q >n   ------------------------------  % Date: Thu, 29 Jun 2000 13:06:23 -0400m+ From: Tim Shoppa <shoppa@trailing-edge.com>i@ Subject: Re: Northern Light vs. Google (and the winner is . . .)1 Message-ID: <395B49CF.4E882ACD@trailing-edge.com>M   Bill Todd wrote:I > And surely you're not suggesting that the cost of *managing* backups isrK > decreasing at as great a rate as the media cost:  that would be in markedFH > contrast to the generally (and rapidly) increasing percentage of totalM > expenditure devoted to managing storage (rather than simply purchasing it).   F I agree with you here - the cost of managing/creating/maintaining dataA is the dominant component in any computer system these days.  TheaB offline and online storage (hardware and media) costs for the data are in the noise.l  A Still, the reason why it's so (comparatively) difficult to manageRB backups is that many modern platforms come with no integrated tool? for doing this, it's something that managers unfortunately haveo> to "bolt on".  To say that it's difficult to manage backups onE your system-critical Windows NT server or Linux server is to say thathH you've chosen the wrong OS.  As in, "whad'ya mean, there's no way for meD to mark that big scratch file /NOBACKUP"?  What *should* be a simple? job gets turned into a confusing array of fixed-size partitionst@ (themselves a waste of storage space, not so important as it was? ten years ago) and complicated scripting to exclude a couple ofh named files or directories.    Tim.   ------------------------------  # Date: Thu, 29 Jun 2000 17:35:27 GMTa- From: tsm@palindrome.org (tsm@palindrome.org) @ Subject: Re: Northern Light vs. Google (and the winner is . . .)3 Message-ID: <slrn8ln296.91p.tsm@user2.teleport.com>s  / On Thu, 29 Jun 2000 13:06:23 -0400, Tim Shoppa o! <shoppa@trailing-edge.com> wrote:u  E >To say that it's difficult to manage backups on your system-criticalrI >Windows NT server or Linux server is to say that you've chosen the wrong J >OS.  As in, "whad'ya mean, there's no way for me to mark that big scratchC >file /NOBACKUP"?  What *should* be a simple job gets turned into adH >confusing array of fixed-size partitions (themselves a waste of storageA >space, not so important as it was ten years ago) and complicated = >scripting to exclude a couple of named files or directories.e  G Actually, Windows NT does have the concept of /NOBACKUP just as much aswF VMS does (see the "backup" field in any file properties tab). Linux isF only loser in this case. It's funny that the only OS which _relies_ onD backups in order to make do for it's utterly horrible user interface@ (e.g. the "rm *>o" typo) is also the only one which is literallyD unbackupable (at least without shutting down into single user mode).   Regards,  
 Terry Murphy z   ------------------------------  % Date: Thu, 29 Jun 2000 19:22:33 +0200w= From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com> @ Subject: Re: Northern Light vs. Google (and the winner is . . .)) Message-ID: <395B85D9.51E45AF5@gtech.com>l   Bill Todd wrote:N > I'm getting tired of responding to people who haven't taken the time to readK > and understand what's already been written on this subject.  So I'll justo > suggest that you do so.   : And I think you should read what I wrote. I am saying that; replication and backup is two completely independent issues & for anyone who cares about their data.  9 That is a general statement and completely independent of  the Google Linux story.@   Arne   ------------------------------  % Date: Thu, 29 Jun 2000 12:48:47 +0100B/ From: Nigel Arnot <sysmgr@maxwell.ph.kcl.ac.uk>d: Subject: re: Northern Light vs. Google (power consumption)7 Message-ID: <009EC550.D4520326.22@maxwell.ph.kcl.ac.uk>    >  > >>       power consumptionE > >> 	(Environmentally very harmful.  If nothing else , could rile up @ > >> 	a few tree huggers and point out how many resources Google@ > >> 	is wasting... ironically this could be one of the greatest? > >> 	criticisms).  Not to be overlooked is the whole manner of D > >> 	efficiency which points a finger at the power consumption too. > > L > > This is perhaps true with current technology, but peering a little into ( > > the future this argument may invert. > > I > > Imagine that the next-generation server-farm box is using a Transmeta L > > Crusoe processor or similar. That means it dynamically adjusts its clockH > > speed to match the workload and minimize the power drain. The power I > > consumption of the server farm will therefore tend to scale with the hK > > level of activity. This gets even better if spin-down of inactive disk uF > > drives can also be incorporated. If they have multiple nodes with H > > identical disk data mirrored on each for redundancy and peak loads, L > > then at quiet times letting one out of N take all the strain might work O > > well. Or one of the non-rotating storage technologies may replace the disk i2 > > drive making the idle power become negligible. > >  > ? > 	But the difficulty here is that none of them are idle enougha; > 	to even spin down the disk drives.  Couple that with thel> > 	fact they have to spin up the disk drives and you see whereA > 	their edge of a very fast search suddenly goes in the oppositeu? > 	direction and they become one of the slower search services.r  M Even if you can't idle the disk drives, a processor with dynamically variablenL clock speed like the Crusoe will tune its speed down to the point where idleC time is small (on a millisecond timescale). It therefore won't burnhH power processing an idle loop, but will speed itself up to maximum clockE rate as soon as there'e enough work to keep it busy (in the case of a3< server, every time a request comes in!). At present this is M notebook and low-power appliance technology, but if massively parallel servertF farms become the norm it would make sense there too. I'm guessing, butE I suspect building it into a tightly-coupled set of processors like anA Galaxy would be much harder, giving a massively parallel array of 4 loosely coupled boxes a power-consumption advantage.  K As for the disk drives: it would all depend on how things were configured. lL Clearly, spin-up on demand wouldn't be any use, but if requests are vectoredD to nodes in a mirrored set by a control processor, then that controlG processor could adjust the number of active nodes in a set, spinning up C another one when needed slightly in advance of sending it any work.'I Users wouldn't see any delay, because it would never vector a transaction F to a processor that wasn't ready. And of course, this trick works onlyO with read-mostly data where updates (requiring all mirrored nodes to be active)n are batched daily or hourly. o   >  > > J > > Anyway, 10000 standard PCs is what percentage of the world's installedM > > PCs? Most of which are sitting in offices running Windoze, doing not muchtD > > very inefficiently, when they are doing anything at all that is,N > > because the operating system of choice only allows one user per processor.G > > Any tree-huggers out there, Bill Gates is the enemy! Or perhaps the C > > writers of "screen savers" which today cause monitors to remaintL > > active (>200W) continuously, rather than going into blank-screen (<10W) I > > standby mode. (I was surprised when I discovered how much power a 17"r> > > monitor takes, compared to the PC it's sitting on top of). > >  > D > 	Since they aren't in standby mode and mostly active, the question; > 	then is how much power are they consuming?  Yesterday's r< > 	New York Times pointed out (speculated) that maybe Celera? > 	is in such a hurry because they are burning $100,000 a months: > 	in electric costs.  Tongue in cheek speculation to make > 	for a nice article. > C > 	But what if Google is only 5% efficient?  Sure, they are gettingo= > 	the job done and done well.. but they are wasting valuable > > 	resources and only a matter of time before the tree huggers > 	catch on.  G The point I was making is that 5% is far higher than the figure for the K majority of office PCs. What percentage of the time is the average office'soJ MS-workstation Pentium-600 doing anything other than executing the WindozeE idle loop or the screen-saver? A very small percentage. If it weren't:I for the Microsoft 1 PC = 1 User approach (and the small MTBF of MicrosoftrO operating systems even with that simplification) the average office would have .J one processor box shared by several low-power dumb display/keyboard/mouse K stations. (I'm thinking of multiple video boards in one processor box with l? video distribution over UTP, rather than X-terms or Winframes).a  C So the tree-huggers should start work on the 99% of PCs that aren'trM servers of any description (100M of them?) and the *Gigawatts* of power that tA they are consuming. Only after something has been done about this,E prodigous waste does it make any sense to look at a small minority ofs# relatively productive server farms.s  E The single best thing that could happen is for flat-panel displays to F become price-competitive with conventional monitors*. In the meantime,I everyone should be encouraged to get rid of any screensavers that preventhJ the monitor going into standby, and all PCs should ship with the video-offF after 10 minutes BIOS option enabled. This could save *Gigawatts*. TheG next best thing would be universal implementation of dynamic adjustmentcL of processor clock rate -- but that's an order of magnitude down on what youJ save by eliminating vacuum-tube monitors or getting them into standby mode whenever possible.  G *Hum. Just realised that the cost of the electricity that the VDU uses  H should be included in the comparison. The average buyer doesn't know howE to do that. There should be a mandatory energy-efficiency sticker on sG VDUs and other IT gear, like there is on fridges and washing machines! r  ? Another thought. If you locate a megawatt server farm in a big lL building in some northern place where that building would have to be heated I for most of the year, then for 9+ months of the year it's not wasting any B energy at all ... just keeping people warm, which you'd have to doK anyway. So to placate the tree-huggers, just relocate to Alaska or Finland.u  Where is NorthernLights, anyway?   	Yours,h
 		Nigel Arnotn- 		NRA@MAXWELL.PH.KCL.AC.UK                   o  7 		"In the beginning there was nothing, which exploded."i   ------------------------------  # Date: Thu, 29 Jun 2000 15:50:15 GMTo- From: tsm@palindrome.org (tsm@palindrome.org)c: Subject: Re: Northern Light vs. Google (power consumption)3 Message-ID: <slrn8lms45.nth.tsm@user2.teleport.com>h  0 On Thu, 29 Jun 2000 12:48:47 +0100, Nigel Arnot $ <sysmgr@maxwell.ph.kcl.ac.uk> wrote:  N >Even if you can't idle the disk drives, a processor with dynamically variableM >clock speed like the Crusoe will tune its speed down to the point where idlefD >time is small (on a millisecond timescale). It therefore won't burnI >power processing an idle loop, but will speed itself up to maximum clockoF >rate as soon as there'e enough work to keep it busy (in the case of a* >server, every time a request comes in!).   H FYI ... Intel processors (including desktop/server parts) do indeed haveH low power features, which are used. They have an "HLT" instruction whichA is executed during the idle loop of the processor, which puts the B processor to sleep until the next interrupt comes in, and consumesJ virtually no power. This instruction is used in the idle loop of Linux. SoE when the Google machines are idle 95% of the time, they are consumingoG very little power. Seriously, even if the 5% of the time they are doingiF something, the task is almost entirely I/O bound and they are still in powerdown mode.u  J The power of the CPU's in Google is almost negligible. The real problem isH the disks, but I don't see how this could be lessened by using huge disk* arrays instead of a bunch of little disks.   Regards,   Terry Murphy   ------------------------------    Date: 29 Jun 2000 13:30:13 -0500* From: young_r@eisner.decus.org (Rob Young): Subject: Re: Northern Light vs. Google (power consumption)+ Message-ID: <t+e3UPlJWuNO@eisner.decus.org>n  c In article <slrn8lms45.nth.tsm@user2.teleport.com>, tsm@palindrome.org (tsm@palindrome.org) writes:e2 > On Thu, 29 Jun 2000 12:48:47 +0100, Nigel Arnot    > L > The power of the CPU's in Google is almost negligible. The real problem isJ > the disks, but I don't see how this could be lessened by using huge disk, > arrays instead of a bunch of little disks. >   8 	It's not the size but the number.  Google will be using@ 	(according to publically stated plans) 10000 servers containing8 	2 disks each for 20000 disks.  From what you find here:  : http://www.openvms.digital.com/openvms/brochures/NORTHERN/    I "Today, the Compaq NonStopT eBusiness solution deployed at Northern LightnJ includes four Compaq OpenVMS Alpha clusters. Each cluster consists of fourO AlphaServer GS140 systems, each with 10 EV6 525 MHz processors, 12 Gigabytes ofVN memory, and an average of 7.75 Terabytes of Compaq StorageWorks storage. FaultN tolerance is achieved through mirroring, RAID5 discs, HW-level redundancy, andO dual power supplies. The company is just bringing another three clusters online H in Waltham, Massachusetts, and is planning a third site outside of their geographical area."     < 	That's 31 Terabytes of storage.  Less than 2000 18 Gigabyte> 	disks.  That consumes a lot less power than 20000 disks.  One@ 	could speculate they may double (if they haven't already) theirD 	data disks to 36 Gigs and would be using far fewer than 2000 disks.  9 	But just as significant, the 16 AlphaServers (more soon,gA 	if not now) consume much less power than the 4000 servers Googlet 	has currently.    				Roba   ------------------------------  % Date: Thu, 29 Jun 2000 10:02:36 +0100nB From: Andrew Harrison SUNUK Consultancy <andrew.nospam@uk.sun.com>7 Subject: Re: OpenVMS clusters vs other systems clustersd* Message-ID: <395B10AC.6927BCE0@uk.sun.com>   Marty Kuhrt wrote:  p > In article <395A26EF.6A4F1D3@uk.sun.com>, Andrew Harrison SUNUK Consultancy <andrew.nospam@uk.sun.com> writes: > > jlsue wrote: > > J > >> On Wed, 21 Jun 2000 01:01:26 -0400, "Bill Todd" <billtodd@foo.mv.com> > >> wrote:  > >>J > >> To be completely fair, many of the reasons that VMS lost what hold itH > >> did have on the market was because of all the marketing lies (snake6 > >> oil) that the Unix segment was putting out there. > >>7 > > Just for a moment just consider packaging, all part 4 > > of marketing. How out of touch do you heve to be6 > > to try to market an OS in this day and age without< > > including an IP stack and all the accompanying utilities9 > > with it. What sort of message do you think this sends,7 > > to people building internet/intranet based systems.q >n= > So the mouthpiece for the wanna-be solid OS company refutes ? > unix marketing lies by telling an outright lie?  Is it irony, > > a complete detachment from reality, or just plain stupidity?  5 Oh come on, UCX is packaged now with OpenVMS but this : is a very belated move by Compaq. The Internet opportunity6 for OpenVMS was long gone before Compaq got arround to. including an IP stack in the basic OS package.  6 Remember the poster I responded to said that UNIX lies0 took (note the past tense) OpenVMS's market away from it.  3 Was it packaged at the low end with 7.1 and the 6.xa0 releases of OpenVMS, think about it the Internet9 revolution took off in the timeframe of both releases andr5 Compaq/Digital were still resolutely trying to charge  extra for UCX.  8 You could argue that at over 1000 dollars for an OpenVMS9 license that includes UCX bundled with it that Compaq arer. are still resolutely trying to charge for UCX.  6 There is an argumet raging on this newsgroup about the5 Google architecture which uses 1000's of Linux boxes.g  4 A sobering thought for you should be that Google can1 get a Linux box with the OS (and an IP stack) for 0 less than the cost of an OpenVMS license with an* IP stack (excepting the hobbyist program).     -- Andrew Harrisons Enterprise IT ArchitectT   ------------------------------  % Date: Thu, 29 Jun 2000 10:49:35 +0100hB From: Andrew Harrison SUNUK Consultancy <andrew.nospam@uk.sun.com>7 Subject: Re: OpenVMS clusters vs other systems clustersr* Message-ID: <395B1BAF.86509737@uk.sun.com>   "Main, Kerry" wrote:  	 > Andrew,e >n" > Thank you for the great setup !! >5K > >>> because DEC totally missed the move from people making apps to peopleoM > buying apps and most crucially it lost its position because DEC and laterlyaN > Compaq though that you didn't need to market the OS and its capabilities. << >d) > Applications, marketing .. absolutely !  >e/ > Check it out: (remember, you asked for it ..)s >lN > <http://www.oracle.com/cgi-bin/press/printpr.cgi?file=200006281130.25961.htmN > l&mode=corp&td=1&tm=7&fd=1&fm=5&status=Search&ty=2000&limit=50&fy=2000> (one > url) >nJ > "Compaq Adopts the Oracle Internet Platform Across Server Product Lines" >   O Oh Kerry, Kerry, Kerry. How many feet do you have, you relentless cheerfullness   N plus a willingless to shoot yourself in the foot/feet on a regular basis makes me* think you have more than the usual number.  N Firstly the Oracle Internet platform is not an application its an applicationsG development and deployment platform. Oracle do have end user commercial O applications like Oracle Applications 11/CRM etc et c none of which are covered"   by this announcement.p  M So you rushed in and proved my point by not seeming to be able to distinguish D between a commercial application (erp, hr, fi etc) and a development environment.  L I was unaware that the Oracle Applications server and the other parts of theF Internet platform did not run on OpenVMS, the picture was worse than I thought.  G So you now have what is analogous to the  "compilers and libraries" nownH all you have to do is convince the ISV's including Oracle who supply theO applications that use these "compilers and libraries" to port their application>   to your platform.h   So thank you Kerry in return   Regards  Andrew Harrison  Enterprise IT Architectu   ------------------------------  % Date: Thu, 29 Jun 2000 12:58:38 +0100n/ From: Nigel Arnot <sysmgr@maxwell.ph.kcl.ac.uk> 7 Subject: re: OpenVMS clusters vs other systems clustersn6 Message-ID: <009EC552.34AAE77D.9@maxwell.ph.kcl.ac.uk>   Andrew Harrison wrote:   > jlsue wrote: > I > > On Wed, 21 Jun 2000 01:01:26 -0400, "Bill Todd" <billtodd@foo.mv.com>o
 > > wrote: > >eP > > >VMS has no lock on high availability any more - well, it never did, but nowQ > > >its competition includes 'standard' Unix and even Windows environments.  ThesO > > >eagerness with which customers are accepting such solutions makes it clear Q > > >that VMS's advantages, real though they may be, are not decisive compared topL > > >other considerations (in other words, the competition offers *adequate*; > > >availability plus many attractions that VMS does not).y > > >- > >-I > > To be completely fair, many of the reasons that VMS lost what hold iteG > > did have on the market was because of all the marketing lies (snaker5 > > oil) that the Unix segment was putting out there.D > >5 > 5 > Ahh to old conspiracy theory raises its head again.n > 8 > OpenVMS didn't lose its position in the market because; > of lies by UNIX competitors. It lost its position becauseb= > it cost too much at the low end, because DEC totally missed 8 > the move from people making apps to people buying apps9 > and most crucially it lost its position because DEC ando: > laterly Compaq though that you didn't need to market the > OS and its capabilities. >  [snip - all sensible stuff]   @ For once I find myself in almost complete agreement with Andrew.  : The only marketing lies concerning unix that I encountered> emanated from Digital. If there ever was a conspiracy, it was , within the Digital Corporation, not outside.  & Sadly, all water under the bridge now.   	Yours,E
 		Nigel Arnotv- 		NRA@MAXWELL.PH.KCL.AC.UK                      7 		"In the beginning there was nothing, which exploded."o   ------------------------------  % Date: Thu, 29 Jun 2000 14:52:02 +0200>= From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com> 7 Subject: Re: OpenVMS clusters vs other systems clusterst) Message-ID: <395B4672.350C7DA9@gtech.com>t  ( Andrew Harrison SUNUK Consultancy wrote: > jlsue wrote:I > > To be completely fair, many of the reasons that VMS lost what hold itsG > > did have on the market was because of all the marketing lies (snakea5 > > oil) that the Unix segment was putting out there.m > 5 > Ahh to old conspiracy theory raises its head again.  > 8 > OpenVMS didn't lose its position in the market because > of lies by UNIX competitors.  ? You mean it is completely unthinkable that f.ex. a SUN employeev= posts to a VMS news-group that a certain mail-program for VMS 8 has the same security flaws as MS Outlook, and then when> no VMSers seems to be able to reproduce it then just silince ?    ; >                              It lost its position becauseg= > it cost too much at the low end, because DEC totally missedt8 > the move from people making apps to people buying apps9 > and most crucially it lost its position because DEC andv: > laterly Compaq though that you didn't need to market the > OS and its capabilities.   I would add a few things like:#   - gettng too much mixed up wth MSn7   - a very unclear OS strategy (with VMS as the looser)e+   - selloff many good products way to cheapo  5 But yes - Digital/Compaq must take a huge part of they blame themselves.p   Arne   ------------------------------  % Date: Thu, 29 Jun 2000 14:55:21 +0200 = From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>r7 Subject: Re: OpenVMS clusters vs other systems clustersm) Message-ID: <395B4738.11D22ACA@gtech.com>   ( Andrew Harrison SUNUK Consultancy wrote:7 > Oh come on, UCX is packaged now with OpenVMS but thise< > is a very belated move by Compaq. The Internet opportunity8 > for OpenVMS was long gone before Compaq got arround to0 > including an IP stack in the basic OS package.  5 > Was it packaged at the low end with 7.1 and the 6.xw2 > releases of OpenVMS, think about it the Internet; > revolution took off in the timeframe of both releases and 7 > Compaq/Digital were still resolutely trying to chargeg > extra for UCX.   ????  ; AFAIK Alphas started to be sold with NAS packages including < UCX after WWW was invented but before it became a commercial8 issue (the good old days where WWW was used to distrbute3 interesting information collected by volonteers !).e   Arne   ------------------------------  # Date: Thu, 29 Jun 2000 13:03:54 GMTe* From: Art Rice <arice.NOcSPAM@ue.itug.org>7 Subject: Re: OpenVMS clusters vs other systems clustersc8 Message-ID: <v6hmls8su11o9pa4jpkjteuanrkai8v67o@4ax.com>  1 On Wed, 28 Jun 2000 18:50:43 -0400, "Main, Kerry"l <Kerry.Main@compaq.com> wrote:   >Andrew, >t! >Thank you for the great setup !!e >rJ >>>> because DEC totally missed the move from people making apps to peopleL >buying apps and most crucially it lost its position because DEC and laterlyM >Compaq though that you didn't need to market the OS and its capabilities. <<  >c( >Applications, marketing .. absolutely ! >G. >Check it out: (remember, you asked for it ..) >"M ><http://www.oracle.com/cgi-bin/press/printpr.cgi?file=200006281130.25961.htmuM >l&mode=corp&td=1&tm=7&fd=1&fm=5&status=Search&ty=2000&limit=50&fy=2000> (oner >url)w > I >"Compaq Adopts the Oracle Internet Platform Across Server Product Lines"t >sM >The Oracle Internet Platform will be optimized for Compaq servers--includingcI >ProLiant(TM) running Windows NT(R)/Windows 2000(TM), AlphaServer runningyF >Tru64 UNIX(TM) or OpenVMS(TM), and the companies plan to optimize for >Himalaya(TM) K > running NonStop(TM) Kernel (NSK)--and StorageWorks(TM) storage offerings.n  F That's actually a step backwards in performance where the NSK PlatformD is concerned.  But it will allow some enterprise-wide uniformity.  IC don't look forward to Oracle being offered for NSK anytime soon and # then only in the OSS environment.  s  E Also, last time I checked, NonStopSQL was about 18,300 to $36,000(pereB system) + about $500/mo service.  Absolutly NO Licenses needed forC Client workstations.  No Licenses needed for developers.  From what B I've seen around here in the Unix side of the house, Oracle is way2 beyond that when you start counting everything up.  D (note) NSK products can Be paid for with a Monthly License charge orA One Time Cost.  The One Time Cost gets you a "perpetual" license.e9 Upgrades don't cost another 18,000 here and 12,000 there.l  I >This Compaq and Oracle E-Business Platform will be available from CompaqrM >during calendar 4Q00. With this agreement, Compaq will also adopt internallye >the Oracle Internet Platform. >sL > "Today's announcement represents a major step in our strategy to work withF >partners to build and deliver the best Internet infrastructure to ourJ >customers," said Michael Capellas, Compaq's President and CEO. "CustomersC >want e-business solutions that are reliable, scalable, and easy tosI >deploy--with the security of knowing they are backed by vendors they can C >trust. With the tight integration of our Internet technologies andmJ >professional services, no two vendors are better positioned to meet these0 >customer requirements than Compaq and Oracle."  >d: ><http://www.oracle.com/tellmemore/?228177> (Tell me more) > M ><http://www1.compaq.com/pressrelease/0,1494,wp%7E14583_2%21ob%7E31866_1_1,00t >.html>n >cH >[another thread has been started already, so I'll add my comments there >....] >t	 >Regards,b >l >Kerry Mainu >Senior Consultant,e >Compaq Canada >Professional Services >Voice : 613-592-4660r >FAX   : 819-772-7036f >Email : kerry.main@compaq.com >  >h >d >-----Original Message-----n( >From: Andrew Harrison SUNUK Consultancy" >[mailto:andrew.nospam@uk.sun.com]( >Sent: Wednesday, June 28, 2000 12:25 PM >To: Info-VAX@Mvb.Saic.Com8 >Subject: Re: OpenVMS clusters vs other systems clusters >  >a
 >jlsue wrote:e >rH >> On Wed, 21 Jun 2000 01:01:26 -0400, "Bill Todd" <billtodd@foo.mv.com>	 >> wrote:e >>K >> >VMS has no lock on high availability any more - well, it never did, butn >nowK >> >its competition includes 'standard' Unix and even Windows environments.t >TheH >> >eagerness with which customers are accepting such solutions makes it >clearM >> >that VMS's advantages, real though they may be, are not decisive compared  >tooK >> >other considerations (in other words, the competition offers *adequate*-: >> >availability plus many attractions that VMS does not). >> > >>H >> To be completely fair, many of the reasons that VMS lost what hold itF >> did have on the market was because of all the marketing lies (snake4 >> oil) that the Unix segment was putting out there. >> >g4 >Ahh to old conspiracy theory raises its head again. > 7 >OpenVMS didn't lose its position in the market becausec: >of lies by UNIX competitors. It lost its position because< >it cost too much at the low end, because DEC totally missed7 >the move from people making apps to people buying apps 8 >and most crucially it lost its position because DEC and9 >laterly Compaq though that you didn't need to market thes >OS and its capabilities.  >e4 >And by market I don't just mean demand creation and3 >brand awareness creation through advertising whichh8 >rather quaintly some people in this group semm to think  >will solve the OpenVMS problem. >c6 >I mean sales programs, loan equipment for developers,5 >special deals for developer systems. Co marketing of 8 >the OS and selected applications into that applications8 >vendors customer base. Cooperative deals to rationalise7 >software licensing charges (parity with NT applicationo7 >licensing for example). Marketing programs to selected : >vertical markets backed by things like the willingness to4 >pay for ports of apps ot in some cases the purchase9 >of the ISV itself that are key to sucess in that market.  >l< >This is all marketing and virtually none of it is addressed? >by the rather sad excercise so beloved of this group of trying-1 >to spot OpenVMS references in Compaq literature.m >o9 >Given the lack of any of the above in Digitals marketingi- >of OpenVMS and market development of OpenVMS 3 >one has to conclude that Digital/Compaq handed thel' >market in a plate to the UNIX vendors.i >o4 >Just for a moment just consider packaging, all part1 >of marketing. How out of touch do you heve to bee3 >to try to market an OS in this day and age withouta9 >including an IP stack and all the accompanying utilitiese6 >with it. What sort of message do you think this sends4 >to people building internet/intranet based systems. >e >Regards >Andrew Harrison >Enterprise IT Architect >d   --  
 Art Rice   **o# Special Data Processing Corporationk& --------------------------------------' All opinions expressed are mine and do N% not reflect the views of my employer.p   ------------------------------  # Date: Thu, 29 Jun 2000 12:46:45 GMTe* From: Art Rice <arice.NOcSPAM@ue.itug.org>7 Subject: Re: OpenVMS clusters vs other systems clustersH8 Message-ID: <nsgmls85prmam0p7ka1oqsvel9ma18rqhk@4ax.com>  E On Wed, 28 Jun 2000 17:25:19 +0100, Andrew Harrison SUNUK Consultancye! <andrew.nospam@uk.sun.com> wrote:W  	 <snipped>a >,4 >Just for a moment just consider packaging, all part1 >of marketing. How out of touch do you heve to bec3 >to try to market an OS in this day and age withoutr9 >including an IP stack and all the accompanying utilities.6 >with it. What sort of message do you think this sends4 >to people building internet/intranet based systems.    A It's not the only OS sold without TCP/IP.  Some of the others aresF doing quite well thank you.   Some niche markets do not require TCP/IPD and management in those areas really do not like paying for featuresF they are not going to use.  However,  things are changing because mostB of the new management software only runs on Microcrap machines andF TCP/IP is the most convienient method of interconnecting thos machines with the "real" computer.      -- s
 Art Rice   **d# Special Data Processing Corporationu& --------------------------------------' All opinions expressed are mine and do w% not reflect the views of my employer.p   ------------------------------  % Date: Thu, 29 Jun 2000 15:01:14 +0100:- From: Tim Llewellyn <tim.llewellyn@bbc.co.uk> 7 Subject: Re: OpenVMS clusters vs other systems clustersh) Message-ID: <395B56AA.FED125AA@bbc.co.uk>d   Nigel Arnot wrote:   > Andrew Harrison wrote: >g > > jlsue wrote: > > K > > > On Wed, 21 Jun 2000 01:01:26 -0400, "Bill Todd" <billtodd@foo.mv.com>t > > > wrote: > > >aR > > > >VMS has no lock on high availability any more - well, it never did, but nowS > > > >its competition includes 'standard' Unix and even Windows environments.  ThegQ > > > >eagerness with which customers are accepting such solutions makes it clearlS > > > >that VMS's advantages, real though they may be, are not decisive compared toeN > > > >other considerations (in other words, the competition offers *adequate*= > > > >availability plus many attractions that VMS does not).i > > > >t > > >hK > > > To be completely fair, many of the reasons that VMS lost what hold itbI > > > did have on the market was because of all the marketing lies (snake 7 > > > oil) that the Unix segment was putting out there.k > > >  > >b7 > > Ahh to old conspiracy theory raises its head again.o > >e: > > OpenVMS didn't lose its position in the market because= > > of lies by UNIX competitors. It lost its position becauset? > > it cost too much at the low end, because DEC totally missedo: > > the move from people making apps to people buying apps; > > and most crucially it lost its position because DEC andv< > > laterly Compaq though that you didn't need to market the > > OS and its capabilities. > >t > [snip - all sensible stuff]  > B > For once I find myself in almost complete agreement with Andrew.  2 Me too. Hey, whatever else, VMS needs the low end.   >e >g< > The only marketing lies concerning unix that I encountered > emanated from Digital.  F You mean "You'll be much better off replacing that clapped out old VMS box with a shiny unix one?" :-)-  ( > If there ever was a conspiracy, it was. > within the Digital Corporation, not outside. > ( > Sadly, all water under the bridge now. >o >         Yours, >                 Nigel Arnotu* >                 NRA@MAXWELL.PH.KCL.AC.UK >uG >                 "In the beginning there was nothing, which exploded.":   --6 Tim Llewellyn, OpenVMS Infrastructure, Remarcs Project0 MedAS at the BBC, Whiteladies Road, Bristol, UK.A Email tim.llewellyn@bbc.co.uk. Home tim.llewellyn@cableinet.co.uki  A I speak for myself only and my views in no way represent those of  MedAS or the BBC.O   ------------------------------  % Date: Thu, 29 Jun 2000 10:20:28 -0400 + From: "Main, Kerry" <Kerry.Main@compaq.com>i7 Subject: RE: OpenVMS clusters vs other systems clustersrJ Message-ID: <910612C07BCAD1119AF40000F86AF0D8052844AD@kaoexc4.kao.dec.com>   Andrew,0  8 Always such a cheery and positive response from you  :-)   Lets see now ..f  > <http://www.oracle.com/corporate/press/index.html?228214.html>E "Oracle iAS (see separate release) is not only a powerful applicationgG Server, but it includes all the middleware and server-based applicationn< functionality needed by virtually any e-business, including:   . Business intelligencea . Dynamic data caching . Forms services> . Component services (e.g., support for Enterprise Java Beans) . Portal servicesr" . Application integration services  H - Oracle positions changes its application strategy for the future to be$ more simplified and more integrated   I - this new strategy includes full support for all of the Compaq platformsdJ ie. Tru64, NT, NSK (for those that require it) AND OpenVMS. Customers haveJ choice of platforms depending on their specific requirements. This differs. from single solution vendors on the market :-)  L - Compaq is the only vendor at the announcement with Larry Ellison with thisK major Oracle announcement (Yes, I recognize that Oracle is a multi-platformsI vendor like Compaq, but, it is a sign of much deeper co-operation betweent the two companies)  H And with all of this, you say ho-hum ... course you likely felt the sameH about the other big Alpha news in the last few days as well - the Genome project:L http://www1.compaq.com/pressrelease/0,1494,wp%7E14583_2%21ob%7E31686_1_1,00. html  I Ah well, keep trying, I am sure you can come up with some better fud thanr the attached ..:   :-)l   Regards,  
 Kerry Main Senior Consultant,
 Compaq Canada  Professional Servicesc Voice : 613-592-4660 FAX   : 819-772-7036 Email : kerry.main@compaq.com-       -----Original Message-----' From: Andrew Harrison SUNUK Consultancy ! [mailto:andrew.nospam@uk.sun.com]s% Sent: Thursday, June 29, 2000 5:50 AMr To: Info-VAX@Mvb.Saic.Comm7 Subject: Re: OpenVMS clusters vs other systems clustersh     "Main, Kerry" wrote:  	 > Andrew,p > " > Thank you for the great setup !! >,K > >>> because DEC totally missed the move from people making apps to peoplehE > buying apps and most crucially it lost its position because DEC ande laterlytK > Compaq though that you didn't need to market the OS and its capabilities.  << >n) > Applications, marketing .. absolutely !e >w/ > Check it out: (remember, you asked for it ..)n >r > L <http://www.oracle.com/cgi-bin/press/printpr.cgi?file=200006281130.25961.htmI > l&mode=corp&td=1&tm=7&fd=1&fm=5&status=Search&ty=2000&limit=50&fy=2000>b (one > url) >uJ > "Compaq Adopts the Oracle Internet Platform Across Server Product Lines" >p  A Oh Kerry, Kerry, Kerry. How many feet do you have, you relentlesse
 cheerfullness   H plus a willingless to shoot yourself in the foot/feet on a regular basis makese me* think you have more than the usual number.  A Firstly the Oracle Internet platform is not an application its anm applicationsG development and deployment platform. Oracle do have end user commercialwG applications like Oracle Applications 11/CRM etc et c none of which are  coveredl   by this announcement.o  A So you rushed in and proved my point by not seeming to be able to  distinguish	D between a commercial application (erp, hr, fi etc) and a development environment.  L I was unaware that the Oracle Applications server and the other parts of theF Internet platform did not run on OpenVMS, the picture was worse than I thought.  G So you now have what is analogous to the  "compilers and libraries" now H all you have to do is convince the ISV's including Oracle who supply theC applications that use these "compilers and libraries" to port theirb application    to your platform.n   So thank you Kerry in return   Regardst Andrew Harrison  Enterprise IT Architectc   ------------------------------  % Date: Thu, 29 Jun 2000 15:19:22 +0100 - From: Tim Llewellyn <tim.llewellyn@bbc.co.uk>r; Subject: Re: OpenVMS loses big, was:  RE: Compaq advertisese) Message-ID: <395B5AEA.8C6D396D@bbc.co.uk>c   David Mathog wrote:   i > In article <001301bfe10e$74e4b2c0$14b324a6@CJ4733A>, arturo saavedra <arturo.saavedra@wcom.com> writes:e > >Compaq A Winner In Gene Race  > K > And OpenVMS is the big loser - because none of these genomics folks chosenD > OpenVMS to run on a single box, going instead with Tru64 or Linux.   >e >yM > And guys, these first few genomics sales are just the beginning.  There are K > going to be some huge sales to the pharmaceutical industry.  They will be I > building/using tools to manipulate and analyze the vast amounts of data H > present in the genomes of various organisms, gene expression patterns,M > protein strucures and every other type of biological information.  But thishI > isn't transaction processing, and it isn't even clear to me that OraclehG > style relational databases are going to play a large role.  So unlesseJ > OpenVMS gets head to head with Tru64 in performance, Unix compatibility,F > and price, I don't see any future for it in this field.  Or for that= > matter, in any other new field that may come down the road.V >b  K David, you make some very valid points. For non-mission critical scientificpM computing where security, stability and scalability is not such and issue andeJ where system admin effort is essentially free (academia) then VMS lost its footing in the late 80's.i  ? However, surely this genome work should not be treated with kid L gloves, from the security and data integrity angle? As I understand it, thisZ "breakthrough"  is not as major as it is being played up to be at the moment. Still, a lotV of work has to go on at the theoretical level, and at some point real experimentation, will be necessary.  J However, once you start having real people's genome data on your computer,J surely you have a data protection issue? Already I am hearing rumours thatK insurance companies might want genetic samples, which is quite frightening.sH It seems to me this sort of data is very personal and private and shouldO not be stored on a linux box set up by a grad student with only a limited grasptK of the security issues etc (the reason my old dept dumped Tru64 was becausetI exactly this happened, OK not with sensitive data, but the box got hackedm badly).a  G btw, you are always slagging off VMS performance for unix code comparediM to unix. Did you try actually developing any applications for VMS using VMS'si strengths recently?G     --6 Tim Llewellyn, OpenVMS Infrastructure, Remarcs Project0 MedAS at the BBC, Whiteladies Road, Bristol, UK.A Email tim.llewellyn@bbc.co.uk. Home tim.llewellyn@cableinet.co.ukt  A I speak for myself only and my views in no way represent those ofe MedAS or the BBC.A   ------------------------------  # Date: Thu, 29 Jun 2000 15:20:04 GMTp0 From: "Terry C. Shannon" <shannon@world.std.com>; Subject: Re: OpenVMS loses big, was:  RE: Compaq advertisesn& Message-ID: <Fwx92y.3qt@world.std.com>  6 "Main, Kerry" <Kerry.Main@compaq.com> wrote in messageD news:910612C07BCAD1119AF40000F86AF0D8052844A6@kaoexc4.kao.dec.com... > David, >tL > No one will argue that OpenVMS marketing suffered under the latter portionJ > of the old Digital regime or even the early Compaq days when things were > just settling in.f  B Yup. I have heard this directly from Capellas, Heil, and Marcello.   >t > However -e >NL > Does the COE project discussed in Terry Shannons articles not address someL > of the issues you raise in the attached ie. combining the good features ofD > OpenVMS with some of the core functions associated with UNIX OS's?  L (Terry here...) COE not only will equip OpenVMS with "Solaris-like" APIs, itE will guarantee that OpenVMS remains viable for a minimum of 15 years.  Probability Factor: 0.9999...e   >uJ > Would this not seem like the best of both worlds? The RASS (reliability,K > availability, scalability and security) and unique features (like Galaxy)m of> > OpenVMS combined with the Linux/UNIX applications available. >aI > In terms of marketing and applications, is not todays announcements forn newrI > Oracle IAS (Internet Application Server) middleware software being madel? > available on OpenVMS a good start to making up for lost time?i  5 Yep, as is the Tier One status now enjoyed by the OS!w   >-J > Are not the recent endorsements from major Customers like E*Trade on the neweI > Alpha GS Series a good sign that Customers (and ISV's) are re-examiningg" > their strategies around OpenVMS?  L Seems to me that $100M in INCREMENTAL NEW OPENVMS business in 1FQ00 ought to say something...   cheers,n   terry sp   ------------------------------   Date: 29 Jun 2000 15:47:55 GMT2 From: mathog@seqaxp.bio.caltech.edu (David Mathog); Subject: RE: OpenVMS loses big, was:  RE: Compaq advertises[, Message-ID: <8jfr3b$h48@gap.cco.caltech.edu>  x In article <910612C07BCAD1119AF40000F86AF0D8052844A6@kaoexc4.kao.dec.com>, "Main, Kerry" <Kerry.Main@compaq.com> writes: >David,i >:K >No one will argue that OpenVMS marketing suffered under the latter portionhI >of the old Digital regime or even the early Compaq days when things werey >just settling in.  I It turns out it wasn't just the marketing, was it?  Seems the engineering.G lagged rather a lot too - nobody bothered to keep the OS competitive ina< terms of IO performance.  That's by far the bigger failing.    >T
 >However - >fK >Does the COE project discussed in Terry Shannons articles not address some K >of the issues you raise in the attached ie. combining the good features ofxD >OpenVMS with some of the core functions associated with UNIX OS's?   H I don't know yet.  The requirements for a competitive general purpose OSL right now are very simple - it must be able to build and run code developed J on Unix systems, with no more than a trivial amount of code changes (such H as that between Unixes) required.  That's because all code that runs on I OpenVMS now comes from Unix, and anything that creates an incompatibilitymJ is a bug, not a feature.  OpenVMS has a chance of getting back in the game! once you a customer can do this: g  B $ ftp/user=anonymous/password="me@here.edu" ftp.lot_o_software.com ftp> binarys5 ftp> get pub/packages/program.tar.gz []program_tar.gze	 ftp> quitl $ gunzip program_tar.gz  $ tar -xf program_tar.;l $ bash bash> cd program/tru64source
 bash> make bash> ./newprogram -blah -BLAH bash>p$ $ diff result.txt example_result.txt no differences found  H and it completes with no problems.  That is, it builds with nothing moreK than minor modifications required to the makefile, the program source code,dI and when the binary runs it does everything at least as fast as Tru64 (oreH Linux), with the output produced exactly the same as on Tru64 (includingG especially when a "record" sent to a stream text file exceeds 32k.) AndlF yes, there really must be a bash and tcsh, because Makefiles and otherH supporting "glue" for many packages will only work properly within thoseJ shells.  If OpenVMS can't do that, then for small users who are not crazedI by the thought of losing a few bytes of data, Linux and Tru64 will remains the OS of choice for Alpha.   D It would also be nice if Compaq finally realized that its clusteringF technology could be used to sell more OpenVMS machines, but that it isK worth very little more than Unix NFS/NIS style solutions to most customers,iG and since groups of OpenVMS machines basically don't work together welloI unless in VMS clusters, they are presenting the customers with the choicenG of inexpensive clusters (Unix) or with OpenVMS, either no clustering orq unaffordable clustering. p   >tI >Would this not seem like the best of both worlds? The RASS (reliability,aM >availability, scalability and security) and unique features (like Galaxy) of > >OpenVMS combined with the Linux/UNIX applications available.   H I don't ever expect to own a Galaxy class machine - it is IRRELEVANT forJ shops smaller than data centers.  The guys who wrote all the software thatE grew up to run at Celera did it almost entirely on unix workstations.pI OpenVMS RASS is not substantially better than the competition (except for A security, which really is a lot better than on Unix.)  For 99% ofmK the market Unix style NIS/NFS is adequate, and the insanely high prices for6I "real" OpenVMS clustering are an unjustifiable expense.  Rather than pay cL for OpenVMS security they go and hide behind a firewall, which is much less J expensive.  If OpenVMS's features were worth as much as you wish they wereI Compaq would be selling scads of small clusters everywhere, but it isn't,oK the only sales we ever hear about are for datacenter style installations.   E Now you may say, "Celera is running a huge, sort of datacenter, stylerF installation, so why shouldn't they consider OpenVMS?"   Well, besidesK their not wanting to rewrite their code, a lot of what they do involves thetK generation and manipulation of a zillion small files, and OpenVMS is turtlet# slow at that particular operation. k  L >In terms of marketing and applications, is not todays announcements for newH >Oracle IAS (Internet Application Server) middleware software being made? >available on OpenVMS a good start to making up for lost time? a  I No, because I don't use Oracle on my own machines, and there was nothing  K else of interest there.  I suppose it's good news for the data center guys,i> and I'm happy for them, but its irrelevant for everybody else.   >-M >Are not the recent endorsements from major Customers like E*Trade on the newwH >Alpha GS Series a good sign that Customers (and ISV's) are re-examining" >their strategies around OpenVMS?   K Come on guy, E*trade is a new customer, but it's the same old market.  ReadsI R. Marcello's letter on the OpenVMS web site.  The only markets mentioned:K are process control (mostly done on chip assembly lines by VAXes) and largenH scale transaction processing.  Now Compaq may break it down into Health J care, stock exchanges, and so forth, but the bottom line is that they are I all doing basically the same thing on their machines, and E*trade is moreoB of the same.  Nobody is saying that OpenVMS isn't a great OS for aK datacenter doing transaction processing - the problem is with the other 99%o4 of computing applications that it isn't addressing.   L >However, while perhaps not fast enough for readers of this list, surely theK >events of the last few months can be seen as steps in the right direction?   H Umm, let's see.  A few months ago I had poor performance and lousy Unix I compatibility, and today I have the same poor performance and lousy Unix  I compatibility, but hear that steps are being taken which may, or may not, J adequately remedy this situation, and which will dribble out over the nextC (unspecified) time interval.  There has been no official commitmenttE anywhere in precise and clear terms to deliver the improvements whichnI really are needed (see above.)  In the meantime, I can assume that a goodpH fraction of the support costs we pay (admittedly, on the CSLG, much lessJ than business folks pay) will not go towards OpenVMS development, but willI instead by siphoned off to pay for development elsewhere in Compaq and tof% prop up the PC side of the business.    @ Compaq may be taking steps internally, but externally, motion is essentially undetectable.    Regards,   David Mathog mathog@seqaxp.bio.caltech.edu ? Manager, sequence analysis facility, biology division, Caltech a   ------------------------------  % Date: Thu, 29 Jun 2000 13:06:53 -0400 ' From: "Bill Todd" <billtodd@foo.mv.com>C; Subject: Re: OpenVMS loses big, was:  RE: Compaq advertisesi( Message-ID: <8jfvll$1ha$1@pyrite.mv.net>  9 Terry C. Shannon <shannon@world.std.com> wrote in message.  news:Fwx92y.3qt@world.std.com...   ...n  K > (Terry here...) COE not only will equip OpenVMS with "Solaris-like" APIs,t itG > will guarantee that OpenVMS remains viable for a minimum of 15 years.  > Probability Factor: 0.9999....  F 'Viable' is not a synonym for 'alive'.  My (oldish) dictionary definesH 'viable' as 'able to live and develop under normal conditions', 'able to take root and grow'.  I 'Develop' and 'grow' being the operative words, IMO.  To what extent does:K COE guarantee this, rather than just that VMS, while it may stagnate, won't1
 disappear?   - bill   ------------------------------  % Date: Thu, 29 Jun 2000 13:35:02 -0400e' From: "Bill Todd" <billtodd@foo.mv.com> ; Subject: Re: OpenVMS loses big, was:  RE: Compaq advertisesm( Message-ID: <8jg1ab$349$1@pyrite.mv.net>  = David Mathog <mathog@seqaxp.bio.caltech.edu> wrote in message & news:8jfr3b$h48@gap.cco.caltech.edu...L > In article <910612C07BCAD1119AF40000F86AF0D8052844A6@kaoexc4.kao.dec.com>,- "Main, Kerry" <Kerry.Main@compaq.com> writes:e	 > >David,o > >nE > >No one will argue that OpenVMS marketing suffered under the latterr portioncK > >of the old Digital regime or even the early Compaq days when things weret > >just settling in. > K > It turns out it wasn't just the marketing, was it?  Seems the engineeringpI > lagged rather a lot too - nobody bothered to keep the OS competitive inq= > terms of IO performance.  That's by far the bigger failing.d  D In some ways I agree whole-heartedly, but I'm still inclined to say,	 "But...".h  8 The VMS-engineering-paths-not-taken are subtler than theK VMS-marketing-paths-not-even-looked-at.  I've heard that a book called 'The G Innovator's Dilemma' addresses the problems of listening mostly to yourfK existing customer base when soliciting direction:  you never hear where thes> far wider group you'd *like* to have as customers wants to go.  B So while I believe that VMS won't thrive without considerably moreJ development work than is scoped out in its road map, I suspect that it mayL be difficult to get many people here to understand why that may be the case:I they're pretty satisfied with VMS as it stands, and the already-scheduleduH future development is largely based on any areas they're not quite happy with.-   ...6  F > It would also be nice if Compaq finally realized that its clusteringH > technology could be used to sell more OpenVMS machines, but that it isB > worth very little more than Unix NFS/NIS style solutions to most
 customers,I > and since groups of OpenVMS machines basically don't work together wellnK > unless in VMS clusters, they are presenting the customers with the choice I > of inexpensive clusters (Unix) or with OpenVMS, either no clustering orc > unaffordable clustering.  L [Insert yet another plug for a heterogeneous SAN file system that allows VMSG systems to sort-of-cluster, inexpensively, with Unix and NT systems, atlJ least as far as sharing data - and perhaps some simple IPC and distributed lock management - goes.]   >i > >cK > >Would this not seem like the best of both worlds? The RASS (reliability,cL > >availability, scalability and security) and unique features (like Galaxy) of? > >OpenVMS combined with the Linux/UNIX applications available.K >yJ > I don't ever expect to own a Galaxy class machine - it is IRRELEVANT for" > shops smaller than data centers.  J You might have said something similar about SMPs 15 years ago.  But if youJ *do* ever own a Galaxy-class machine, it may well be because it has become? as much of a commodity as SMPs are today, hence in no way a VMSn product-differentiator.c  . [Snipped the rest, all of which I agree with.]   - bill   ------------------------------  % Date: Thu, 29 Jun 2000 18:34:36 +0100y- From: Tim Llewellyn <tim.llewellyn@bbc.co.uk>s; Subject: Re: OpenVMS loses big, was:  RE: Compaq advertisesd) Message-ID: <395B88AC.5AABDFF9@bbc.co.uk>h   David Mathog wrote:o  z > In article <910612C07BCAD1119AF40000F86AF0D8052844A6@kaoexc4.kao.dec.com>, "Main, Kerry" <Kerry.Main@compaq.com> writes:	 > >David,  > >iM > >No one will argue that OpenVMS marketing suffered under the latter portionnK > >of the old Digital regime or even the early Compaq days when things werem > >just settling in. >pK > It turns out it wasn't just the marketing, was it?  Seems the engineeringnI > lagged rather a lot too - nobody bothered to keep the OS competitive in-= > terms of IO performance.  That's by far the bigger failing.  >   P Not that stupid write thru/write back cache issue again, David? If people cannotN appreciate this issues involved in performance versus caching strategy, surely8 they deserve evey lost/corrupted nibble they experience.   >t > >a > >However - > >eM > >Does the COE project discussed in Terry Shannons articles not address sometM > >of the issues you raise in the attached ie. combining the good features of E > >OpenVMS with some of the core functions associated with UNIX OS's?  >sJ > I don't know yet.  The requirements for a competitive general purpose OSM > right now are very simple - it must be able to build and run code developediK > on Unix systems, with no more than a trivial amount of code changes (sucheI > as that between Unixes) required.  That's because all code that runs oneK > OpenVMS now comes from Unix, and anything that creates an incompatibilitytL > is a bug, not a feature.  OpenVMS has a chance of getting back in the game" > once you a customer can do this: >l  I Yeah, well, it always amused me when HEP was jumping off VMS (and VM/CMS)oQ onto unix in the interests of code portability they did not work at all very hardt, to make their unix code portable to VMS etc.   >wD > $ ftp/user=anonymous/password="me@here.edu" ftp.lot_o_software.com
 > ftp> binarys7 > ftp> get pub/packages/program.tar.gz []program_tar.gzb > ftp> quits > $ gunzip program_tar.gzy > $ tar -xf program_tar.;i > $ bash > bash> cd program/tru64source > bash> make  > bash> ./newprogram -blah -BLAH > bash>o& > $ diff result.txt example_result.txt > no differences found >a  G Didn't anyone tell you not to run code you found on the net :-) (joke).o  J > and it completes with no problems.  That is, it builds with nothing moreM > than minor modifications required to the makefile, the program source code, K > and when the binary runs it does everything at least as fast as Tru64 (orrJ > Linux), with the output produced exactly the same as on Tru64 (includingI > especially when a "record" sent to a stream text file exceeds 32k.) AndfH > yes, there really must be a bash and tcsh, because Makefiles and otherJ > supporting "glue" for many packages will only work properly within thoseL > shells.  If OpenVMS can't do that, then for small users who are not crazedK > by the thought of losing a few bytes of data, Linux and Tru64 will remaine > the OS of choice for Alpha.t  I David, and those few bytes of data could mean the difference between lifeiN and death for someone with a terminal illness that some geneticists are tryingI to help a few years down the road of the genome project, because they aree4 running on a windoze box without ECCram or whatever.  F As someone who learned to do real life reatime science on RSX and VMS,Q Unix and Windows (whatever variety) scared the shit out of me then (10 years ago)  as it does now.-  6 Of course, if you really don't care about your data...     >rF > It would also be nice if Compaq finally realized that its clusteringH > technology could be used to sell more OpenVMS machines, but that it isM > worth very little more than Unix NFS/NIS style solutions to most customers,0I > and since groups of OpenVMS machines basically don't work together welluK > unless in VMS clusters, they are presenting the customers with the choicetI > of inexpensive clusters (Unix) or with OpenVMS, either no clustering orn > unaffordable clustering. >r  K Agreed, as Nigel Arnot I think claimed recently, to have a cluster you needEO to buy at least two, probably 3 boxes, so why pay thru the nose for the clusteroL licence when you've already doubled or tripled the vendor's hardware margin.J Cluster development costs at the lowend MUST have been recouped many timesL now. OK, maybe some high end cannibalization might occur, but ultimately VMSO will be more competitive because more people are using it for a greater variety  of applications.  F Billy boy and unix have made people think rebooting is a normal thing.   >  > >bK > >Would this not seem like the best of both worlds? The RASS (reliability,dO > >availability, scalability and security) and unique features (like Galaxy) ofl? > >OpenVMS combined with the Linux/UNIX applications available.n >nJ > I don't ever expect to own a Galaxy class machine - it is IRRELEVANT forL > shops smaller than data centers.  The guys who wrote all the software thatG > grew up to run at Celera did it almost entirely on unix workstations.jK > OpenVMS RASS is not substantially better than the competition (except forsC > security, which really is a lot better than on Unix.)  For 99% ofMM > the market Unix style NIS/NFS is adequate, and the insanely high prices for J > "real" OpenVMS clustering are an unjustifiable expense.  Rather than payM > for OpenVMS security they go and hide behind a firewall, which is much lessdL > expensive.  If OpenVMS's features were worth as much as you wish they wereK > Compaq would be selling scads of small clusters everywhere, but it isn't,0K > the only sales we ever hear about are for datacenter style installations.wG > Now you may say, "Celera is running a huge, sort of datacenter, styleuH > installation, so why shouldn't they consider OpenVMS?"   Well, besidesM > their not wanting to rewrite their code, a lot of what they do involves thehM > generation and manipulation of a zillion small files, and OpenVMS is turtleb$ > slow at that particular operation. >e  Q Application design issue. Don't apps developed on unix scare the shit out of you?e   >  > I > Umm, let's see.  A few months ago I had poor performance and lousy UnixuJ > compatibility, and today I have the same poor performance and lousy UnixK > compatibility, but hear that steps are being taken which may, or may not,tL > adequately remedy this situation, and which will dribble out over the nextE > (unspecified) time interval.  There has been no official commitmentmG > anywhere in precise and clear terms to deliver the improvements whichwK > really are needed (see above.)  In the meantime, I can assume that a goodeJ > fraction of the support costs we pay (admittedly, on the CSLG, much lessL > than business folks pay) will not go towards OpenVMS development, but willK > instead by siphoned off to pay for development elsewhere in Compaq and tot& > prop up the PC side of the business. >eB > Compaq may be taking steps internally, but externally, motion is > essentially undetectable.l >   I All we can do is wait and see, it seems Compaq is certainly taking noticeHN of Alpha now (spotted an job for senior Alphaserver salebods in the UK today),; just how much of an extra push VMS gets, we'll have to see.o  H hmmm, you didn't mention the graphics hardware support issue either yet.   --6 Tim Llewellyn, OpenVMS Infrastructure, Remarcs Project0 MedAS at the BBC, Whiteladies Road, Bristol, UK.A Email tim.llewellyn@bbc.co.uk. Home tim.llewellyn@cableinet.co.ukr  A I speak for myself only and my views in no way represent those ofu MedAS or the BBC.s   ------------------------------  % Date: Thu, 29 Jun 2000 16:38:51 +0100l  From: steven.reece@quintiles.com: Subject: Re: OpenVMS loses big, was: RE: Compaq advertises> Message-ID: <8025690D.005616A5.00@qedilc01.qedi.quintiles.com>   Tim Llwellyn wrote: M >>>However, once you start having real people's genome data on your computer,dJ surely you have a data protection issue? Already I am hearing rumours thatK insurance companies might want genetic samples, which is quite frightening. H It seems to me this sort of data is very personal and private and shouldO not be stored on a linux box set up by a grad student with only a limited grasptK of the security issues etc (the reason my old dept dumped Tru64 was becauseiI exactly this happened, OK not with sensitive data, but the box got hackedt
 badly).<<<  @ Are you sure it was hacked badly, rather than hacked well?  :-))  M Anyway, to be serious, I would imagine that there is more than one issue herep for data security.P Obviously the genome data must have come from somewhere, but the individuals mayN not be named or coded in any way.  In many research markets though the time toO market or time to paper is a key differential.  If company x gets their product N out six months ahead of company y then x will clean up and y will not do quiteN so well.  Similarly, the first man to publicise the atom bomb or the first manP to walk on the moon get remembered and get the kudos, the second and third there don't.  K Then there's the key issue (as with any system) to maintain data integrity.fJ There's no point in doing all this work on human genomes only to find thatK something got corrupt in the first week and you've been working with cruddy 
 data......   ------------------------------  # Date: Thu, 29 Jun 2000 14:14:22 GMTe# From: Tom Dockray <dockray@acm.org>" Subject: Re: OVMS 7.2 updateso8 Message-ID: <icmmls87fajlj3m9oh07irq94ooqiq9fjs@4ax.com>  @ On 27 Jun 2000 20:13:45 GMT, hoffman@xdelta.zko.dec.nospam (Hoff Hoffman) wrote:    >tL >  Um, which updates are you refering to?  (Sorry, I left my psychic powers ) >  on the kitchen table this morning. :-)a >pG >  Are you seeing any particular problems with your OpenVMS VAX system?u >mO > --------------------------- pure personal opinion --------------------------- M >   Hoff (Stephen) Hoffman   OpenVMS Engineering   hoffman#xdelta.zko.dec.come   Those that can be found at:o3 http://ftp1.support.compaq.com/public/vms/vax/v7.2/t  F So far the only problem I have encountered is some odd behavior by the COPY/FTP command.    -- a Tom Dockrayo Dockray Consulting 610-356-0624 dockray@acm.orgo   ------------------------------  % Date: Thu, 29 Jun 2000 13:30:33 +0100a- From: Tim Llewellyn <tim.llewellyn@bbc.co.uk>r Subject: Re: PARITY ERROR ) Message-ID: <395B4169.39872914@bbc.co.uk>    Nivlesh Chandra wrote:   > Need help.. please respondN > I have a tape drive that is used for backups. I have a script that runs at aM > specified time, get the operators to load the tapes and then it checks thatiK > the tapes are accessible. During this it initialises the tapes so that it K > starts fresh. Now this is where my problem lies. Sometimes the tapes thatiM > are used result in a parity error.. this is sometimes due to the tape media N > and sometimes it is just the tape drive that is playing up. I need to detectM > that a parity error has occurred and then take steps to find out what it isgN > due to .. I have thought of this ... when a parity error occurs.. the scriptN > will ask the operators to load another tape that passed the testing and thenJ > this will be loaded. If this tape passes the test then it means that theM > previous tape was faulty and thus it needs to be replaced (a new tape needsoN > to be initialised with the same label as that tape and the faulty tape needsG > to be trashed). If the second tape that is loaded also gives a paritysL > error.. then there is a high chance that the drive has given up on us thusK > the technical people need to be informed (well there will need to be some 4 > more tests done before this conclusion is reached)   >rN > Anyway when I get the parity error.. I do not know how I could capture it inJ > my script so that I can detect that a parity error has occurred and thus > react on it appropriately..l >eC > Can someone please tell me as to how I could go about doing this? % > Your help will be most appreciated.l >b  O This doesn't actually check for parity error but could easily be modified to doa so.s  - $ define/user sys$output sys$scratch:tape.tmps$ $ mount/noassist/over=id/write $TAPE $ !-0 $ ! Check return status from the MOUNT commnand.0 $ ! ============================================$ $ message = F$MESSAGE ("''$status'")( $ message1 = F$ELEMENT (0, " ", message) $ show symb message1/ $ IF (message1 .EQS. "%MOUNT-F-VOLINV,") .OR. -.5      (message1 .EQS. "%SYSTEM-F-DEVNOTMOUNT,") .OR. -o1      (message1 .EQS. "%MOUNT-F-DEVALLOC,") .OR. - -  (message1 .EQS. "%MOUNT-F-DEVALLOC,") .OR. - (      (message1 .EQS. "%MOUNT-F-MEDOFL,") $ THEN $ !m+ $ ! There is no tape loaded for the backup..+ $ ! =======================================  $   goto no_tape_loadedf $ ENDIFa $ !a $ ! The tape is not labelled.e $ ! =========================  $ tape_not_labelled = 0D, $ IF (message1 .EQS. "%MOUNT-F-NOTLABELMT,") $ THEN1 $   WRITE SYS$OUTPUT "-W-, tape is not labelled."  $   tape_not_labelled = 1  $ ENDIF    >1 > Thanks > Niv    --6 Tim Llewellyn, OpenVMS Infrastructure, Remarcs Project0 MedAS at the BBC, Whiteladies Road, Bristol, UK.A Email tim.llewellyn@bbc.co.uk. Home tim.llewellyn@cableinet.co.uki  A I speak for myself only and my views in no way represent those ofg MedAS or the BBC..   ------------------------------  # Date: Thu, 29 Jun 2000 13:01:57 GMTc/ From: Anders Wallin <anders_wallin@my-deja.com>x* Subject: Perl suddenly hangs on VMS system) Message-ID: <8jfhbq$6d8$1@nnrp1.deja.com>N  D I have installed Perl and built it from sources on my OVMS 6.2 Alpha system.o  C The distribution of Perl has worked beautifully on this machine fortD several months, but now does not. When invoking Perl my process justF "hangs". Well, things do happen (lots of CPU is consumed) by no outputH is created. Even miniperl does not work. It hangs just like the ordinary Perl.   H There seems to be nothing wrong with Perl itself (sort of). This machineH is part of a Alpha cluster. When logging in to another node and invokingC Perl things work just as expected. I am definately running the sameh EXE's from the different nodes.'   SUMMARY:>         The distribution worked beautifully for several monthsA         but then someone did something with this node that causeda@         Perl to "hang". The node has been rebooted several times!         but the problem persists.i  ?         For example the following example will only eat CPU andn         do nothing:e           $ PERL garbage.plt  @         The file "garbage.pl" does not exist. Even this will not%         be reported. Nothing happens.n  F         On the other (working) node the following immediately happens:           $ PERL garbage.plyF         Can't open perl script "garbage.pl": no such file or directory"         %RMS-E-FNF, file not found	         $t    A         If anyone has had a similar experience, and knows what tot>         do about it, I most certainly would appreciate to know         how to fix it.           Best Regards           Anders Wallins  . I attach an extract of a report from "perlbug"> ==============================================================  0 Site configuration information for perl 5.00503:  5 Configured by unknown at Mon sep 06 14:36:14.29 1999.N  B Summary of my perl5 (5.0 patchlevel 5 subversion 3) configuration:   Platform:a1     osname=VMS, osvers=V6.2-1H3, archname=VMS_AXPa7     uname='VMS seb100 V6.2-1H3 AlphaServer 1000A 5/400'n0     hint=none, useposix=false, d_sigaction=undef1     usethreads=undef useperlio=undef d_sfio=undef    Compiler:t4     cc='CC/DECC', optimize='undef', gccversion=undef     cppflags='undef'     ccflagsi? ='/Include=[]/Standard=Relaxed_ANSI/Prefix=All/Obj=.obj/NoList'p4     stdchar='char', d_stdstdio=define, usevfork=true2     intsize=4, longsize=4, ptrsize=4, doublesize=8G     d_longlong=define, longlongsize=8, d_longdbl=define, longdblsize=16e1     alignbytes=8, usemymalloc=N, prototype=definee   Linker and Libraries:f(     ld='Link', ldflags ='/NoTrace/NoMap'"     libpth=/sys$share /sys$library	     libs=p;     libc=(DECCRTL), so=exe, useshrplib=undef, libperl=undef-   Dynamic Linking:<     dlsrc=dl_vms.c, dlext=exe, d_dlsymun=undef, ccdlflags=''%     cccdlflags='', lddlflags='/Share':    & Sent via Deja.com http://www.deja.com/ Before you buy.-   ------------------------------  # Date: Thu, 29 Jun 2000 17:41:55 GMTe From: nospam@nohost.no.net* Subject: RS232 CTS/RTS flow control on VAX/ Message-ID: <sln2j3s7jev164@corp.supernews.com>   
 Greetings,=           is the use of RS232 hardware flow control on a VAX, 2 in particular, the CXY08 board described anywhere.  ; This is needed for flow control for a device on our system. 3 XON/XOFF cannot be used because of the application.e  7 The documentation on modem control which I got from DECt: aka Compaq states that if CTS is not "on" then VMS assumes? a login. If this is so that it would seem that CTS would not ber6 usable for flow control. I have set the terminal port,6 "nointeractive" but aside from changing more than one 8 terminal attribute I don't know what it does since it's  not described in the help.  = I also looked at DSR/DTR but the same documentation indicatess, that DTR is a signal to the modem to hangup.  ' Any clarification would be appreciated.h   Thanks.    ------------------------------  # Date: Thu, 29 Jun 2000 14:12:33 GMTs From: rleathe@my-deja.coms Subject: Satallite not booting) Message-ID: <8jflg0$9li$1@nnrp1.deja.com>S  G We installed a new Alpha 1200 running VMS 7.2. Part of the installationoH included applying the vms-v7.2_upgrade, vms-v7.2_sys, vms-v7.2_hardware,@ and vms-v7.1_graphics patches.  After installing the patches andC necessary layered products, we ran cluster_config.com and made thiscC node a boot node.  Next we ran cluster_config.com again and added a C satellite node, (another Alpha 1200).  We booted the satellite node:A via ethernet and watched as the node joined the cluster.  HoweverSE after the satellite joined the cluster the following error displ;ayed " on the satellite's console screen.  E *PKQDRIVER-W PKB0, pkq_alloc_queues: run autogen with feedback to setr NPAG_BAP_Paramst  A ePKQDRIVER-W- PKB0, pkq_unit_init_fork: Port can NOT come online,.% pkq_alloc_queues() returened failure.   6 %MSCPLOAD-I-CONFIGSCAN, enabled automatic disk serving   Waiting to tune system...u Waiting to tune system...a Waiting to tune system...H Waiting to tune system.... Waiting to tune system...o Waiting to tune system...i    F Then it just sits there. I tried running autogen with the following...  $ @sys$system:AUTOGEN GETDATA FEEDBACK  * then i rebooted but no luck, same problem.        Any ideas on how I can fix this?   thanks   russ    & Sent via Deja.com http://www.deja.com/ Before you buy.a   ------------------------------  # Date: Thu, 29 Jun 2000 16:06:57 GMTs1 From: "Mark D. Jilson" <jilly@clarityconnect.com>c" Subject: Re: Satallite not booting1 Message-ID: <395B73F6.D2EC5AF@clarityconnect.com>a  F Try this.  Remove the node and readd it but say no when it asks you ifF this satellite will be doing any MSCP serving and create the page/swapG files on the system disk.  If this work then you can run CLUSTER_CONFIGaD on the satellite to enable MSCP disk serving and manually create the$ page/swap files on a different disk.   rleathe@my-deja.com wrote: > I > We installed a new Alpha 1200 running VMS 7.2. Part of the installationeJ > included applying the vms-v7.2_upgrade, vms-v7.2_sys, vms-v7.2_hardware,B > and vms-v7.1_graphics patches.  After installing the patches andE > necessary layered products, we ran cluster_config.com and made this E > node a boot node.  Next we ran cluster_config.com again and added a E > satellite node, (another Alpha 1200).  We booted the satellite nodetC > via ethernet and watched as the node joined the cluster.  HoweveriG > after the satellite joined the cluster the following error displ;ayedr$ > on the satellite's console screen. > G > *PKQDRIVER-W PKB0, pkq_alloc_queues: run autogen with feedback to set  > NPAG_BAP_Paramso > C > ePKQDRIVER-W- PKB0, pkq_unit_init_fork: Port can NOT come online,l' > pkq_alloc_queues() returened failure.  > 8 > %MSCPLOAD-I-CONFIGSCAN, enabled automatic disk serving >  > Waiting to tune system...  > Waiting to tune system...  > Waiting to tune system...i > Waiting to tune system...a > Waiting to tune system...  > Waiting to tune system...s > H > Then it just sits there. I tried running autogen with the following... > & > @sys$system:AUTOGEN GETDATA FEEDBACK > , > then i rebooted but no luck, same problem. > " > Any ideas on how I can fix this? >  > thanks >  > russ > ( > Sent via Deja.com http://www.deja.com/ > Before you buy.u   --  D Jilly	- Working from Home in the Chemung River Valley - Lockwood, NY0 	- jilly@clarityconnect.com			- Brett Bodine fan. 	- Mark.Jilson@Compaq.com			- since 1975 or so, 	- http://www.jilly.baka.com               -   ------------------------------  % Date: Thu, 29 Jun 2000 16:50:59 +0000($ From: Steve.Spires@yellowpages.co.uk" Subject: Re: Satellite not booting/ Message-ID: <0025690D.005CA8AD.00@quegw01.btyp>x  = Contact:   Tel: 3063  -  VSSG, 1st Floor, Bridge Street Plaza    Russ,   O I had similar problems when moving to 7.2 upwards. Try setting these parameterse
 as I have;   NPAG_BAP_MIN=0 NPAG_BAP_MAX=0 NPAG_BAP_MAX_PA=1073741824  L This is NOT in line with what Compaq have put on DSN/WIS in the only articleF I could find relevant, except for the PA parameter which is set to the recommended 1Gb value.  L Setting the others to zero was the only way I could get any of my systems to boot.   P I had a call logged with the CSC in Reading (UK) asking for either a solution to  the initial problem, orJ an explanation of how it resolved the problem (if in fact it did) with the! settings shown above, but someone J decided that the call could be closed as the system was now working, and I haven't had the time or theaM inclination yet to chase it up. I did post here, you could try looking in they" DEJA archives for the discussions.  M So, again, this worked for me, but I am unsure as to why, or whether it wouldi# work for you, but it's worth a try.l   Steve Spires VMS System Manager BT/Yellow Pagesb        * rleathe@my-deja.com on 29/06/2000 14:12:33    To:        Info-VAX@Mvb.Saic.Com+ cc:         (bcc: Steve Spires/YellowPages) 7 From:      rleathe@my-deja.com, 29 June 2000, 2:12 p.m.o   Satallite not booting         G We installed a new Alpha 1200 running VMS 7.2. Part of the installationtH included applying the vms-v7.2_upgrade, vms-v7.2_sys, vms-v7.2_hardware,@ and vms-v7.1_graphics patches.  After installing the patches andC necessary layered products, we ran cluster_config.com and made thismC node a boot node.  Next we ran cluster_config.com again and added aeC satellite node, (another Alpha 1200).  We booted the satellite nodenA via ethernet and watched as the node joined the cluster.  HowevertE after the satellite joined the cluster the following error displ;ayedw" on the satellite's console screen.  E *PKQDRIVER-W PKB0, pkq_alloc_queues: run autogen with feedback to sett NPAG_BAP_Params   A ePKQDRIVER-W- PKB0, pkq_unit_init_fork: Port can NOT come online,,% pkq_alloc_queues() returened failure.l  6 %MSCPLOAD-I-CONFIGSCAN, enabled automatic disk serving   Waiting to tune system...i Waiting to tune system...p Waiting to tune system...h Waiting to tune system...t Waiting to tune system...  Waiting to tune system...h    F Then it just sits there. I tried running autogen with the following...  $ @sys$system:AUTOGEN GETDATA FEEDBACK  * then i rebooted but no luck, same problem.        Any ideas on how I can fix this?   thanks   russ    & Sent via Deja.com http://www.deja.com/ Before you buy.o   ------------------------------  % Date: Thu, 29 Jun 2000 10:11:53 +0100s2 From: Chris Sharman <Chris.Sharman@CCAgroup.co.uk>( Subject: Re: See how a device is spooled. Message-ID: <395B20E9.45174848@CCAgroup.co.uk>  , This is a multi-part message in MIME format.  & --------------329FE6E179740C0664B722C7* Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bite   Shael Richmond wrote:r > 5 > Is there a way to see where a device is spooled to?e > A > We have been creating LAT ports spooling to LPD/TelnetSym printcD > queues.  But doing a sho dev/full or sho term won't show the queue! > name - just that it is spooled.e  D Here's the source for my priv shareable image to do just that: usualH disclaimers apply. It would be nice if it could be added to $getdvi, but until then, here it is.e   Chriso  & --------------329FE6E179740C0664B722C7D Content-Type: text/plain; charset=us-ascii; name="ss_spoolqueue.mar" Content-Transfer-Encoding: 7bite9 Content-Disposition: inline; filename="ss_spoolqueue.mar"r  0 	.TITLE	SS_spoolqueue Spool Queue system service 	.IDENT	/V1.0/  o ;++r ; FUNCTIONAL DESCRIPTION:  ; H ;	 This module declares a system service to extract the spool queue name ;	 from a spooled device.s ; / ; AUTHOR:   Chris  , CREATION-DATE: 20-JUL-1992  ;  ; MODIFIED BY: d ;  ;--n ;  - 	.link		"sys$system:sys.stb"/selective_searchw 	.library	"sys$share:lib"m   	$devdef 	$plvdef 	$prdefr 	$shrdef 	$ssdefw 	$ucbdef 	$vcbdef  ? 	.external	sch$iolockr, sch$iounlock, ioc$searchdev, ctl$gl_pcb    	.dsabl	GLOBAL i .if defined plv$l_kernel ; VAXt$ 	.macro	xfr	name, cod=<<jmp	name+2>>- 	.psect	$$$transfer_vector,page,pic,nowrt,exeo 	.align	quad 	.transfer	namep 	.mask	namet 	.irp	lin, \codb 		link 	.endr 	.endm   	.macro	kss	name,narg=0-( 	xfr	name	<<chmk	#<kbase+kcount>>,<ret>> 	kcount = kcount + 1    	.psect	knarg,byte,pic,nowrt,exe 	.byte	nargM  ' 	.psect	kernel_disp1,byte,pic,nowrt,exem 	.signed_word	2+name-kcase_basen 	.endm  
 kcount = 0    	.psect	knarg,byte,pic,nowrt,exe knarg:   	kss	getdvi_queue,2   
 kbase = -1024s  , 	.psect	user_services,page,vec,pic,nowrt,exe  > 	.long	plv$c_typ_cmod, 0		; vector type change mode dispatcher% 	.long	kdisp-., 0			; k,e dispatchersy4 	.long	0, 0, 0, 0			; user rundown, resv, rms, check  ' 	.psect	kernel_disp0,byte,pic,nowrt,exe-   knotme:	rsb   / kdisp:	movab	w^-kbase(r0), r1		; normalize codef 	blss	knotme 	cmpl	r1, #kcount 
 	bgequ	knotmen   ; call is for us   	movzbl	knarg[r1], r1/ 	moval	@#4[r1], r1 	ifnord	r1, (ap), kaccvioe 	cmpb	(ap), knarg-kbase[r0]v 	blssu	kinsfargF 	casew	r0, #kbase, #<kcount-1> kcase_base:o' 	.psect	kernel_disp2,byte,pic,nowrt,exe  	.external	bug$_ivssrvrqst 	bug_check	ivssrvrqst, FATAL  	 kaccvio1:0 	brw	kaccvio' 	.psect	kernel_disp0,byte,pic,nowrt,exej :	 .if_false  ; Alphae( 	.psect	kernel_disp,long,pic,nowrt,noexe
 kcount = 0 ktable:t 	.macro	kss	name,narg=0e 	kcount = kcount + 1 	.address	name 	.endm   	kss	getdvi_queue,2r  , 	.psect	user_services,page,vec,pic,nowrt,exe  > 	.long	plv$c_typ_cmod, 0		; vector type change mode dispatcher' 	.long	kcount, 0			; k, e routine countt5 	.address ktable, 0, 0			; k, e routine list, rundown + 	.long	0, 0, 0, 0			; resv; rms; k, e flags   * 	.psect	kernel_routines,long,pic,nowrt,exe .endc   	 kinsfarg:p 	movzwl	#ss$_insfarg, r0 	ret   kaccvio: 	movzwl	#ss$_accvio, r0p 	ret  % 	.entry	getdvi_queue, ^m<r2,r3,r4,r5>c .if not_defined plv$l_kernelD ;	.call_entry label=getdvi_queue, preserve=<r2,r3,r4,r5>, max_args=3 	cmpb	(ap), #2 	blssu	kinsfarg  .endck ; Args: device, queue, queuelenp   	clrl	-(sp)h 	moval	(sp), (sp)    	ifnord	#8, @4(ap), kaccviot 	movl	4(ap), r1s 	movzwl	(r1), r0 	ifnord	r0, @4(r1), kaccvioa   	ifnord	#8, @8(ap), kaccviop 	movl	8(ap), r1  	movzwl	(r1), r0 	ifnowrt	r0, @4(r1), kaccvio   	cmpb	(ap), #3	 	blssu	1$a 	movl	12(ap), r0 	beql	1$ 	ifnowrt	#2, (r0), kaccvio 	movl	12(ap), (sp)  < ; Lock the I/O database for read (required by ioc$searchdev)   1$:	jsb	g^sch$iolockrh  ; returns mutex addr, not status, ; IPL = astdel (to prevent process deletion)7 ; Fatal bug to exit Kernel mode while holding any mutexn  7 ; Search for the device (with logical name translation)d   	movl	4(ap), r1y 	jsb	g^ioc$searchdev 	blbc	r0, 8$ ; R1 = UCB, R2 = DDB, R3 trash" ; We've found it - is it spooled ?   	movzwl	#shr$_noqueue, r0 & 	bbc	#dev$v_spl, ucb$l_devchar(r1), 8$  $ ; Copy the queue name out of the VCB   	movl	ucb$l_vcb(r1), r2a 	movzbl	vcb$b_qnamecnt(r2), r0 	movl	8(ap), r1d 	movw	r0, @(sp)+0 	movc5	r0, vcb$t_qname(r2), #^a" ", (r1), @4(r1) 	movl	#1, r0  ; ; All done - unlock the I/O DB, drop IPL, and return statuse   8$:	pushl	r0 	movl	g^ctl$gl_pcb, r4 	jsb	g^sch$iounlock=  	setipl	#0, environ=UNIPROCESSOR 	popl	r0 	ret   	.ENDE  ( --------------329FE6E179740C0664B722C7--   ------------------------------   Date: 29 Jun 2000 07:52:12 CDT= From: wayne@tachysoft.xxx.183943.killspam.00bb (Wayne Sewell)" Subject: Re: SMTP Mail. Message-ID: <vcJqSCwl7oU9@tachxxsoftxxconsult>  \ In article <395A4979.3B95745B@videotron.ca>, JF Mezei <jfmezei.spamnot@videotron.ca> writes: > Jerry Leslie wrote:=E >> The "smtp%" is no longer necessary as of VMS 6.2, per the VMS FAQ:  >>  = >>     http://www.openvms.digital.com/wizard/openvms_faq.html  > J > Ok, so when MAIL sees an @ and some dots in a email address (is that theP > actual logic used ?), it calls the SMTP foreign transport automatically. (just2 > tested it on a system without dec's SMTP on it). > N > Question: it is possible to configure mail to use a different transport when( > it thinks it has an internet address ?    < ------------------------------------------------------------J MAIL2.  How do I get IN% or MX% added automatically to Internet addresses?  E For older OpenVMS releases, you can acquire the MAILSHR_PATCH package B (there's one each for VAX and Alpha) from the WKU FILESERV server  (see question SOFT1.).  H As of OpenVMS V6.2, this is not necessary -- simply enter the SMTP emailK address directly.  If the address specified to MAIL contains an embeded "@"uJ character in it (a quoted string is not needed), MAIL will look to see if I the logical name MAIL$INTERNET_TRANSPORT is defined.  If it is, then MAILdJ will use the translation as the transport protocol, otherwise it will use ? the SMTP transport as is used by TCP/IP Services for OpenVMS.     B To alter this, for example, if you wanted IN% added, you'd define   MAIL$INTERNET_TRANSPORT as "IN".       I use:  ' $ define/sys mail$internet_transport mxi     -- mO =============================================================================== M Wayne Sewell, Tachyon Software Consulting  (281)812-0738  wayne@tachysoft.xxxs: http://www.tachysoft.xxx/www/tachyon.html and wayne.html  K change .xxx to .com in addresses above, assuming you are not a spambot  :-)xO =============================================================================== O Otter, on dining with Bluto:"It's perfectly safe if you keep your arms and legsi 			away from his mouth."   ------------------------------  % Date: Thu, 29 Jun 2000 09:19:09 +0200m5 From: Oswald Knoppers <Oswald.Knoppers@whitehouse.nl> ! Subject: Re: SMTP Mail - followupl, Message-ID: <395AF86D.9CDE0CD@whitehouse.nl>   Charlie Hammond wrote: > ) > In article <8jc6br$5tl$1@joe.rice.edu>,x- > leslie@clio.rice.edu (Jerry Leslie) writes:i > ..7 > >Upgrade to TCPIP V5.0A, unless you already have. ...x > .. > E > To avoid potential confusion, be aware that the POLYCENTER SoftwareaC > Installation (PCSI) utility calls this version of TCP/IP Services= > V5.0-10.  e.g.   No it doesn't:     $ prod show prod tcpip< ----------------------------------- ----------- ------------5 PRODUCT                             KIT TYPE    STATE < ----------------------------------- ----------- ------------9 DEC VAXVMS TCPIP V5.0-11            Full LP     Availablev< ----------------------------------- ----------- ------------  f 1 item found $   C The V5.0-10 version is 'almost' V5.0-11, all images are the same. Ii& think some header files are different.   Regards,   Oswald   ------------------------------  % Date: Thu, 29 Jun 2000 14:45:10 +0200 = From: Arne =?iso-8859-1?Q?Vajh=F8j?= <arne.vajhoej@gtech.com>,S Subject: Re: Sun Porting Clearinghouse for Staroffice, anybody going to port toVMS?l) Message-ID: <395B44D5.5580EE5D@gtech.com>    Wayne Sewell wrote:e` > In article <slk71d3pjev66@corp.supernews.com>, "Peter Weaver" <peter.weaver@stelco.ca> writes:J > > If you check out http://soldc.sun.com/staroffice/#porting you will seeG > > that Sun is looking for people who want to port StarOffice to othernD > > platforms. I know several people in the past have said they wereF > > interested in getting StarOffice ported to VMS, is anyone going toI > > attempt it? I do not have the C skills to even think about it myself,.% > > but maybe someone out there does.t > N > Wasn't the licensing scheme for this rather unfavorable?  I remember hearingQ > about it a few months ago.  Something about going to the effort of porting withaQ > the possibility that Sun could restrict/terminate distribution of the port on am > whim.O   Yep.  C There is no particular reason to fear that they will actually do t,n! but the license consdition sucks.p   Arne   ------------------------------  # Date: Thu, 29 Jun 2000 13:07:55 GMTt* From: Art Rice <arice.NOcSPAM@ue.itug.org> Subject: Re: VAX on Intel?8 Message-ID: <jaimlso06bj84oo6udfujuttnophufjs56@4ax.com>  7 On Wed, 28 Jun 2000 15:16:17 -0700, "Larry D Bohan, Jr"o" <LBohan@dbc.spam_less..com> wrote:  0 >On Wed, 28 Jun 2000 11:47:33 -0600, Glen Martin' ><GLENMARK@utxvms.cc.utexas.edu> wrote:l > C >>A VMS bigot who has been swamped with NT/Exchange admin duties ofh	 >>late...-H >>(Whatever idiot within the bowels of Redmond who decided that it was aJ >>good idea to have EVERYBODY'S mailbox in a SINGLE DATABASE should, IMHO,6 >>have evil things done to him. But I'm not bitter...)  ! Sounds more like MS Access to me.t   > % >My understanding is that the MSEXCH  0 >was built from a SQL engine, modified for speed1 >(prolbably at the expense of..., you guessed it,e >reliability/recoverability) >c. >the white papers I've seen, describing MSEXCH1 >recovery, illustrate that folks best suited  to  1 >MSEXCH admin, are those with DBA/admin training.d >e7 >seems like what was part-time duty on VMS (pmdf admin)i6 >translates into 3 full-time (MSEXCH admin) positions;* >Hidden expenses MSFT never mentions much. >t   -- l
 Art Rice   **s# Special Data Processing Corporationc& --------------------------------------' All opinions expressed are mine and do 0% not reflect the views of my employer.    ------------------------------   End of INFO-VAX 2000.361 ************************tic disk serving   Waiting to tune system...u Waiting to tune system...a Waiting to tune system...H Waiting to tune system.... Waiting to tune system...o Waiting to tune system...i    F Then it just sits there. I triedcol=ip	addr=10.43.243.67	disc-cause=1	disc-cause-ext=1045	pre-bytes-in=151	pre-bytes-out=91	pre-paks-in=6	pre-paks-out=4	bytes_in=40754	bytes_out=322114	paks_in=522	paks_out=821	pre-session-time=8	nas-rx-speed=0	nas-tx-speed=0	elapsed_time=742	nas-rx-speed=0	nas-tx-speed=0 CP-AS1.gw.saic.com	kopetp	Async49	async	stop	server=netsec-cs01	time=18:25:10	date=02/23/2000	task_id=20288	start_time=951358870	timezone=pst	service=ppp	protocol=ip	addr=10.11.242.17	disc-cause=2	disc-cause-ext=1011	pre-bytes-in=179	pr