1 INFO-VAX	Fri, 25 Jul 2003	Volume 2003 : Issue 407       Contents: cache 	 Re: cache 	 Re: cache 	 Re: cache 	 Re: cache 	 Re: cache 	 Re: cache 	 Re: cache 	 Re: cache  Compiling PINE 3.91-2 for AXP.O Re: D.H. Brown: EV7 AlphaServers Deliver Enhanced RAS and Powerful  Performance O Re: D.H. Brown: EV7 AlphaServers Deliver Enhanced RAS and Powerful  Performance O Re: D.H. Brown: EV7 AlphaServers Deliver Enhanced RAS and Powerful  Performance N Re: D.H. Brown: EV7 AlphaServers Deliver Enhanced RAS and Powerful PerformanceP Re: D.H. Brown: EV7 AlphaServers Deliver Enhanced RAS and Powerful Performance PP Re: D.H. Brown: EV7 AlphaServers Deliver Enhanced RAS and Powerful Performance PP Re: D.H. Brown: EV7 AlphaServers Deliver Enhanced RAS and PowerfulPerformance PoP Re: DEC-AXPVMS-VMS731_CDRECORD-V0100--4.PCSI v. DEC-AXPVMS-VMS731_PCSI-V0100--4.P Re: DEC-AXPVMS-VMS731_CDRECORD-V0100--4.PCSI v. DEC-AXPVMS-VMS731_PCSI-V0100--4.P Re: DEC-AXPVMS-VMS731_CDRECORD-V0100--4.PCSI v. DEC-AXPVMS-VMS731_PCSI-V0100--4. Re: EMC on VMS Encompass Members - Please Help / Re: Help Required Regarding DECNET Copy Command 1 Help required to run Java applications in OpenVMS 5 Re: Help required to run Java applications in OpenVMS E Re: IDC reports PC Market Share: Dell over HP by a nose; both pulling E Re: IDC reports PC Market Share: Dell over HP by a nose; both pulling , Re: IDC: HP ranks No. 1 in 2002 HPTC revenueP Re: Intel's had Dirty Bombs Anthrax and everything! (Was: Re: HP World: Why AlphG RE: Legal status of EVE source code ?, was: RE: Creating an EVE o	r TPU 2 Re: Migrate obsolete VAX/VMS SW to OpenVMS Itanium NEWBIE: DECWindows Bit Depth Re: OpenVms Backup Re: OpenVms Backup Re: OpenVMS Cluster overhead1 Re: OpenVMS homepage and 'sales.liveperson.net' ? 1 Re: OpenVMS homepage and 'sales.liveperson.net' ? " RE: Packed decimal arithmetic in C" Re: Packed decimal arithmetic in C" Re: Packed decimal arithmetic in C" Re: Packed decimal arithmetic in C" Re: Packed decimal arithmetic in C" Re: Packed decimal arithmetic in C" Re: Packed decimal arithmetic in C" RE: Packed decimal arithmetic in C PWS500au Boot failing  RE: PWS500au Boot failing  Re: PWS500au Boot failing C Re: Questions regarding OVMS 7.3-1 and Pathworks v6.0 compatibility # Re: Sun Micro Profit, Revenues Fall # Re: Sun Micro Profit, Revenues Fall # Re: Sun Micro Profit, Revenues Fall  Re: Sybase client  Re: Sybase client  TFTP timeout on tcp/ip services  Re: The PDP lives & Re: XFC, IO Size and Backup - Take Two& Re: XFC, IO Size and Backup - Take TwoA Re: [ANNOUNCEMENT] Migrate obsolete VAX/VMS SW to OpenVMS Itanium   F ----------------------------------------------------------------------    Date: 24 Jul 2003 12:24:20 -0700' From: nikki_wire@yahoo.com (nikki_wire)  Subject: cache= Message-ID: <60fdd9c9.0307240320.3073da10@posting.google.com>   B Okay I'm stumped - I typed 'cache' on the  vms command line and itC dropped me to an INS> prompt which i can't get out of!!! There's no 6 symbol or logcial for it and no help. Can anyone help? Thanks   ------------------------------  # Date: Thu, 24 Jul 2003 19:52:32 GMT # From: hoff@hp.nospam (Hoff Hoffman)  Subject: Re: cache0 Message-ID: <4OWTa.842$9u4.685@news.cpqcorp.net>  g In article <60fdd9c9.0307240320.3073da10@posting.google.com>, nikki_wire@yahoo.com (nikki_wire) writes: C :Okay I'm stumped - I typed 'cache' on the  vms command line and it D :dropped me to an INS> prompt which i can't get out of!!! There's no7 :symbol or logcial for it and no help. Can anyone help?   B   ^Z, ^Y, EXIT, HELP and such would be the normal approaches here.@   Failing that, you can potentially STOP/ID against the process,;   issuing that STOP command from another logged-in session.   C   You apparently have a Cache* DCL command verb around or -- if you A   should have the automatic foreign command mechanism's DCL$PATH  A   logical name defined -- you have a CACHE.COM or CACHE.EXE image '   in the command search path specified.   E   There is a third-party Cach product found on some OpenVMS systems; F   search for that or for Intersystems for related details.  Check yourC   local documentation and your system's installed product lists for    details, too.       N  ---------------------------- #include <rtfaq.h> -----------------------------K     For additional, please see the OpenVMS FAQ -- www.hp.com/go/openvms/faq N  --------------------------- pure personal opinion ---------------------------E         Hoff (Stephen) Hoffman   OpenVMS Engineering   hoff[at]hp.com    ------------------------------  # Date: Thu, 24 Jul 2003 20:11:34 GMT 3 From: hammond@not@peek.ssr.hp.com (Charlie Hammond)  Subject: Re: cache0 Message-ID: <W3XTa.844$cz4.835@news.cpqcorp.net>  g In article <60fdd9c9.0307240320.3073da10@posting.google.com>, nikki_wire@yahoo.com (nikki_wire) writes: C >Okay I'm stumped - I typed 'cache' on the  vms command line and it D >dropped me to an INS> prompt which i can't get out of!!! There's no7 >symbol or logcial for it and no help. Can anyone help?  >Thanks   > From another process do a SHOW PROCESS /ID=xxxxxx /CONTINUOUS.- This should tell you what image is being run.    --  J       Charlie Hammond -- Hewlett-Packard Company -- Ft Lauderdale  FL  USAF           (hammond@not@peek.ssr.hp.com -- remove "@not" when replying)J       All opinions expressed are my own and not necessarily my employer's.   ------------------------------    Date: 24 Jul 2003 16:21:57 -0500- From: Kilgallen@SpamCop.net (Larry Kilgallen)  Subject: Re: cache3 Message-ID: <LH9vcXFUxwSy@eisner.encompasserve.org>   V In article <4OWTa.842$9u4.685@news.cpqcorp.net>, hoff@hp.nospam (Hoff Hoffman) writes:i > In article <60fdd9c9.0307240320.3073da10@posting.google.com>, nikki_wire@yahoo.com (nikki_wire) writes: E > :Okay I'm stumped - I typed 'cache' on the  vms command line and it F > :dropped me to an INS> prompt which i can't get out of!!! There's no9 > :symbol or logcial for it and no help. Can anyone help?  > D >   ^Z, ^Y, EXIT, HELP and such would be the normal approaches here.  A I tried those when I got into that situation, and as I recall the A answer was not a standard VMS method but a standard MUMPS method. ? You might try QUIT (or more likely Q).  Or check back, somebody 6 with experience using Cache must check this newsgroup.  & Or you could ask in a MUMPS newsgroup.   ------------------------------    Date: 24 Jul 2003 17:41:43 -0500- From: Kilgallen@SpamCop.net (Larry Kilgallen)  Subject: Re: cache3 Message-ID: <KbOC3rQ7f0mJ@eisner.encompasserve.org>   v In article <p9ZTa.777$XI2.510446@newsfep1-win.server.ntli.net>, "Chris Casey" <chris.caseyNOSPAM@ntlworld.com> writes: > < > "Larry Kilgallen" <Kilgallen@SpamCop.net> wrote in message/ > news:LH9vcXFUxwSy@eisner.encompasserve.org... H >> In article <4OWTa.842$9u4.685@news.cpqcorp.net>, hoff@hp.nospam (Hoff > Hoffman) writes:B >> > In article <60fdd9c9.0307240320.3073da10@posting.google.com>,+ > nikki_wire@yahoo.com (nikki_wire) writes: H >> > :Okay I'm stumped - I typed 'cache' on the  vms command line and itI >> > :dropped me to an INS> prompt which i can't get out of!!! There's no < >> > :symbol or logcial for it and no help. Can anyone help? >> >G >> >   ^Z, ^Y, EXIT, HELP and such would be the normal approaches here.  >>D >> I tried those when I got into that situation, and as I recall theD >> answer was not a standard VMS method but a standard MUMPS method.B >> You might try QUIT (or more likely Q).  Or check back, somebody9 >> with experience using Cache must check this newsgroup.  >>) >> Or you could ask in a MUMPS newsgroup.  >  > Larry, > M > that is not a normal prompt for Mumps or Cache however you could try H (for + > Halt) or ? to see if any help is offered.  > Alto try <CTRL>T  B Perhaps not a normal prompt for MUMPS, but when I encountered thisJ problem in the presence of a MUMPS expert he said "Oh it's...<I forget>...6 which is what you would expect as a MUMPS programmer."   ------------------------------  % Date: Thu, 24 Jul 2003 23:34:11 +0100 4 From: "Chris Casey" <chris.caseyNOSPAM@ntlworld.com> Subject: Re: cache? Message-ID: <p9ZTa.777$XI2.510446@newsfep1-win.server.ntli.net>   : "Larry Kilgallen" <Kilgallen@SpamCop.net> wrote in message- news:LH9vcXFUxwSy@eisner.encompasserve.org... G > In article <4OWTa.842$9u4.685@news.cpqcorp.net>, hoff@hp.nospam (Hoff  Hoffman) writes:A > > In article <60fdd9c9.0307240320.3073da10@posting.google.com>, ) nikki_wire@yahoo.com (nikki_wire) writes: G > > :Okay I'm stumped - I typed 'cache' on the  vms command line and it H > > :dropped me to an INS> prompt which i can't get out of!!! There's no; > > :symbol or logcial for it and no help. Can anyone help?  > > F > >   ^Z, ^Y, EXIT, HELP and such would be the normal approaches here. > C > I tried those when I got into that situation, and as I recall the C > answer was not a standard VMS method but a standard MUMPS method. A > You might try QUIT (or more likely Q).  Or check back, somebody 8 > with experience using Cache must check this newsgroup. > ( > Or you could ask in a MUMPS newsgroup.   Larry,  K that is not a normal prompt for Mumps or Cache however you could try H (for ) Halt) or ? to see if any help is offered.  Alto try <CTRL>T   ------------------------------  % Date: Thu, 24 Jul 2003 23:53:05 +0100 4 From: "Chris Casey" <chris.caseyNOSPAM@ntlworld.com> Subject: Re: cache? Message-ID: <6rZTa.794$XI2.516775@newsfep1-win.server.ntli.net>   : "Larry Kilgallen" <Kilgallen@SpamCop.net> wrote in message- news:KbOC3rQ7f0mJ@eisner.encompasserve.org... H > In article <p9ZTa.777$XI2.510446@newsfep1-win.server.ntli.net>, "Chris/ Casey" <chris.caseyNOSPAM@ntlworld.com> writes:  > > > > > "Larry Kilgallen" <Kilgallen@SpamCop.net> wrote in message1 > > news:LH9vcXFUxwSy@eisner.encompasserve.org... J > >> In article <4OWTa.842$9u4.685@news.cpqcorp.net>, hoff@hp.nospam (Hoff > > Hoffman) writes:D > >> > In article <60fdd9c9.0307240320.3073da10@posting.google.com>,- > > nikki_wire@yahoo.com (nikki_wire) writes: J > >> > :Okay I'm stumped - I typed 'cache' on the  vms command line and itK > >> > :dropped me to an INS> prompt which i can't get out of!!! There's no > > >> > :symbol or logcial for it and no help. Can anyone help? > >> >I > >> >   ^Z, ^Y, EXIT, HELP and such would be the normal approaches here.  > >>F > >> I tried those when I got into that situation, and as I recall theF > >> answer was not a standard VMS method but a standard MUMPS method.D > >> You might try QUIT (or more likely Q).  Or check back, somebody; > >> with experience using Cache must check this newsgroup.  > >>+ > >> Or you could ask in a MUMPS newsgroup.  > > 
 > > Larry, > > J > > that is not a normal prompt for Mumps or Cache however you could try H (for- > > Halt) or ? to see if any help is offered.  > > Alto try <CTRL>T > D > Perhaps not a normal prompt for MUMPS, but when I encountered thisL > problem in the presence of a MUMPS expert he said "Oh it's...<I forget>...8 > which is what you would expect as a MUMPS programmer."  K Actually thinking about it, it could be a standard prompt if it referred to C a database named INS. In which case H<RET> should exit the process.    ------------------------------  % Date: Thu, 24 Jul 2003 19:27:31 -0500 1 From: "David J. Dachtera" <djesys.nospam@fsi.net>  Subject: Re: cache' Message-ID: <3F207973.FA474904@fsi.net>    nikki_wire wrote:  > D > Okay I'm stumped - I typed 'cache' on the  vms command line and itE > dropped me to an INS> prompt which i can't get out of!!! There's no 8 > symbol or logcial for it and no help. Can anyone help? > Thanks  . Cache' is the new name for InterSystems MUMPS.  - There is no on-line help in that environment.   . To exit, an "H" should do it ("HALT", get it?)   >H $    --   David J. Dachtera  dba DJE Systems  http://www.djesys.com/  ( Unofficial Affordable OpenVMS Home Page: http://www.djesys.com/vms/soho/    ------------------------------  % Date: Thu, 24 Jul 2003 19:28:43 -0500 1 From: "David J. Dachtera" <djesys.nospam@fsi.net>  Subject: Re: cache' Message-ID: <3F2079BB.3EDC855A@fsi.net>    Chris Casey wrote: > < > "Larry Kilgallen" <Kilgallen@SpamCop.net> wrote in message/ > news:KbOC3rQ7f0mJ@eisner.encompasserve.org... J > > In article <p9ZTa.777$XI2.510446@newsfep1-win.server.ntli.net>, "Chris1 > Casey" <chris.caseyNOSPAM@ntlworld.com> writes:  > > > @ > > > "Larry Kilgallen" <Kilgallen@SpamCop.net> wrote in message3 > > > news:LH9vcXFUxwSy@eisner.encompasserve.org... L > > >> In article <4OWTa.842$9u4.685@news.cpqcorp.net>, hoff@hp.nospam (Hoff > > > Hoffman) writes:F > > >> > In article <60fdd9c9.0307240320.3073da10@posting.google.com>,/ > > > nikki_wire@yahoo.com (nikki_wire) writes: L > > >> > :Okay I'm stumped - I typed 'cache' on the  vms command line and itM > > >> > :dropped me to an INS> prompt which i can't get out of!!! There's no @ > > >> > :symbol or logcial for it and no help. Can anyone help? > > >> >K > > >> >   ^Z, ^Y, EXIT, HELP and such would be the normal approaches here.  > > >>H > > >> I tried those when I got into that situation, and as I recall theH > > >> answer was not a standard VMS method but a standard MUMPS method.F > > >> You might try QUIT (or more likely Q).  Or check back, somebody= > > >> with experience using Cache must check this newsgroup.  > > >>- > > >> Or you could ask in a MUMPS newsgroup.  > > >  > > > Larry, > > > L > > > that is not a normal prompt for Mumps or Cache however you could try H > (for/ > > > Halt) or ? to see if any help is offered.  > > > Alto try <CTRL>T > > F > > Perhaps not a normal prompt for MUMPS, but when I encountered thisN > > problem in the presence of a MUMPS expert he said "Oh it's...<I forget>...: > > which is what you would expect as a MUMPS programmer." > M > Actually thinking about it, it could be a standard prompt if it referred to E > a database named INS. In which case H<RET> should exit the process.   A You're thinking of "namespace". This is displayed before the ">".    --   David J. Dachtera  dba DJE Systems  http://www.djesys.com/  ( Unofficial Affordable OpenVMS Home Page: http://www.djesys.com/vms/soho/    ------------------------------  % Date: Thu, 24 Jul 2003 22:08:30 -0400 # From: Tom Rymes <tomnews@rymes.net> ' Subject: Compiling PINE 3.91-2 for AXP. E Message-ID: <tomnews-44FFA8.22083024072003@news.comcast.giganews.com>   	 Hi folks,   I I am trying to compile PINE for use on my system (VMS 7.3-1 AXP). I will  ; be using it to check IMAP accounts on a different server...   I However, when I run the vmsbuild.com script, it complains that I have no  H C compiler installed. No surprise there. Now, I need to get my hands on G DEC C, or GCC, or something else. Of course, I'm fairly certain that I  G will need to purchase a PAK to install and use the DEC C compiler, and  A the only version of GCC that I can find seems to be old and only  E compiled for VAX. (Though I suppose it might be binary compatible...)   E Am I correct that I need a PAK for the DEC/Compaq/HP (or whatever) C  I compiler? If so, where can I find the best version of GCC for VMS on AXP?   A Sorry if these are FAQs, but I can't seem to dig this info up....   
 Thank you,   Tom    ------------------------------  # Date: Thu, 24 Jul 2003 18:05:17 GMT 9 From: "Fred Kleinsorge" <my-last-name@stardotzko.dec.com> X Subject: Re: D.H. Brown: EV7 AlphaServers Deliver Enhanced RAS and Powerful  Performance0 Message-ID: <xdVTa.826$fo4.566@news.cpqcorp.net>  7 "JF Mezei" <jfmezei.spamnot@istop.com> wrote in message # news:3F201A93.B43E95A6@istop.com...  > Fred Kleinsorge wrote:K > > The games Sun used in boosting a Spec result are widely known, and have B > > re-sparc'ed a lively debate on how this type of stuff might be
 prevented.I > > In short, a specific compiler optimization for a very limited type of ' > > application juiced up your numbers.  > I > Interesting you would mention this. I strongly suspect that IA64 is far  moreL > guilty of this type of behaviour than SPARC or other chips, simply because the G > EPIC concept requires the compiler to do very specific work to make a  program  > run fast.   J Eh?  Yes, I suppose that feedback optimized code works better on IA64 thanL other architectures.  Not quite sure that counts as "bad behavior" - even ifI it is used (I really don't know).  It's something that *generally* can be L applied to *any* application, and get the benefit.  Knowing how to recognizeK a particular array organization and access pattern in a very specific case, L and then knowing that the constraints allow very complext transformations toK be done... well, I congratulate the benchmarking language developers... but : don't see how it is applied to any real world application.   ------------------------------  % Date: Thu, 24 Jul 2003 13:13:51 -0700 3 From: Robert Klute <robert_klute_removethis@hp.com> X Subject: Re: D.H. Brown: EV7 AlphaServers Deliver Enhanced RAS and Powerful  Performance8 Message-ID: <moe0ivg0punds6g3en8qqpjp6deb5fr4u5@4ax.com>  H On Thu, 24 Jul 2003 13:42:44 -0400, JF Mezei <jfmezei.spamnot@istop.com> wrote:   >Fred Kleinsorge wrote: J >> The games Sun used in boosting a Spec result are widely known, and haveL >> re-sparc'ed a lively debate on how this type of stuff might be prevented.H >> In short, a specific compiler optimization for a very limited type of& >> application juiced up your numbers. > M >Interesting you would mention this. I strongly suspect that IA64 is far more O >guilty of this type of behaviour than SPARC or other chips, simply because the N >EPIC concept requires the compiler to do very specific work to make a program
 >run fast.  F Isn't the job of any compiler to do specific work that makes a programG run fast?  Isn't that what all those compiler flags are for?  I thought H the uproar was over the way the Sun compiler modifies data structures to optimize the 179.art segment?   M http://www.spec.org/osg/cpu2000/results/res2003q2/cpu2000-20030326-02001.html   H See John McCalpin's excellent post in Ace's Hardware for an explanation:0 http://www.aceshardware.com/forum?read=105021973  G Basically though, "The huge boost [in 179.art] comes from a combination H of a large cache and a set of code transformations that convert an array$ of structs into a struct of arrays".   ------------------------------  # Date: Thu, 24 Jul 2003 21:12:34 GMT 9 From: "Fred Kleinsorge" <my-last-name@stardotzko.dec.com> X Subject: Re: D.H. Brown: EV7 AlphaServers Deliver Enhanced RAS and Powerful  Performance0 Message-ID: <6ZXTa.850$VB4.569@news.cpqcorp.net>  L Yes.  My understanding is that the compiler is smart enough to determine theI specific cicumstances that allow it to safely do an optimization that has J dramatic effects on the benchmark.  Of course, the benchmark wasn't tryingJ to test how well that a specific structure layout could be optimized.  AndJ it's fragile in the sense that the circumstances are very limited (can youA say, nearly tailored for this test) in which it can safely do the F optimization.  The uproar is that optimization in general is good, butL optimizing for a specific benchmark using mechanism that can't be applied toG real-world applications is bad -- so did Sun cross that line?  A lot of F people think they did - that were it not for this benchmark (and theirH scores), they would not have tried to solve this problem.  They may haveK been within the letter of the rules, but most people trying to use the SPEC L values to compare systems, and compare versions of the same types of systemsD see this as chipping away at the legitimacy of the current benchmarkH numbers.  Of course, Sun's performance still is way behind even with theK optimization - it just took them out of the "not on the same planet" range.     @ "Robert Klute" <robert_klute_removethis@hp.com> wrote in message2 news:moe0ivg0punds6g3en8qqpjp6deb5fr4u5@4ax.com...J > On Thu, 24 Jul 2003 13:42:44 -0400, JF Mezei <jfmezei.spamnot@istop.com> > wrote: >  > >Fred Kleinsorge wrote: L > >> The games Sun used in boosting a Spec result are widely known, and haveC > >> re-sparc'ed a lively debate on how this type of stuff might be 
 prevented.J > >> In short, a specific compiler optimization for a very limited type of( > >> application juiced up your numbers. > > J > >Interesting you would mention this. I strongly suspect that IA64 is far moreE > >guilty of this type of behaviour than SPARC or other chips, simply  because the H > >EPIC concept requires the compiler to do very specific work to make a program  > >run fast. > H > Isn't the job of any compiler to do specific work that makes a programI > run fast?  Isn't that what all those compiler flags are for?  I thought J > the uproar was over the way the Sun compiler modifies data structures to > optimize the 179.art segment?  >  > L http://www.spec.org/osg/cpu2000/results/res2003q2/cpu2000-20030326-02001.htm l  > J > See John McCalpin's excellent post in Ace's Hardware for an explanation:2 > http://www.aceshardware.com/forum?read=105021973 > I > Basically though, "The huge boost [in 179.art] comes from a combination J > of a large cache and a set of code transformations that convert an array& > of structs into a struct of arrays".   ------------------------------    Date: 24 Jul 2003 22:07:57 -07001 From: keithparris_NOSPAM@yahoo.com (Keith Parris) W Subject: Re: D.H. Brown: EV7 AlphaServers Deliver Enhanced RAS and Powerful Performance = Message-ID: <cf15391e.0307240942.21f2192c@posting.google.com>   f young_r@encompasserve.org (Rob Young) wrote in message news:<Vv7ebBqO+8Dc@eisner.encompasserve.org>...G > 	And imagine how much better their margins will be when all their R&D C > 	for CPU is gone.  Sun did $482 million in R&D this last quarter. @ > 	They can't sustain that.  It isn't rocket science to conclude; > 	a large portion of that $482 million is CPU related R&D:   ? Semiconductor design is pretty tricky, too.  Witness the recent @ problems with undetected floating-point errors in UltraSparc III chips:? http://sunsolve.sun.com/pub-cgi/retrieve.pl?doc=fsalert%2F55081   @ Gartner recently lowered Sun's overall vendor rating to Caution:C http://www4.gartner.com/1_researchanalysis/vendor_rating/vr_sun.jsp    ------------------------------  % Date: Thu, 24 Jul 2003 13:42:44 -0400 * From: JF Mezei <jfmezei.spamnot@istop.com>Y Subject: Re: D.H. Brown: EV7 AlphaServers Deliver Enhanced RAS and Powerful Performance P ) Message-ID: <3F201A93.B43E95A6@istop.com>    Fred Kleinsorge wrote:I > The games Sun used in boosting a Spec result are widely known, and have K > re-sparc'ed a lively debate on how this type of stuff might be prevented. G > In short, a specific compiler optimization for a very limited type of % > application juiced up your numbers.   L Interesting you would mention this. I strongly suspect that IA64 is far moreN guilty of this type of behaviour than SPARC or other chips, simply because theM EPIC concept requires the compiler to do very specific work to make a program 	 run fast.    ------------------------------  % Date: Thu, 24 Jul 2003 18:00:56 -0400 * From: JF Mezei <jfmezei.spamnot@istop.com>Y Subject: Re: D.H. Brown: EV7 AlphaServers Deliver Enhanced RAS and Powerful Performance P ) Message-ID: <3F20570C.8D482DC3@istop.com>    Fred Kleinsorge wrote:H > optimization.  The uproar is that optimization in general is good, butN > optimizing for a specific benchmark using mechanism that can't be applied to@ > real-world applications is bad -- so did Sun cross that line?   J I assume everyone crosses the line, except those with a vested interest toI downgrade their own chip in order not to offend their competitor they are  trying so much to help.   F It is a bit like fuel consumption values for american cars. They don'tM represent real world values, they are derived from static tests in a lab with * the car on rollers and perfect conditions.  M Comparing SPEC values gives you an indication, but like fuel values, you have H to take those with a big grain of salt and use them to compare only. AndH because inaccuracies in the benchmark (and borderline cheating by all), N benchmarks that are reasonably close to each other should be considered equal.    F I reckon it is a bit similar to velodrome cycling. There are differentI copmpetitions which require VERY different skills. The Kilometre is brute K force and endurance to pain. (yes, even though it lasts less than a minute, G the rider is in excruciating pain after about 40 seconds). Sprint races K require les brute force and lots of adaptability (that is when the 2 riders H travel very slowly to try to outwit the other and control who starts the sprint first).    I With the advent of EPIC, I think that sprint-type benchmarks will have to L become more important. EPIC is brute force, pre-scripted behaviour since theJ smarts are in the compiler. Just like the coach who has told the kilometreL rider exactly how to start, accelerate and finish. That rider has absolutelyL nothing else to focus on than his own body/race, no competitor nothing else.  L But in real world computing, when you can have lots of possible outcomes andN perhaps unexpected turns, perhaps the sprint-type competition to measure how aN system reacts to unexpected turns would be best. And perhaps this is where theI real RISC systems would really win over those EPIC systems which lack the J smarts to continue to be extremely fast when conditions unpredicted by the compilers arise.   ------------------------------  % Date: Thu, 24 Jul 2003 14:01:14 -0400 * From: JF Mezei <jfmezei.spamnot@istop.com>Y Subject: Re: D.H. Brown: EV7 AlphaServers Deliver Enhanced RAS and PowerfulPerformance Po ) Message-ID: <3F201EE7.90AD1EBB@istop.com>   ( Andrew Harrison SUNUK Consultancy wrote:= > If you think back to the battles over the CPQ-HP merger you ? > will also remember that Walter Hewlett one of your main board < > members was clear that systems did not make HP money hence> > his opposition to merging with another company that produced > computer systems.   J My recollection was that Hewlett was opposed because the merger would haveG made HP focus too much on the unprofitable wintel market and not on the + profitable enterprise and printer business.   J And I strongly suspect that as part of the deal with Deutsche Bank/BankersG Trust, Carly probably had to agree to put a greater focus on enterprise N systems instead of wintel, a feeling that was clearly reflected in the may 7th! speeches that downplayed windows.   N The problem with HP's enterprise business is that it has no products and won'tL have for a couple of years. All its enterprise products are in transition orK in the process of being killed. So it is no surprise that it, combined with " recession, wouldn't be profitable.   ------------------------------  # Date: Thu, 24 Jul 2003 19:17:10 GMT 3 From: hammond@not@peek.ssr.hp.com (Charlie Hammond) Y Subject: Re: DEC-AXPVMS-VMS731_CDRECORD-V0100--4.PCSI v. DEC-AXPVMS-VMS731_PCSI-V0100--4. 0 Message-ID: <WgWTa.833$8t4.289@news.cpqcorp.net>  @ In article <OF42BE5FEF.294FE607-ON85256D6D.004A22BF@metso.com>,  norm.raphael@metso.com writes: ..J >Does PCSI check for enough space for the recovery data on the system disk >before proceeding? > >Is there a message informing how much space will be required?H >Can the option be changed on-the-fly if the space is deemed inadequate?= >Is the process terminated if the space is deemed inadequate? : >Can the recovery data be stored/accessed on another disk?= >[This seems akin to the dump-off-the-system-disk situation.]   K So far as I know, the answer is NO to all of the above -- at least for now. D I have forwarded these questions to the developers.  If they provice' additional information, I will post it.   I >If I have it correctly, if save is not done on the current eco, then all " >ability to undo previous eco's isB >forfeit, and all recoverysets are deleted, which seems draconian.  F You have it correctly.  Whether it is "draconinan" or not, this is theF necessary way to provide a generalized roll-back capability.  You willC find this in database management and trasaction processing systems.  (Even in RMS journaling!)   G What would it mean to "undo" a patch if you do not know what additional % patches have effected the same files?     K >Of course, all this begs the question of why one would need to undo a PCSI @ >install, since QA should have obviated any need for undoing....  E C'mon, Norm!  You might as well ask why one would need patches at all ! if the product had sufficient QA.   B              "Perfection is an abstraction; I'm merely damn good!"   --  J       Charlie Hammond -- Hewlett-Packard Company -- Ft Lauderdale  FL  USAF           (hammond@not@peek.ssr.hp.com -- remove "@not" when replying)J       All opinions expressed are my own and not necessarily my employer's.   ------------------------------  % Date: Thu, 24 Jul 2003 16:15:47 -0400  From: norm.raphael@metso.comY Subject: Re: DEC-AXPVMS-VMS731_CDRECORD-V0100--4.PCSI v. DEC-AXPVMS-VMS731_PCSI-V0100--4. ? Message-ID: <OF682882E7.500D9678-ON85256D6D.006EFC72@metso.com>   . > From:     hammond@not on 07/24/2003 03:17 PM  A > Please respond to hammond@not@peek.ssr.hp.com (Charlie Hammond)    > To:  Info-VAX@Mvb.Saic.Com > cc:  > ; > Subject:  Re: DEC-AXPVMS-VMS731_CDRECORD-V0100--4.PCSI v. +        DEC-AXPVMS-VMS731_PCSI-V0100--4.PCSI  >  > A > In article <OF42BE5FEF.294FE607-ON85256D6D.004A22BF@metso.com>,   > norm.raphael@metso.com writes: <snip>H > >Of course, all this begs the question of why one would need to undo a PCSIB > >install, since QA should have obviated any need for undoing.... > G > C'mon, Norm!  You might as well ask why one would need patches at all # > if the product had sufficient QA.  > D >              "Perfection is an abstraction; I'm merely damn good!" > G I'd respond to that, Charlie, but my tongue is still stuck firmly in my  cheek!J Actually, ECO's for improved functionality and new hardware would still be okay.   G "Everything worthwhile eventually degenerates into real work." - Murphy    -Norm  > --G >       Charlie Hammond -- Hewlett-Packard Company -- Ft Lauderdale  FL  USA J > >           (hammond@not@peek.ssr.hp.com -- remove "@not" when replying)J       All opinions expressed are my own and not necessarily my employer's.   ------------------------------  % Date: Thu, 24 Jul 2003 17:47:22 -0500 / From: Chris Scheers <chris@applied-synergy.com> Y Subject: Re: DEC-AXPVMS-VMS731_CDRECORD-V0100--4.PCSI v. DEC-AXPVMS-VMS731_PCSI-V0100--4. 3 Message-ID: <3F2061FA.A3C8B965@applied-synergy.com>    Charlie Hammond wrote: > ) > In article <3F1F4CB8.424DF5DD@fsi.net>, 5 > "David J. Dachtera" <djesys.nospam@fsi.net> writes:  > K > >I was going to suggest extracting the the CLD for the PRODUCT verb using C > >the freeware VERB utility and simply editing the .CLD to add the J > >"DEFAULT" attribute to the SAVE qualifier and then modifying DCLTABLES.< > >/SAVE does not appear to be in my V7.2-2 system, however. > F > That is not the problem.  It is this: There are old/existing commandD > procedures that execute PRODUCT INSTALL commands.  Since the /SAVEC > qaulifier did not exist, they do not use it.  In effect, they get C > /NOSAVE functionality, because there is no /SAVE functionality in % > older versions of the PCSI utility.  > L > If the default were to be made /SAVE, these old/exiting command proceduresJ > would now work differently.  Is that bad?  Yes!  It is a legitimate userK > consideration to determine whether or not the space required for recovery K > data sets is available.  (I have a copy of one patch kit --  tcpip_eco -- H > that is over 100,000 disk blocks.  I think some of the system remedial > kits can get pretty big too.)  > M > Same logic applies to /[NO]RECOVERY_MODE.  Imagine if an OpenVMS upgrade -- F > including DECwindows, TCP/IP and DECnet Plus -- suddenly operated inL > RECOVERY_MODE.  The recovery data set would take about as much room as theI > installed operating system and LPs.  I doubt that many people have that ' > much free space on their system disk!     D Way back when INSTALL changed, it was necessary to do something like   	INSTALL := INSTALL/COMMAND   H so that previously existing procedures would continue to work.  So thereE is precedence for changing the defaults of existing management tools.   H Why not define a logical that sets the default sense of the /SAVE flag? E It could be something like the MAIL$SYSTEM_FLAGS logical.  That would 1 give you a hook for other switches in the future.   C That way a site could opt to modify the behavior if they wished to.   G -----------------------------------------------------------------------a$ Chris Scheers, Applied Synergy, Inc.  C Voice: 817-237-3360            Internet: chris@applied-synergy.com l   Fax: 817-237-3074    ------------------------------    Date: 24 Jul 2003 18:16:39 -07001 From: keithparris_NOSPAM@yahoo.com (Keith Parris)  Subject: Re: EMC on VMSE= Message-ID: <cf15391e.0307240819.21d41960@posting.google.com>h  f young_r@encompasserve.org (Rob Young) wrote in message news:<DbRWP4Fq3P7e@eisner.encompasserve.org>...@ > 	Okay.  How about example?  In nearly 2 years of HBVS and EMC ; > 	Symmetrix, no bad blocks.  Am I shocked, surprised?  No.n > B > 	After all, if the Sym never has bad blocks (1), the Zen of the @ > 	situation is HBVS doesn't ever have to perform shadowing data/ > 	repair, READL/WRITEL never come into play.  -H > (1)  cache and disk scrubbing, everything a hyper (RAID 1 hardware) at > a minimum.  E The most-obvious problem area is in a shadow copy from a non-EMC disk D (which might have one or more forced-error flags set) to an EMC diskD -- Shadowing would kick out the EMC member out of the shadowset whenF it tried to write the forced-error flag and found it could not do so. C But that would only occur when first moving from non-EMC storage to D EMC storage; once all EMC storage was in place, that couldn't happen anymore, of course.   > Another, less-obvious problem area is that disks do experienceF uncorrectable ECC errors on reads occasionally in practice.  Tracks onD today's disks are very close together, and the fringe magnetic fieldE generated when writing one track overwrites the edges of the adjacentlE tracks to some extent.  Over time, especially if head alignment isn't E always spot on, the signal for a sector can weaken to the point whereh+ it might not be retrievable without errors.C  C Rob's advice to always keep a minimum of a 2-member mirrorsets, andCB use cache and disk scrubbing on EMC, is excellent advice, because:E 1) With the non-mirrored cache that EMC has, if enough bit errors due C to alpha particles or whatever accumulate, the cache contents for auA sector could conceivably be damaged beyond the limitations of the.E cache ECC scheme to detect and correct errors.  By scanning cache andr@ rewriting data that has any errors, you statistically reduce theE probability (but don't entirely eliminate the chance) of errors which 9 are beyond the ECC scheme's ability to detect or correct. A 2) As described above, with any single disk, there is a chance oflA getting uncorrectable ECC errors which result in lost (corrupted)-F data.  By constantly reading the data in the background (and rewritingA it if it has more correctable errors than a given threshold), yousE decrease (but don't entirely eliminate) the chance of a sector comings
 up unreadabledD 3) By keeping a 2-member mirrorset, if you ever do get an unreadableC sector, presumably the EMC does the right thing, gets the data fromdE the other mirrorset member, replaces the bad sector on the unreadable  disk, and all is well.  C But if you ever DO get an uncorrectable read error from a disk, the F EMC (and this problem is not limited to EMC -- any other ordinary SCSID or FC disk has the same problem) has no method of notifying the hostF that the data is still bad each time host software tries to access theC now-bad data.  It will presumably return an error on the very first!= I/O that returns bad data, but all subsequent accesses to the F now-corrupted data would have no error returned, so it would seem likeF the error had "gone away" (and in the VMS environment we refer to this7 as "undetected data corruption" and consider it totallyn unacceptable.).e  F This basic underlying problem was the rationale for coming up with theD "forced-error flag" in the Digital Storage Architecture in the firstF place.  Since DSA came out in the mid-'80s, I suspect the patents haveD now expired, and other vendors could offer similar functionality.  IF presume this function could now even be added to a SCSI standard.  ButA as things stand, HBVS simulates FE on SCSI/FC using READL/WRITEL.t  " >       EMC isn't violating a SCSI5 > 	protocol, READL/WRITEL are optional SCSI commands:a  @ I find it interesting that the MSA1000, which is based on CompaqC SmartArray technology, has had the READL/WRITEL capability added to-E its firmware recently in support of HBVS.  That's what EMC should do,t I think.   ------------------------------  % Date: Thu, 24 Jul 2003 19:08:39 -0600-' From: "Dale E. Coy" <dalecoy@spinn.net> ( Subject: Encompass Members - Please Help/ Message-ID: <vi10oon4ubaoa9@corp.supernews.com>X  I I want to run for election to the Encompass Board of Directors this year.pB I have been an active member of DECUS/Encompass for over 20 years.  L If you are a concerned Encompass member, please send the following petition:   ----------------- I As a member in good standing of Encompass, I nominate Dale E. Coy, membern2 166993, for the 2004 Encompass Board of Directors.  %     Signed (and printed if necessary)h     Date+     Encompass member number or Company Name  -----------------   B The petition must get to Encompass by August 8.  It must be either1 postal-mailed or faxed to Encompass headquarters.n4 E-mail is not allowed for this purpose by Encompass.     Encompass US Headquarters-% 401 North Michigan Avenue, 22nd FloorJ Chicago, IL 606115   fax.312-673-4609  D Thanks.  I've tried to be brief.  I apologize if you are bothered by multiple copiesmK of this request.  If you have questions, either ask here or e-mail to me ats dalecoy@spinn.netw    Dale    ------------------------------    Date: 24 Jul 2003 13:08:02 -0700' From: hari@transcomm.uk.com (HariHaran)a8 Subject: Re: Help Required Regarding DECNET Copy Command= Message-ID: <2de080f3.0307240429.1aced287@posting.google.com>f  g "John Travell" <john@travell.uk.net> wrote in message news:<bd9qj6$pthrm$1@ID-120847.news.dfncis.de>...s6 > "HariHaran" <hari@transcomm.uk.com> wrote in message9 > news:2de080f3.0306240709.64335003@posting.google.com...wK > > We have 2 VAx 4100 and 2 Alpha DS10 machines connected to our corporate  >  Lan.mM > > To copy a file from ALpha to Vax I have to use the following copy commandk> > > copy *.* nodename"userid password"::disk1:[destination]*.* > >-I > > Is there any possibility to override the username and password in the@ >  above command.KD > > I suspect that there should be a configuration change in Decnet. > > Is that true?n > > Can you please advice....n > > Thanks in Advance. >  > Look at DECnet proxies.y > N > DECnet 'knows' who you are and if a valid proxy for your user identity (bothH > node and username) is found on the destination node you do not need to > supply the login information.. > N > If 'disk1:[destination]' is the same as sys$login for the specified usernameK > on the target node you do not need to specify that detail. If you want to F > preserve the same filename you can omit that as well. So, with proxyJ > working, copy to the same filename in (remote) sys$login is as simple as > $ copy *.* nodename::  >  >  > -- > John Travell" > VMS crashdump expertise for hire > john@travell.uk.neth > http://www.travell.uk.net/ >  >  >  > ---w( > Outgoing mail is certified Virus Free.< > Checked by AVG anti-virus system (http://www.grisoft.com).C > Version: 6.0.489 / Virus Database: 288 - Release Date: 10/06/2003      Thank you very much indeed.e Regards, Hari   ------------------------------    Date: 24 Jul 2003 12:03:34 -0700' From: hari@transcomm.uk.com (HariHaran)m: Subject: Help required to run Java applications in OpenVMS= Message-ID: <2de080f3.0307240252.20fa109a@posting.google.com>T  B I am facing some techincal problems for running java applications.D I have installed Java 1.1.8 on our alpha ds10 server. I have applied. all the patches specified in the OpenVMS site.> I am able to compile my code properly but unable to run them. > please can some kind soul advice me how to solve these type of	 problems?.   Problem Description......s  D FileOP.java reads a file given as commandline argument and creates a	 new file.K5 Same code is working fine on another OpenVMS machine.S   screen dump follows.....   ' MBXSRV_Test: What next? sh lo classpatho<    "CLASSPATH" = "/sys$common/java/lib/JDK118_CLASSES.ZIP:." (LNM$PROCESS_TABLE)2+ MBXSRV_Test: What next? javac "FileOp.java"y MBXSRV_Test: What next? dir    Directory DQA0:[HARAN]   BACKUP16MAY2003.BCK;227 5                      7535367  16-MAY-2003 12:24:52.49  BACKUP16MAY2003.DIR;1f5                            1  16-MAY-2003 11:00:45.15l! DEC-AXPVMS-JAVA-V0101-85-1.PCSI;1Z5                        65488  10-JUL-2003 14:06:51.71s$ DEC-AXPVMS-JAVA131-V0103-13-1.PCSI;15                       177552   3-OCT-2002 11:10:53.43s/ DEC-AXPVMS-JAVA131-V0103-13-1.PCSI_SFX_AXPEXE;2e5                        61248  10-JUL-2003 14:21:03.24e* DEC-AXPVMS-JAVAAPIDOC131-V0103-11-1.PCSI;15                       201824   3-AUG-2001 12:41:37.14k5 DEC-AXPVMS-JAVAAPIDOC131-V0103-11-1.PCSI_SFX_AXPEXE;2l5                        17334  10-JUL-2003 14:21:39.11 ' DEC-AXPVMS-TCPIP_ECO-V0500-113-4.PCSI;1d5                        46400   7-JUN-2001 13:31:21.63 2 DEC-AXPVMS-TCPIP_ECO-V0500-113-4.PCSI-DCX_AXPEXE;25                        26975  10-JUL-2003 14:21:49.18:( DEC-AXPVMS-VMS721_LIBRTL-V0400--4.PCSI;25                        11888  10-DEC-2001 09:32:16.69l3 DEC-AXPVMS-VMS721_LIBRTL-V0400--4.PCSI-DCX_AXPEXE;2m5                         8012  10-JUL-2003 14:22:05.31$& DEC-AXPVMS-VMS721_PCSI-V0100--4.PCSI;15                         3600  13-APR-2000 09:15:23.35s1 DEC-AXPVMS-VMS721_PCSI-V0100--4.PCSI-DCX_AXPEXE;3n5                         2376  10-JUL-2003 14:22:10.09m( DEC-AXPVMS-VMS721_UPDATE-V0300--4.PCSI;15                       173856  25-JUN-2001 08:17:33.44 3 DEC-AXPVMS-VMS721_UPDATE-V0300--4.PCSI-DCX_AXPEXE;335                       117829  10-JUL-2003 14:22:11.51c5 DISC_BACKUP.COM;3          1  16-MAY-2003 12:23:56.52t5 DQA1_BACKUP.LIS;2       3932  16-MAY-2003 12:24:52.42r5 FILEOP.CLASS;1             3  24-JUL-2003 11:35:41.46T5 FILEOP.JAVA;2              2  30-MAY-2003 15:39:20.98p5 LOGIN.COM;17               3  26-MAR-2003 15:48:21.28 5 NET$SERVER.LOG;33          2  24-JUN-2003 18:44:09.31  UNZIP-ALPHA_V5.EXE;1  5                          232  24-JUL-2003 11:30:23.97e  " Total of 23 files, 8453928 blocks.# MBXSRV_Test: What next? dir *.classt   Directory DQA0:[HARAN]  6 FILEOP.CLASS;1             3  24-JUL-2003 11:35:41.46  Total of 1 files, 3 blocks.p5 MBXSRV_Test: What next? java "FileOP" disc_backup.com  Can't find class FileOP  MBXSRV_Test: What next? wh   DQA0:[HARAN]   ------------------------------  # Date: Thu, 24 Jul 2003 22:26:44 GMTo' From: Don Sykes <anonymous@pacbell.net>h> Subject: Re: Help required to run Java applications in OpenVMS+ Message-ID: <3F205B4C.123867C4@pacbell.net>e  / Alter your CLASSPATH log name to JAVA$CLASSPATH  then x@ @sys$common:[java]setup (or whatever the name of the setup file) then $ java "class.path.name"   HariHaran wrote: > D > I am facing some techincal problems for running java applications.F > I have installed Java 1.1.8 on our alpha ds10 server. I have applied0 > all the patches specified in the OpenVMS site.? > I am able to compile my code properly but unable to run them.i@ > please can some kind soul advice me how to solve these type of > problems?n >  > Problem Description......m > F > FileOP.java reads a file given as commandline argument and creates a > new file.n7 > Same code is working fine on another OpenVMS machine.y >  > screen dump follows....l > ) > MBXSRV_Test: What next? sh lo classpathH> >    "CLASSPATH" = "/sys$common/java/lib/JDK118_CLASSES.ZIP:." > (LNM$PROCESS_TABLE) - > MBXSRV_Test: What next? javac "FileOp.java"l > MBXSRV_Test: What next? dirw >  > Directory DQA0:[HARAN] >  > BACKUP16MAY2003.BCK;227a7 >                      7535367  16-MAY-2003 12:24:52.49h > BACKUP16MAY2003.DIR;1n7 >                            1  16-MAY-2003 11:00:45.15f# > DEC-AXPVMS-JAVA-V0101-85-1.PCSI;1 7 >                        65488  10-JUL-2003 14:06:51.71e& > DEC-AXPVMS-JAVA131-V0103-13-1.PCSI;17 >                       177552   3-OCT-2002 11:10:53.43-1 > DEC-AXPVMS-JAVA131-V0103-13-1.PCSI_SFX_AXPEXE;2N7 >                        61248  10-JUL-2003 14:21:03.24l, > DEC-AXPVMS-JAVAAPIDOC131-V0103-11-1.PCSI;17 >                       201824   3-AUG-2001 12:41:37.14 7 > DEC-AXPVMS-JAVAAPIDOC131-V0103-11-1.PCSI_SFX_AXPEXE;2h7 >                        17334  10-JUL-2003 14:21:39.11l) > DEC-AXPVMS-TCPIP_ECO-V0500-113-4.PCSI;1g7 >                        46400   7-JUN-2001 13:31:21.63 4 > DEC-AXPVMS-TCPIP_ECO-V0500-113-4.PCSI-DCX_AXPEXE;27 >                        26975  10-JUL-2003 14:21:49.18 * > DEC-AXPVMS-VMS721_LIBRTL-V0400--4.PCSI;27 >                        11888  10-DEC-2001 09:32:16.69n5 > DEC-AXPVMS-VMS721_LIBRTL-V0400--4.PCSI-DCX_AXPEXE;2 7 >                         8012  10-JUL-2003 14:22:05.31>( > DEC-AXPVMS-VMS721_PCSI-V0100--4.PCSI;17 >                         3600  13-APR-2000 09:15:23.35d3 > DEC-AXPVMS-VMS721_PCSI-V0100--4.PCSI-DCX_AXPEXE;3m7 >                         2376  10-JUL-2003 14:22:10.09y* > DEC-AXPVMS-VMS721_UPDATE-V0300--4.PCSI;17 >                       173856  25-JUN-2001 08:17:33.44a5 > DEC-AXPVMS-VMS721_UPDATE-V0300--4.PCSI-DCX_AXPEXE;3 7 >                       117829  10-JUL-2003 14:22:11.51u7 > DISC_BACKUP.COM;3          1  16-MAY-2003 12:23:56.52 7 > DQA1_BACKUP.LIS;2       3932  16-MAY-2003 12:24:52.42o7 > FILEOP.CLASS;1             3  24-JUL-2003 11:35:41.46 7 > FILEOP.JAVA;2              2  30-MAY-2003 15:39:20.98 7 > LOGIN.COM;17               3  26-MAR-2003 15:48:21.28a7 > NET$SERVER.LOG;33          2  24-JUN-2003 18:44:09.31  > UNZIP-ALPHA_V5.EXE;1 > 7 >                          232  24-JUL-2003 11:30:23.97s > $ > Total of 23 files, 8453928 blocks.% > MBXSRV_Test: What next? dir *.classp >  > Directory DQA0:[HARAN] > 7 > FILEOP.CLASS;1             3  24-JUL-2003 11:35:41.46/ > Total of 1 files, 3 blocks.M7 > MBXSRV_Test: What next? java "FileOP" disc_backup.com- > Can't find class FileOPa > MBXSRV_Test: What next? wh >   DQA0:[HARAN]   -- t   Have VMS, Will Travelc Wire paladin, San FranciscoA   (paladinATalphaseDOTcom)   ------------------------------  # Date: Thu, 24 Jul 2003 17:58:59 GMTy9 From: "Fred Kleinsorge" <my-last-name@stardotzko.dec.com> N Subject: Re: IDC reports PC Market Share: Dell over HP by a nose; both pulling0 Message-ID: <D7VTa.823$Vj4.713@news.cpqcorp.net>  K "Andrew Harrison SUNUK Consultancy" <Andrew_No.Harrison_No@nospamn.sun.com>y; wrote in message news:bfopsj$ff8$1@new-usenet.uk.sun.com...f > Rob Young wrote:   > >lA > > Face it - 9 straight quarters of declining revenue highlights  > > a company in decline.h > >c > >n >e1 > Odd, Sun's revenue last quarter was higher than 7 > it was the previous quarter and the previous quarterse0 > revenues were higher than the one before that. >e5 > Face it that isn't 9 straight quarters of decliningi
 > revenue. >M  K Let me help you.  Go to MSNBC, click on money, ask for a quote on SUNW, and 5 the click on Financials.  Takes all of a few seconds.-  G Here is the revenue for 01, 02, and 03 (I think the fiscal year ends inn June):  E [I hope this cut & paste works... but if you have problems just go tocL http://moneycentral.msn.com/investor/invsub/results/hilite.asp?Symbol=SUNW ]    :                         03                02            01  +             1st Qtr 2,747.0 2,861.0 5,045.0K+             2nd Qtr 2,915.0 3,108.0 5,115.0o+             3rd Qtr 2,790.0 3,107.0 4,095.0v+             4th Qtr 2,982.0 3,420.0 3,995.0i,             Total 11,434.0 12,496.0 18,250.0    K       If you look across each quarter from right to left (oldest to newest)oD You will see that in each quarter for the last 3 years - revenue hasL declined.  In each year, revenue has increased quarter-to-quarter - but thatJ is the normal buying cycles.  Sun has not managed to have a single quarterK in the last 3 years where revenue has increased from the previous year, and 5 total annual revenues are almost $7 billion US lower.e  %       Earnings... are also a problem:   1                  FY (06/03) FY (06/02) FY (06/01)t-                   1st Qtr -$0.04 -$0.06 $0.15 -                   2nd Qtr -$0.72 -$0.13 $0.12i,                   3rd Qtr $0.00 -$0.01 $0.04,                   4th Qtr $0.00 $0.02 -$0.02+                   Total -$0.76 -$0.18 $0.29                  Ohh dear.r    0 > Ohh dear, you can't be relied on to hype Alpha& > You can't be relied on to hype IA-642 > and now you can't be relied on to read a balance > sheet. >r   So what did he miss?   ------------------------------  % Date: Thu, 24 Jul 2003 16:56:48 -0400 * From: JF Mezei <jfmezei.spamnot@istop.com>N Subject: Re: IDC reports PC Market Share: Dell over HP by a nose; both pulling) Message-ID: <3F204808.9B47BDE9@istop.com>n   Fred Kleinsorge wrote:M > Let me help you.  Go to MSNBC, click on money, ask for a quote on SUNW, ando7 > the click on Financials.  Takes all of a few seconds.u  N What do the first two letters of MSNBC stand for ? And your trust Microsoft to) give you accurate information about Sun ?r   :-) :-) :-) :-) :-) :-) :-) :-)o  M NBC lost much credibility in its news gathering when it cosied up with Gates.b   ------------------------------    Date: 24 Jul 2003 21:24:32 -0700' From: icerq4a@spray.se (David Svensson)o5 Subject: Re: IDC: HP ranks No. 1 in 2002 HPTC revenueh= Message-ID: <734da31c.0307240929.43be2a0e@posting.google.com>k   Andrew Harrison SUNUK Consultancy <Andrew_No.Harrison_No@nospamn.sun.com> wrote in message news:<bfoph3$fae$1@new-usenet.uk.sun.com>...o > David Svensson wrote:? > > Andrew Harrison SUNUK Consultancy <Andrew_No.Harrison_No@nospamn.sun.com> wrote in message news:<bfm842$gqn$2@new-usenet.uk.sun.com>.../ > >  > >>David Svensson wrote:2 > >> > >>>Andrew Harrison SUNUK Consultancy <Andrew_No.Harrison_No@nospamn.sun.com> wrote in message news:<bfh0jr$lsd$1@new-usenet.uk.sun.com>... > >>>a > >>>  > >>>>Keith Parris wrote:  > >>>># > >>>>>From HP World News, July 17:e >  eA > >>>>>"HP ranks No. 1 in 2002 high-performance computing revenues > >>>>>tU > >>>>>HP cited industry research firm IDC to claim it is the worldwide market leaderMU > >>>>>for High Performance Technical Computing (HPTC) in 2002. IDC's "Worldwide High:P > >>>>>Performance Systems Technical Computing Census 2002" report shows that HPU > >>>>>commands 34 percent revenue share, more than a five percentage point lead overt > >>>>>IBM." > >>>>>e3 > >>>>>http://biz.yahoo.com/bw/030714/145265_1.html  > >>>> > >>>>G > >>>>Shame that the 3 big DARPA contracts for HPTC systems in the HPCS'A > >>>>program have gone in a pretty even split of ~50 million pere# > >>>>company to IBM, Sun and Cray.  > >>>h > >>>cF > >>>I can understand that Sun and Cray are in need for help, but IBM? > >>D > >>Interesting Idea HP didn't get a DARPA award because they didn't
 > >>need it !a > >> > >>Nice spin. > >> > >  > > E > > Yes. :) I was looking at the latest TOP500 supercomputer list andaF > > there were very few Sun and Cray machines there. On the other hand+ > > there were lots of IBM and HP machines.a >  > It gets better.  > A > So now your contention is that DARPA have a remit which extends B > to trying to balance out the top500 list entries on a per vendor: > basis by funding vendors that have a lower market share. > E > This is almost like traditional socialist redistribution economics.b > C > What a very ammusing theory I hope you didn't spend too much time  > thinking it up.i > 	 > Regardso > Andrew Harrisonn   :)E Those "contracts" are mainly used to keep the US HPC industry in goodpE shape. It was you who started the spin by saying it was not good that. HP didn't got a contract.3   ------------------------------   Date: 25 Jul 2003 06:40:57 GMT/ From: "Dave Weatherall" <djweath@attglobal.net> Y Subject: Re: Intel's had Dirty Bombs Anthrax and everything! (Was: Re: HP World: Why Alpho5 Message-ID: <DTiotGxQ0bj6-pn2-MMimwmfRsTkP@localhost>r  C On Wed, 23 Jul 2003 16:56:52 UTC, jlsue <jefflsxxxz@sbcglobal.net> p wrote:  J > On Fri, 18 Jul 2003 18:10:44 -0400, "Bill Todd" <billtodd@metrocast.net> > wrote: >  > >66 > >"jlsue" <jefflsxxxz@sbcglobal.net> wrote in message5 > >news:h5aghvkavsukjcto77hkv3qsd1ciu3g7le@4ax.com...o > ry.e > >> > > >>; > >> Right.  Somehow you feel they owe you an apology.(???)s > > M > >No, I feel they owe those who made decisions based upon their specific andu1 > >unequivocal 'commitments to Alpha' an apology.r >  <snip>  0 > >>  No real "treachery" either, just business. > >d > >Ah - the Mafia ethic. > L > Yeah, right.  They decide that the direction the company should go to meetM > business goals, and they incur the risks to take it there.  Hey, it's theiroJ > responsibility to keep the company profitable, if they don't do it, it's7 > their job on the line.  Armchair refs need not apply.   C Ok, then just accept Bill and the other people who've feel they've  F been let down, or other, stronger, emotions, as one of the risks. They3 will complain and seek to document their complaint.    -- 2 Cheers - Dave.   ------------------------------    Date: 24 Jul 2003 12:37:36 -0500B From: clubley@remove_me.eisner.decus.org-Earth.UFP (Simon Clubley)P Subject: RE: Legal status of EVE source code ?, was: RE: Creating an EVE o	r TPU3 Message-ID: <hLVRz$jsw9Dw@eisner.encompasserve.org>V   In article <95544AF90752D211AC0300A0C9E01D32019FFA1C@reaes4.sema.co.uk>, "POWERS, John" <John.POWERS@reading.sema.slb.com> writes:H > I suspect they have the legal status well sorted. I hope I am not bre= > aking F > any rules by redistributing the following  11 lines from EVE$BUILD.. >  > !r3 > !                      COPYRIGHT =A9 1987,1992 BY,G > !               DIGITAL EQUIPMENT CORPORATION, MAYNARD, MASSACHUSETTSs1 > !                           ALL RIGHTS RESERVED/ > !gH > ! THIS SOFTWARE IS FURNISHED UNDER A LICENSE AND MAY BE USED AND COPI= > EDB > ! ONLY IN ACCORDANCE WITH THE TERMS OF SUCH LICENSE AND WITH THEH > ! INCLUSION OF THE ABOVE COPYRIGHT NOTICE.  THIS SOFTWARE OR ANY OTHE= > R H > ! COPIES THEREOF MAY NOT BE PROVIDED OR OTHERWISE MADE AVAILABLE TO A= > NYF > ! OTHER PERSON.  NO TITLE TO AND OWNERSHIP OF THE SOFTWARE IS HEREBY > ! TRANSFERRED. > B > Sounds like high-class legalese to me. I know 'Digital EquipmentC > Corporation' know longer exist, but I bet they made sure all thisa= > copyright stuff got transferred to the new owners properly!t >   J Thanks for the pointer. I had already noticed the license but was not 100%J that it would stop you from taking the code and running it on Linux (but I think that it does).  M The reason that I was not sure was that, because you are encouraged to modifyiN EVE for your own needs, and are given the source code so that you _can_ modifyK it (unlike everything else in VMS unless you buy a source license), whether2N that means that all the usual restrictions on copying source code would apply.  L I thought that the usual restrictions would still apply, but was not sure so hence the question.5  L The implication is that a hobbyist could not create an implementation of TPU' on Linux and then run EVE on top of it.b   Simon.  M PS: I've looked at doing a Linux TPU implementation in the past, but it's toowO big a job for one person (and I became more comfortable with the Unix editors),sH so no, I do not have a Linux implementation of TPU ready to release. :-)   --  B Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP       L VMS advocate: One who makes a Mac advocate look like a beginner at advocacy.   ------------------------------  % Date: Thu, 24 Jul 2003 21:13:34 -0400w* From: "Stanley F. Quayle" <stan@stanq.com>; Subject: Re: Migrate obsolete VAX/VMS SW to OpenVMS Itaniumk. Message-ID: <3F204BFE.18671.50C3457@localhost>  1 On 24 Jul 2003 at 19:22, David J. Dachtera wrote:eI > Charon-VAX for OpenVMS-I64 (when/if it appears) may be of some value toc > some (but likely not all).    It's more a "when" than an "if".  D > A VAX -> I64 (direct) binary translator may have an audience also, > albeit somewhat limited.  B There won't be one of those.  But HP has many times said that the A Alpha -> IA-64 translator will support translated (VAX -> Alpha)  @ images.  That would be equivalent (still not 100% coverage, but  nothing is).  
 --Stan Quaylec Quayle Consulting Inc.  
 ----------C Stanley F. Quayle, P.E. N8SQ  +1 614-868-1363  Fax: +1 614 868-1671o1 8572 North Spring Ct. NW, Pickerington, OH  43147a= Preferred address:  stan@stanq.com       http://www.stanq.com    ------------------------------  % Date: Thu, 24 Jul 2003 22:10:02 -0400t# From: Tom Rymes <tomnews@rymes.net>l% Subject: NEWBIE: DECWindows Bit DeptheE Message-ID: <tomnews-C23664.22100224072003@news.comcast.giganews.com>o  	 Hi folks,n  H After much wind and fury, out trusty MicroVAX 3100/95 has been replaced  by a refurbed AS DS20. Woohoo!  I I have the machine start DECWindows on startup and all is well. However,  D I would like to change the Bit depth and/or resolution. My only X11 @ experience is with Linux, so I have no idea how to do this. Any H suggestions (without getting into the merits of running DECWindows on a G server machine...) ? I have searched on google, google groups, and the l VMS Docs, but to no avail....-  H The box in question is an AlphaServer DS-20 with an Elsa-Gloria Synergy ) card. The system is running OpenVMS 7.3-1r   Any help appreciated....    
 Thank you,   Tomr   ------------------------------  # Date: Thu, 24 Jul 2003 17:54:45 GMTHA From: "Colin Butcher" <colin_DOT.butcher_AT@xdelta_DOT.co_DOT.uk>s Subject: Re: OpenVms Backup = Message-ID: <F3VTa.2340$Ng5.22589694@news-text.cableinet.net>   L One tape per disc is a good thing. Makes life much easier with the restores.    Yup - use these backup commands:   to make savesets:r    $$$ BACKUP DKA300:/IMAGE/NOALIAS? <tape1><dka300_saveset_name>/SAVE/INIT/IGNORE=LABEL/BLOCK=32768u    $$$ BACKUP DKA100:/IMAGE/NOALIAS? <tape2><dka100_saveset_name>/SAVE/INIT/IGNORE=LABEL/BLOCK=32768e   to restore savesets:  G $$$ BACKUP <tape1><dka300_saveset_name>/SAVE/IMAGE/NOALIAS DKA300:/INITo  G $$$ BACKUP <tape2><dka100_saveset_name>/SAVE/IMAGE/NOALIAS DKA100:/INITf  L Then it should boot and run, assuming that those two backups worked and thatK those two discs hold everything that you need in order to be able to boot anI running system. If not then you'll have to find things you need, probablyt1 copy other discs, or start fixing things by hand.l    A -----------------------------------------------------------------e Hope this helps, Colin.-) colin DOT butcher AT xdelta DOT co DOT uk0L Systems Archaeologist - Investigation & troubleshooting of older systems and	 networks.A   ------------------------------  % Date: Thu, 24 Jul 2003 18:24:21 -0400e  From: John Santos <JOHN@egh.com> Subject: Re: OpenVms BackupF5 Message-ID: <1030724181849.1344B-100000@Ives.egh.com>m  ) On Thu, 24 Jul 2003, Colin Butcher wrote:e  N > One tape per disc is a good thing. Makes life much easier with the restores. > " > Yup - use these backup commands: >  > to make savesets:  > " > $$$ BACKUP DKA300:/IMAGE/NOALIASA > <tape1><dka300_saveset_name>/SAVE/INIT/IGNORE=LABEL/BLOCK=32768  > " > $$$ BACKUP DKA100:/IMAGE/NOALIASA > <tape2><dka100_saveset_name>/SAVE/INIT/IGNORE=LABEL/BLOCK=32768  >  > to restore savesets: > I > $$$ BACKUP <tape1><dka300_saveset_name>/SAVE/IMAGE/NOALIAS DKA300:/INITt > I > $$$ BACKUP <tape2><dka100_saveset_name>/SAVE/IMAGE/NOALIAS DKA100:/INITp  E The assumption here is that the new system has 2 disks with SCSI ID'sl= 1 and 3, both on controller A, just like the original system.t  B To return to the mysterious disappearance of the PAGEFILE, VMS canB find a page file on the system disk all by itself, but if the pageA file is on the other disk, there has to be a mount and an INSTALLdB command for it, normally in SYPAGSWPFILES.COM.  If the 2nd disk on@ the new system has a different controller letter or unit number,( then that file will need to be modified.  > The other issue is "are you booting from the same root?"  Look) at the console variable "boot_osflags"...s  nN > Then it should boot and run, assuming that those two backups worked and thatM > those two discs hold everything that you need in order to be able to boot aeK > running system. If not then you'll have to find things you need, probablyc3 > copy other discs, or start fixing things by hand.  >  > C > -----------------------------------------------------------------  > Hope this helps, Colin..+ > colin DOT butcher AT xdelta DOT co DOT uk N > Systems Archaeologist - Investigation & troubleshooting of older systems and > networks.-   -- - John Santos  Evans Griffiths & Hart, Inc. 781-861-0670 ext 539   ------------------------------  # Date: Thu, 24 Jul 2003 19:41:16 GMTA( From: "konabear" <maurert@ameritech.net>% Subject: Re: OpenVMS Cluster overhead2@ Message-ID: <wDWTa.12557$Vx2.6144605@newssvr28.news.prodigy.com>  J Keith is right on the money.  But there is are two more considerations I'd add.  I How is the 2nd system getting to the disk drives? If the 2nd node is only-I attached by Gigabit to the 1st, and if the 1st system must MSCP serve thecJ disks to the 2nd system, then mileage gained by adding the 2nd system will
 be even less.   G Is the first system an SMP system?  If so you may want to check out thenL Interrupt state on the primary processor too.  At least until v7.3-2 lockingJ traffic is only going to add to the Interrupt activity on the primary CPU.   Todd    > "Keith Parris" <keithparris_NOSPAM@yahoo.com> wrote in message6 news:cf15391e.0307221121.ae14c01@posting.google.com...C > andrew.rycroft@intrinsitech.com (Andrew Rycroft) wrote in message 9 news:<58ba0101.0307220546.4c818de6@posting.google.com>... J > > We have a Oracle/Rdb application which we are thinking of putting intoI > > a two node cluster with gigabit interconnect to enhance availability.PE > > But the question we need to address is whether by clustering thispE > > application the overall performance will be negatively impacted ?e >dG > If you simply added a 2nd node and formed a cluster, but continued to-C > run your application on the 1st node, you'd likely not notice any % > difference from being in a cluster.R >DH > If you put half of the workload onto the new 2nd node, two things will	 > happen:-G > 1) One of the nodes will be the lock-master for a given resource/lock H > tree, and users of that same resource/lock tree on the other node willH > have their lock requests satisfied in perhaps 200 microseconds insteadH > of perhaps 5 microseconds for a lock request on the same node which isA > mastering the lock tree.  Your users may or may not notice thisq
 > difference.e? > 2) The number of users on the first node will go down by 1/2,-E > potentially improving their response times.  But whether or not yourF > notice this will depend on how heavily-loaded the system was before. > F > A lot depends on where your performance bottleneck currently is.  IfF > you presently have a shortage of CPU or memory, adding a second nodeH > will likely result in significantly-better performance for your users.F >  If your current bottleneck is in the storage subsystem, performanceD > is likely to be exactly the same, and would get worse if you added > additional workload.   ------------------------------  % Date: Thu, 24 Jul 2003 16:14:11 -0400 , From: "warren sander" <warren.sander@hp.com>: Subject: Re: OpenVMS homepage and 'sales.liveperson.net' ?, Message-ID: <3f204390$1@usenet01.boi.hp.com>  I we do tell you.. the 'privacy' link and the 'terms of use' links have allS	 the info.   J right now 'liveperson' is monitoring a number of the 'bcs' (OpenVMS, Tru648 UNIX, AlphaServer, HP-UX, Superdome, Nonstop etc) sites.  E This is to build a clickstream database and enable them to be able top? 'pop-up' a window and help customers with sales related issues.a  I This is currently being used very successfully on several other HP sites.L  I This is because we are increasingly getting customer requests for help in 7 'buying' stuff. And since we 'sell stuff' it seems liketG a good idea to have this available.. Currently we are using the shotgune= approach and putting the monitoring tags on every page but astI time goes on and 'liveperson' goes live there will probably only be 10-20h: pages with this on it. Mostly the pages where we see folksL hitting the 'buy form' and other 'salesy' oriented pages. Ask the wizard and= the doc site will probably not have these tags at that point.n  C We also have Omniture tags on every page and collect data via 'sitem: catalyst'. I used to have keylime and webtrendslive on the& site also but those have been retired.  5 I also keep log files via the server (CSWS (apache)).R  L But everything is spelled out in the privacy policy. Infact the info-privacy7 folks are very active on a lot of fronts. Some of whichnK seem overboard to me. Like I want to collect the 'url' (reffering url) that ? a customer is on when the 'website feeback' link is hit (I have I my own for the OpenVMS site but I manage several other sites). We are notb< allowed to put the url into the mail message if the customerK doesn't put it in themselves because that is considered a privacy violation > because we can put the url together with a name/email address.J but it does make it easier to find/fix a problem if you know the url where@ it occurs or where the customer is wondering around the web siteG when they decide to give feedback. So if you ever do the feedback routeo> PLEASE voluntarily paste the url where you found the error etc onto the form.     -- dK --------------------------------------------------------------------------- 7 Warren Sander                   WW E-Marketing (HP.COM)dB Hewlett-Packard Company         Work:  warren.sander@remove.hp.comK 200 Forest Street MR01-3/K8     Personal: sander.ma.ultranet@remove.rcn.comy. Marlboro, MA 01752              (508) 467-48755    My opinions are my own and I only speak for myselfe*          Read http://www.hp.com/go/openvmsK ---------------------------------------------------------------------------g      1 "Michael Unger" <unger@decus.de> wrote in message 3 news:bfp0sg$gdprs$1@ID-152801.news.uni-berlin.de...lI > I just noticed this code snippet on the OpenVMS homepage source (at thei > very bottom):s >aL > > <!-- BEGIN HumanTag Monitor. DO NOT MOVE! MUST BE PLACED JUST BEFORE THE
 /BODY TAG -->t$ > > <!-- BEGIN HumanTag time out -->" > > <script language="javascript"> > > var lpInviteTimeout=120;
 > > </script>u" > > <!-- End HumanTag time out -->" > > <script language="javascript"> > > var lpPosY = 100;t > > var lpPosX = 100;a
 > > </script>t! > > <script language='javascript'tL src='http://sales.liveperson.net/hc/27889365/x.js?cmd=file&file=chatScript3&L site=27889365&imageUrl=http://www.hp.com/products1/servers/liveperson/proact ive/'>B >                                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ >oH > > </script><!-- END HumanTag Monitor. DO NOT MOVE! MUST BE PLACED JUST BEFORE THE /BODY TAG --> > >g >n9 > What about that "sales company" named "liveperson.net"?   > What about that "chatScript3"?E > Why doesn't HP tell the visitors of that page they are "monitored"?a >lI > Really strange. (JavaScript turned off for security reasons of course!)e >r	 > Michaely >e > -- e >hB > Please do *not* send "Security Patch Notifications" or "SecurityC > Updates"; this system isn't running a Micro$oft operating system. ? > And don't annoy me <mailto:postmaster@[127.0.0.1]> please ;-)  >    ------------------------------  % Date: Thu, 24 Jul 2003 17:01:34 -0400i* From: JF Mezei <jfmezei.spamnot@istop.com>: Subject: Re: OpenVMS homepage and 'sales.liveperson.net' ?( Message-ID: <3F204926.21722C5@istop.com>   warren sander wrote:L > right now 'liveperson' is monitoring a number of the 'bcs' (OpenVMS, Tru64: > UNIX, AlphaServer, HP-UX, Superdome, Nonstop etc) sites.  N But considering that a non trivial number of users accessing the VMS pages areK smart enough to turn off javascript, your big probably expensive statisticsnI are worthless since "liveperson" isn't getting any information from thoseoJ smart users with javascript off. But if you used your web server logs, youI would get some idea of how many visitors you really have compared to whatt liveperson tells you you have.   ------------------------------    Date: 24 Jul 2003 13:23:32 -0500 From: briggs@encompasserve.org+ Subject: RE: Packed decimal arithmetic in Ce3 Message-ID: <Y2LQEzjPviUi@eisner.encompasserve.org>s  _ In article <CIEJLCMNHNNDLLOOGNJIKEBGHKAA.tom@kednos.com>, "Tom Linden" <tom@kednos.com> writes:n >  >  >>-----Original Message-----B >>From: briggs@encompasserve.org [mailto:briggs@encompasserve.org]' >>Sent: Thursday, July 24, 2003 7:57 AMt >>To: Info-VAX@Mvb.Saic.Como- >>Subject: RE: Packed decimal arithmetic in C- >> >>@ >>In article <CIEJLCMNHNNDLLOOGNJIKEBCHKAA.tom@kednos.com>, "Tom" >>Linden" <tom@kednos.com> writes: >>>s >>>e >>>>-----Original Message-----D >>>>From: briggs@encompasserve.org [mailto:briggs@encompasserve.org]) >>>>Sent: Thursday, July 24, 2003 5:45 AMb >>>>To: Info-VAX@Mvb.Saic.Comn/ >>>>Subject: Re: Packed decimal arithmetic in Cn >>>> >>>>6 >>>>In article <3F1EE1FB.FE8CBFC4@istop.com>, JF Mezei' >>>><jfmezei.spamnot@istop.com> writes:o >>>>> John Reagan wrote:D >>>>>> If they were only that smart...  I've seen dozens of examples >>of peopled? >>>>>> using binary floating to track currency (and each one ise >>broken).  I'veL >>>>>> even seen them in languages that provided scaled integers like BASIC. >>>>>iE >>>>> Interesting since there are often standards on exactly how many  >>>>decimals topL >>>>> use when specifying currency and conversion factors, as in the case ofF >>>>> international SWIFT transactions. With floating point, you can't >>>>specify how.G >>>>> many decimals to consider during arithmetic, only during display.  >>>>F >>>>Ahhh.  The point being that floating point considers _more_ digitsE >>>>in some cases than the standards specify and thus gets the answer: >>>>"wrong" for that reason? >>>oL >>> No the point being that it doesn't consider digits, it deals in bits and	 >>> thereeJ >>> is not an isomorphism between the two bounded spaces, i.e. no closure. >>C >>You'll have to get a whole lot more specific than that before you> >>can make your case.r >>@ >>Sure, scaled decimals implemented with integer-valued floatingA >>point is not closed under addition, subtraction, multiplication  >>or division. > L > Firstly, the dweeb comments are gratuitous.  Secondly, what do you mean byN > integer-valued floating point?  If you are referring to a float in the senseK > of, say, Fortran, then you can approximate the value of a decimal, but in M > general, not completely represent it.  I refer to this as a float binary toiF > distinguish it from float decimal, both supported data types in PL/I   Integer-valued floating point:   1.0E+0, 2.0E+0, 3.0E+0, etc.  " Non-integer-valued floating point:   1.5E+0, 2.1E+0, 4.7E-1, etc.  C Lemma:  For a n bit mantissa (plus 1 hidden bit) and integer-valuedo> operands in the range from -2^n + 1 .. 2^n - 1, floating point< addition will yield a result that exactly coincides with the9 corresponding result for mathematical integer arithmetic.d  : Similar results obtain for subtraction and multiplication.  , >>But as long as you don't overflow past the> >>continuous range of integers that can be represented exactly= >>in floating point, it is closed under addition, subtractions8 >>and multiplication.  Neither packed decimal nor scaled9 >>decimal using floating point are closed under division.s >>9 >>So there's no case to be made for a lack of isomorphismp >>based on a lack of closure.: >>< >>If you want to make a case for a lack of isomorphism basedF >>on the inconsistency in round off error in division handling between= >>packed decimal and scaled decimal using floating point thenu >>you should say so. > ) > Well, yes, that is certainly the point.n  D Then you shouldn't have mouthed off about closure.  It's irrelevant.  C Are we in agreement that 1.0 + 1.0 = 2.0 exactly in floating point? C Are we in agreement that 2.0 - 1.0 = 1.0 exactly in floating point? C Are we in agreement that 2.0 * 2.0 = 4.0 exactly in floating point?mC Are we in agreement that 4.0 / 2.0 = 2.0 exactly in floating point?d  E Is your whole point that 3.0 / 2.0 = 1.0 or 2.0 depending on roundoff.E and that decimal string arithmetic prescribes one answer and floatings point the other?  > If so, why didn't you say so instead of talking about closure.  C > Not sure what you mean by "scaled decimal in flaoting point" thisP= > is not a data type in any language with which I am familiarS  A If you had read the original post instead of mouthing off withoutW@ having done so, you would have seen that I never claimed it was.  G > scaled decimal is a bonafide data type in PL/I and it is not a float..   That's right.   @ >>Yes, everybody and his brother knows that you can't just stick> >>dollars and decimal fractions into a floating point variable3 >>and go.  $1.01 is a repeating fraction in binary.  >>C >>But what you and most of the other ignorant dweebs out there seem-A >>not to be able to fathom is that you _CAN_ stick pennies into a2B >>floating point variable and get perfectly accurate computations. >>E >>I'm sure you're not really an ignorant dweeb.  Please don't take me-
 >>for one.  A I take the above comment back.  You ARE really an ignorant dweeb.-   	John Briggs   ------------------------------  % Date: Thu, 24 Jul 2003 14:21:45 -0400 * From: "Stanley F. Quayle" <stan@stanq.com>+ Subject: Re: Packed decimal arithmetic in CL- Message-ID: <3F1FEB79.1809.3932DCA@localhost>4  + On 24 Jul 2003 at 17:22, John Reagan wrote:DH > Not to speak directly for John, but I often tell people that they can H > use H_ or X_floating for currency if they track the smallest currency J > amount.  (ie, pennies here in the US).  Given that the X_float mantissa G > has 33 decimal digits of precision, it gives you even more room than  M > using 64-bit integers which only has around 20 decimal digits of precision.   C Since 0.2 can't be expressed precisely in either of those floating-2> point formats, won't there be a rounding error after repeated  operations?0  
 --Stan QuayleI Quayle Consulting Inc.  
 ----------C Stanley F. Quayle, P.E. N8SQ  +1 614-868-1363  Fax: +1 614 868-1671r1 8572 North Spring Ct. NW, Pickerington, OH  431471= Preferred address:  stan@stanq.com       http://www.stanq.comO   ------------------------------    Date: 24 Jul 2003 13:29:22 -0500 From: briggs@encompasserve.org+ Subject: Re: Packed decimal arithmetic in C 3 Message-ID: <aF78G6gcF5s3@eisner.encompasserve.org>2  Y In article <4BUTa.817$wk4.472@news.cpqcorp.net>, John Reagan <john.reagan@hp.com> writes:I > Tom Linden wrote:  >  > Secondly, what do you mean by O >> integer-valued floating point?  If you are referring to a float in the senseaL >> of, say, Fortran, then you can approximate the value of a decimal, but inN >> general, not completely represent it.  I refer to this as a float binary toG >> distinguish it from float decimal, both supported data types in PL/Ii > H > Not to speak directly for John, but I often tell people that they can H > use H_ or X_floating for currency if they track the smallest currency J > amount.  (ie, pennies here in the US).  Given that the X_float mantissa G > has 33 decimal digits of precision, it gives you even more room than pM > using 64-bit integers which only has around 20 decimal digits of precision.i  B Yes.  That is exactly what I had in mind.  In the implementation I( spoke of, I used D floating and pennies.   	John Briggs   ------------------------------    Date: 24 Jul 2003 13:47:14 -0500 From: briggs@encompasserve.org+ Subject: Re: Packed decimal arithmetic in Co3 Message-ID: <Ty1iTuZ3NfdN@eisner.encompasserve.org>e  Z In article <3F1FEB79.1809.3932DCA@localhost>, "Stanley F. Quayle" <stan@stanq.com> writes:- > On 24 Jul 2003 at 17:22, John Reagan wrote:SI >> Not to speak directly for John, but I often tell people that they can  I >> use H_ or X_floating for currency if they track the smallest currency  K >> amount.  (ie, pennies here in the US).  Given that the X_float mantissa  H >> has 33 decimal digits of precision, it gives you even more room than N >> using 64-bit integers which only has around 20 decimal digits of precision. > E > Since 0.2 can't be expressed precisely in either of those floating- @ > point formats, won't there be a rounding error after repeated 
 > operations?   	 0.2 what?   E 0.2 pennies doesn't need to be expressed precisely.  We are operating  on whole pennies by assumption.l  2 0.2 dollars is expressed as 0.20E+2 = 200 pennies.@ (and you may want to take proper care with your I/O routines.  I: think that most will read 0.20E+2 and get a floating point> value exactly equal to two hundred, but you ought to be sure).   	John Briggs   ------------------------------  % Date: Thu, 24 Jul 2003 18:24:03 -0500S/ From: Chris Scheers <chris@applied-synergy.com> + Subject: Re: Packed decimal arithmetic in C13 Message-ID: <3F206A93.27CF7D9B@applied-synergy.com>    briggs@encompasserve.org wrote:0 > X > In article <3F1EE1FB.FE8CBFC4@istop.com>, JF Mezei <jfmezei.spamnot@istop.com> writes: > > John Reagan wrote:L > >> If they were only that smart...  I've seen dozens of examples of peopleL > >> using binary floating to track currency (and each one is broken).  I'veJ > >> even seen them in languages that provided scaled integers like BASIC. > >VO > > Interesting since there are often standards on exactly how many decimals to3J > > use when specifying currency and conversion factors, as in the case ofP > > international SWIFT transactions. With floating point, you can't specify howE > > many decimals to consider during arithmetic, only during display.2 > D > Ahhh.  The point being that floating point considers _more_ digitsC > in some cases than the standards specify and thus gets the answer  > "wrong" for that reason? >  > Interesting. > @ > So if you were going to use floating point correctly in such aG > case, one appropriate approach would be to use it to implement scaled C > integers (make sure your integers fit into the mantissa, round to @ > nearest integer after multiplication or division and track the( > decimal scale factor by hand).  Right? > E > I once used double precision floating point in exactly that manner. H > It made sense since I had compiler and hardware support for D floatingC > arithmetic but no compiler or hardware support for 64 bit integerl? > arithmetic.  32 bit integers were too small, but the mantissah= > on 64 bit floats was big enough for my purposes.  If I werenG > re-implementing today, I'd be using 64 bit native integer arithmetic.o >  >         John Briggsi    B To try to summarize the issues:  (Hopefully I will get this mostly3 right.  I apologize if I am rehashing the obvious.)n  > For now, we are only considering financial applications, i.e.,; applications that track currency and need to do so exactly.t   The first problem is exactness:   H 1) An integer is not appropriate for tracking currency as dollars, since7 an amount such as $1.35 can not be represented exactly.h  C 2) A scaled integer consists of two values: the value and where the G decimal point is (the scale).  This is what you are used to seeing on aMF calculator display.  A scaled integer represents $1.35 as 135s2, i.e.,F 135 with the decimal point shifted left 2 (between the 1 and 3).  ThisH is also known as tracking pennies.  Sometimes a scale larger than two isG required.  Quite often, compound interest calculations require a largertC scale.  Also, different currencies may have different requirements.m  4 A scaled integer meets the requirement of exactness.  G 3) A floating point number is not appropriate for currency calculations H because a floating point number can not exactly represent 1.35.  GettingF technical, the problem is that the base used in floating point numbersE is a power of two whereas currency calculations need a base that is a E power of ten.  In base 2, 1/10 is a repeating fraction and can not bee represented accurately.r  C A floating point number does not meet the requirement of exactness.u  E 3) Comparing floating point and scaled integer, it may become obviousvG that a scaled integer is really a floating point number that is base 10tF instead of base 2.  They are really implementations of the same thing,D but scaled integers are more appropriate for financial applications.   The second problem is range:  B 1) Using a scaled integers, some calculations can exceed the range available in 32 bits.1  B 2) Decimal arithmetic is one solution to this.  Decimal arithmetic@ tracks the number as a string of decimal digits.  (Again, like a> calculator.)  Decimal numbers are scaled integers, but are notF restricted to 32 bits.  Generally, a decimal arithmetic implementation" allows up to (at least) 31 digits.  H 3) Floating point is another solution to this problem.  A floating pointH number is implemented as some number of bits for the power and some bitsE for the mantissa.  The mantissa is the numeric part of the value.  IfbC the power is held at zero, the mantissa is the integer value of the C number.  (Ignoring normalization.)  Basically, you use the floatinge  point number to hold an integer.  C What's the point to this?  Well, a 64 bit or 128 bit floating point H value has more than 32 bits of mantissa.  This allows the floating pointF value to hold an integer larger than 32 bits.  You can also do math onD this number with floating point instructions.  Just manage the powerC appropriately.  If this is done correctly, the floating point value1D becomes exact for our purposes.  The extra bits give us the range weH need.  You still need to maintain an external scale.  You have created aF scaled integer with the integer part stored in a floating point value.     Summary:  G Either an extended precision floating point value or a decimal variable.A can be used to provide a scaled integer with more than 32 bits ofl
 precision.  ; COBOL and PL/I have built in support for decimal variables.$  H The Alpha does not have hardware support for decimal variables, but doesF have hardware support for extended precision floating point values, soE code can be more efficient.  But code libraries need to be written to>8 manipulate the floating point values as scaled integers.   Summary #2:,  F Native 64 bit integer support in the Alpha makes much of the preceding- discussion moot, if you are writing new code.r   Let the discussion continue!  G -----------------------------------------------------------------------n$ Chris Scheers, Applied Synergy, Inc.  C Voice: 817-237-3360            Internet: chris@applied-synergy.com s   Fax: 817-237-3074c   ------------------------------  % Date: Thu, 24 Jul 2003 20:45:30 -0400e( From: David Froble <davef@tsoft-inc.com>+ Subject: Re: Packed decimal arithmetic in C>, Message-ID: <3F207DAA.3040308@tsoft-inc.com>   Chris Scheers wrote:  ! > briggs@encompasserve.org wrote:  > X >>In article <3F1EE1FB.FE8CBFC4@istop.com>, JF Mezei <jfmezei.spamnot@istop.com> writes: >> >>>John Reagan wrote:h >>>nK >>>>If they were only that smart...  I've seen dozens of examples of peopletK >>>>using binary floating to track currency (and each one is broken).  I've I >>>>even seen them in languages that provided scaled integers like BASIC.t >>>>N >>>Interesting since there are often standards on exactly how many decimals toI >>>use when specifying currency and conversion factors, as in the case of O >>>international SWIFT transactions. With floating point, you can't specify howbD >>>many decimals to consider during arithmetic, only during display. >>>oD >>Ahhh.  The point being that floating point considers _more_ digitsC >>in some cases than the standards specify and thus gets the answert >>"wrong" for that reason? >> >>Interesting. >>@ >>So if you were going to use floating point correctly in such aG >>case, one appropriate approach would be to use it to implement scaledaC >>integers (make sure your integers fit into the mantissa, round to @ >>nearest integer after multiplication or division and track the( >>decimal scale factor by hand).  Right? >>E >>I once used double precision floating point in exactly that manner.dH >>It made sense since I had compiler and hardware support for D floatingC >>arithmetic but no compiler or hardware support for 64 bit integerj? >>arithmetic.  32 bit integers were too small, but the mantissa:= >>on 64 bit floats was big enough for my purposes.  If I were G >>re-implementing today, I'd be using 64 bit native integer arithmetic.v >> >>        John Briggse >> >  > D > To try to summarize the issues:  (Hopefully I will get this mostly5 > right.  I apologize if I am rehashing the obvious.)m > @ > For now, we are only considering financial applications, i.e.,= > applications that track currency and need to do so exactly.e > ! > The first problem is exactness:  > J > 1) An integer is not appropriate for tracking currency as dollars, since9 > an amount such as $1.35 can not be represented exactly.r > E > 2) A scaled integer consists of two values: the value and where theaI > decimal point is (the scale).  This is what you are used to seeing on arH > calculator display.  A scaled integer represents $1.35 as 135s2, i.e.,H > 135 with the decimal point shifted left 2 (between the 1 and 3).  ThisJ > is also known as tracking pennies.  Sometimes a scale larger than two isI > required.  Quite often, compound interest calculations require a larger E > scale.  Also, different currencies may have different requirements.o > 6 > A scaled integer meets the requirement of exactness. > I > 3) A floating point number is not appropriate for currency calculationseJ > because a floating point number can not exactly represent 1.35.  GettingH > technical, the problem is that the base used in floating point numbersG > is a power of two whereas currency calculations need a base that is a G > power of ten.  In base 2, 1/10 is a repeating fraction and can not be  > represented accurately.  > E > A floating point number does not meet the requirement of exactness.P > G > 3) Comparing floating point and scaled integer, it may become obviousSI > that a scaled integer is really a floating point number that is base 10DH > instead of base 2.  They are really implementations of the same thing,F > but scaled integers are more appropriate for financial applications. >  > The second problem is range: > D > 1) Using a scaled integers, some calculations can exceed the range > available in 32 bits.O > D > 2) Decimal arithmetic is one solution to this.  Decimal arithmeticB > tracks the number as a string of decimal digits.  (Again, like a@ > calculator.)  Decimal numbers are scaled integers, but are notH > restricted to 32 bits.  Generally, a decimal arithmetic implementation$ > allows up to (at least) 31 digits. > J > 3) Floating point is another solution to this problem.  A floating pointJ > number is implemented as some number of bits for the power and some bitsG > for the mantissa.  The mantissa is the numeric part of the value.  If E > the power is held at zero, the mantissa is the integer value of thenE > number.  (Ignoring normalization.)  Basically, you use the floatinga" > point number to hold an integer. > E > What's the point to this?  Well, a 64 bit or 128 bit floating point J > value has more than 32 bits of mantissa.  This allows the floating pointH > value to hold an integer larger than 32 bits.  You can also do math onF > this number with floating point instructions.  Just manage the powerE > appropriately.  If this is done correctly, the floating point valuenF > becomes exact for our purposes.  The extra bits give us the range weJ > need.  You still need to maintain an external scale.  You have created aH > scaled integer with the integer part stored in a floating point value. >  > 
 > Summary: > I > Either an extended precision floating point value or a decimal variable C > can be used to provide a scaled integer with more than 32 bits ofa > precision. > = > COBOL and PL/I have built in support for decimal variables.g > J > The Alpha does not have hardware support for decimal variables, but doesH > have hardware support for extended precision floating point values, soG > code can be more efficient.  But code libraries need to be written to : > manipulate the floating point values as scaled integers. > 
 > Summary #2:, > H > Native 64 bit integer support in the Alpha makes much of the preceding/ > discussion moot, if you are writing new code.i >  > Let the discussion continue!     Ok.s  O I've found double precision floating point to be entirely adequate for storing QM monetary amounts.  I'll note that I've only used it for US$.  I'll also note 6N that $1,000,000.00 is a rather large monetary amount for any of the work I've P done, and $100 million is more than ever required.  I'll concede that there are P places where much more magnitude is required, while still requiring preciseness 
 to the penny.n  J As for calculations, that's a bit different.  To do so, a programmer must P routinely perform a rounding calculation after, or as part of, any calculation, I or group of calculations.  Sure, there are values that cannot exactly be oP represented in a floating point number, but, what's the difference between 1.35 N and 1.34999999999999?  If an amount is added to this value, and the result is J rounded to 2 digits after the decimal point, that's adequate for counting I dollars and cents.  Multiply it by 100, and still, what's the meaningful IJ difference between 134.99999999999 and 135?  Once rounded, it becomes 135.     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, 24 Jul 2003 23:34:56 -0400 * From: JF Mezei <jfmezei.spamnot@istop.com>+ Subject: Re: Packed decimal arithmetic in Ct) Message-ID: <3F20A53C.CA1A0DEC@istop.com>k   David Froble wrote:hJ > or group of calculations.  Sure, there are values that cannot exactly beQ > represented in a floating point number, but, what's the difference between 1.35E > and 1.34999999999999?   L If a contract specifies a unit price of 1.35 in a specific currency, but theM supplier, due to rounding errors, charges 1.351 when calculating the cost forAE selling 200,000 widgets, that makes a $200 difference in the invoice.e  N Also consider those "urban legend" cases of bank employees diverting sub-pennyM amounts from transactions to their own accounts and collecting tons of money. M This is one reason one would require full specification of monetary precision # when handling other people's money.r  M If your amounts are big enough that they do not fit in a 32 bit integer, then-C the precision *IS* important because any error will be significant.u   ------------------------------  % Date: Thu, 24 Jul 2003 21:50:05 -0700u# From: "Tom Linden" <tom@kednos.com>w+ Subject: RE: Packed decimal arithmetic in Ca9 Message-ID: <CIEJLCMNHNNDLLOOGNJIAEDFHKAA.tom@kednos.com>H   >-----Original Message-----s7 >From: Chris Scheers [mailto:chris@applied-synergy.com]h& >Sent: Thursday, July 24, 2003 4:24 PM >To: Info-VAX@Mvb.Saic.Com, >Subject: Re: Packed decimal arithmetic in C >i >t  >briggs@encompasserve.org wrote: >>5 >> In article <3F1EE1FB.FE8CBFC4@istop.com>, JF MezeiK$ ><jfmezei.spamnot@istop.com> writes: >> > John Reagan wrote:NC >> >> If they were only that smart...  I've seen dozens of examples 
 >of people> >> >> using binary floating to track currency (and each one is >broken).  I've:K >> >> even seen them in languages that provided scaled integers like BASIC.I >> >? >> > Interesting since there are often standards on exactly howA >many decimals tosK >> > use when specifying currency and conversion factors, as in the case oft? >> > international SWIFT transactions. With floating point, youo >can't specify howF >> > many decimals to consider during arithmetic, only during display. >>E >> Ahhh.  The point being that floating point considers _more_ digitsaD >> in some cases than the standards specify and thus gets the answer >> "wrong" for that reason?  >> >> Interesting.t >>A >> So if you were going to use floating point correctly in such afH >> case, one appropriate approach would be to use it to implement scaledD >> integers (make sure your integers fit into the mantissa, round toA >> nearest integer after multiplication or division and track the ) >> decimal scale factor by hand).  Right?r >>F >> I once used double precision floating point in exactly that manner.I >> It made sense since I had compiler and hardware support for D floatingdD >> arithmetic but no compiler or hardware support for 64 bit integer@ >> arithmetic.  32 bit integers were too small, but the mantissa> >> on 64 bit floats was big enough for my purposes.  If I wereH >> re-implementing today, I'd be using 64 bit native integer arithmetic. >> >>         John Briggs >- >-C >To try to summarize the issues:  (Hopefully I will get this mostly04 >right.  I apologize if I am rehashing the obvious.) >>? >For now, we are only considering financial applications, i.e.,7< >applications that track currency and need to do so exactly. >   >The first problem is exactness: >sI >1) An integer is not appropriate for tracking currency as dollars, sincef8 >an amount such as $1.35 can not be represented exactly. >fD >2) A scaled integer consists of two values: the value and where theH >decimal point is (the scale).  This is what you are used to seeing on aG >calculator display.  A scaled integer represents $1.35 as 135s2, i.e.,tG >135 with the decimal point shifted left 2 (between the 1 and 3).  This I >is also known as tracking pennies.  Sometimes a scale larger than two is H >required.  Quite often, compound interest calculations require a largerD >scale.  Also, different currencies may have different requirements. >s5 >A scaled integer meets the requirement of exactness.a >1H >3) A floating point number is not appropriate for currency calculationsI >because a floating point number can not exactly represent 1.35.  Getting G >technical, the problem is that the base used in floating point numberseF >is a power of two whereas currency calculations need a base that is aF >power of ten.  In base 2, 1/10 is a repeating fraction and can not be >represented accurately. >eD >A floating point number does not meet the requirement of exactness. >yF >3) Comparing floating point and scaled integer, it may become obviousH >that a scaled integer is really a floating point number that is base 10G >instead of base 2.  They are really implementations of the same thing, E >but scaled integers are more appropriate for financial applications.i >  >The second problem is range:s >dC >1) Using a scaled integers, some calculations can exceed the range  >available in 32 bits. >nC >2) Decimal arithmetic is one solution to this.  Decimal arithmeticoA >tracks the number as a string of decimal digits.  (Again, like ao? >calculator.)  Decimal numbers are scaled integers, but are nottG >restricted to 32 bits.  Generally, a decimal arithmetic implementationl# >allows up to (at least) 31 digits.  > I >3) Floating point is another solution to this problem.  A floating pointnI >number is implemented as some number of bits for the power and some bitseF >for the mantissa.  The mantissa is the numeric part of the value.  IfD >the power is held at zero, the mantissa is the integer value of theD >number.  (Ignoring normalization.)  Basically, you use the floating! >point number to hold an integer.- >-D >What's the point to this?  Well, a 64 bit or 128 bit floating pointI >value has more than 32 bits of mantissa.  This allows the floating point.G >value to hold an integer larger than 32 bits.  You can also do math ontE >this number with floating point instructions.  Just manage the power D >appropriately.  If this is done correctly, the floating point valueE >becomes exact for our purposes.  The extra bits give us the range weiI >need.  You still need to maintain an external scale.  You have created aaG >scaled integer with the integer part stored in a floating point value.l >e > 	 >Summary:u >sH >Either an extended precision floating point value or a decimal variableB >can be used to provide a scaled integer with more than 32 bits of >precision.  >s< >COBOL and PL/I have built in support for decimal variables. > I >The Alpha does not have hardware support for decimal variables, but doesbG >have hardware support for extended precision floating point values, sobF >code can be more efficient.  But code libraries need to be written to9 >manipulate the floating point values as scaled integers.  >  >Summary #2: >oG >Native 64 bit integer support in the Alpha makes much of the precedingO. >discussion moot, if you are writing new code.  K This gives approx 19 digits of precision, which ought to be enough for mostuF bank accounts!  Currency trading requires 5 significant digits (called pips),I thus for example USD may be express as CHF 1.3405, to 1/100 of a centime.o TheaK drawback to this approach is that you have to roll your own in handling thet? numbers.  Sometimes you want to limit the precison, for exampleuL declare X fixed decimal (12,4) and use the machinery built into the language toG automatically detect overflow, invoking exception handling routines fore errortH recovery.  And yes you can do this without this machinery, but it is far moreL cumbersome, I think this is the point Mezei was making about controlling the arithmetic vs the display.   Good Summary, Chris.   >x >Let the discussion continue!w >eH >-----------------------------------------------------------------------% >Chris Scheers, Applied Synergy, Inc.e > C >Voice: 817-237-3360            Internet: chris@applied-synergy.comu >  Fax: 817-237-3074 >e >---' >Incoming mail is certified Virus Free.-; >Checked by AVG anti-virus system (http://www.grisoft.com). A >Version: 6.0.502 / Virus Database: 300 - Release Date: 7/18/2003  >  ---o& Outgoing mail is certified Virus Free.: Checked by AVG anti-virus system (http://www.grisoft.com).@ Version: 6.0.502 / Virus Database: 300 - Release Date: 7/18/2003   ------------------------------  % Date: Thu, 24 Jul 2003 13:24:45 -0700h# From: "Tom Linden" <tom@kednos.com>- Subject: PWS500au Boot failing9 Message-ID: <CIEJLCMNHNNDLLOOGNJIAECHHKAA.tom@kednos.com>s  J Found one of these down this morning.  Tried rebooting (5 times) it but it always. failed at the same point in the boot sequence,+ $LICENSE START /DATABASE=LMF$LICENSE /NOLOG  so it wasn't a random HW error.O  H I was able to attach the drives to a spare box and boot without problem.K Anybody experience similar, or have any ideas as to the likely culprit that= needs to be replaced?    TomE   ---n& Outgoing mail is certified Virus Free.: Checked by AVG anti-virus system (http://www.grisoft.com).@ Version: 6.0.502 / Virus Database: 300 - Release Date: 7/18/2003   ------------------------------  % Date: Thu, 24 Jul 2003 13:30:46 -07009# From: "Tom Linden" <tom@kednos.com>t" Subject: RE: PWS500au Boot failing9 Message-ID: <CIEJLCMNHNNDLLOOGNJIGECIHKAA.tom@kednos.com>^  K Should have added that on the boot screen it gave a bugcheck when executingE% process sda.exe with code 000215 IIRC      >-----Original Message-----a) >From: Tom Linden [mailto:tom@kednos.com]t& >Sent: Thursday, July 24, 2003 1:25 PM >To: Info-VAX@Mvb.Saic.Com >Subject: PWS500au Boot failingl >s >nK >Found one of these down this morning.  Tried rebooting (5 times) it but it  >alwayse/ >failed at the same point in the boot sequence,f, >$LICENSE START /DATABASE=LMF$LICENSE /NOLOG  >so it wasn't a random HW error. > I >I was able to attach the drives to a spare box and boot without problem. L >Anybody experience similar, or have any ideas as to the likely culprit that >needs to be replaced? >p >Tom >o >---' >Outgoing mail is certified Virus Free.s; >Checked by AVG anti-virus system (http://www.grisoft.com). A >Version: 6.0.502 / Virus Database: 300 - Release Date: 7/18/2003m >  >---' >Incoming mail is certified Virus Free.e; >Checked by AVG anti-virus system (http://www.grisoft.com).hA >Version: 6.0.502 / Virus Database: 300 - Release Date: 7/18/2003  >u --- & Outgoing mail is certified Virus Free.: Checked by AVG anti-virus system (http://www.grisoft.com).@ Version: 6.0.502 / Virus Database: 300 - Release Date: 7/18/2003   ------------------------------  # Date: Thu, 24 Jul 2003 21:38:40 GMTw( From: David Harrold <DHarrold@wi.rr.com>" Subject: Re: PWS500au Boot failing@ Message-ID: <966502b9d34394c55ef33fe22b8d9d4a@free.teranews.com>  H On Thu, 24 Jul 2003 13:30:46 -0700, "Tom Linden" <tom@kednos.com> wrote:   Hi Tom,e  L >Should have added that on the boot screen it gave a bugcheck when executing& >process sda.exe with code 000215 IIRC >" >o  L Just looked as some console logs and it looks like the 215 code is a machineK check bugcheck.  If that is correct, then you have something broken in your2E system.  Memory or CPU are the ones I usually see with this bugcheck.a   Hope that helps,   Dave Harrold    N ..............................................................................N David Harrold                              E-Mail: David_Harrold at aurora.orgI Sr. Software Systems Engineer              Phone:          (414) 647-62040I                                            Pager:          (414) 941-4634 G Aurora Health Care                         Fax:          (414) 647-4999> 3031 W. Montana Street Milwaukee, WI 53215r   ------------------------------  % Date: Fri, 25 Jul 2003 01:12:53 -0400'5 From: "Brad McCusker" <brad.mccuskerNosp@Mcompaq.com>bL Subject: Re: Questions regarding OVMS 7.3-1 and Pathworks v6.0 compatibility/ Message-ID: <vi1f2qph985b5d@corp.supernews.com>n  / "PEN" <paul.nuneznosp@mhp.com> wrote in message & news:bfopm7$lgk$1@hplms2.hpl.hp.com... > Hi,  >>H > Just want to clarify one statement made by Brad in the previous reply: >nL > >> The "new" PATHWORKS client is PATHWORKS 32 V7.3 and it provides support > for  > >> Windows 2003. > K > PATHWORKS 32 v7.3 provides support for Windows XP (Pro and Home) systems,>" > not Windows 2003 Server systems.  G Nice catch Paul.  Not sure what I was thinking about when I typed that.y Sorry for any confusion.   Brad   ------------------------------  % Date: Thu, 24 Jul 2003 13:38:21 -0400e* From: JF Mezei <jfmezei.spamnot@istop.com>, Subject: Re: Sun Micro Profit, Revenues Fall) Message-ID: <3F20198C.7AAA955D@istop.com>h   Bill Todd wrote:N > Compaq gave up Alpha because it didn't want to be in the processor business.  K I am not sure of this. I think it is far more because Compaq didn't want toe= jeoperdize its relationship with Intel that it killed Alpha.    N It seems to me that Digital and Compaq both sunk because they decided to focusJ on their competitor's products, cannabalising their own to maintain a good# relationship with their competitor.i   ------------------------------  % Date: Thu, 24 Jul 2003 13:55:50 -0400t* From: JF Mezei <jfmezei.spamnot@istop.com>, Subject: Re: Sun Micro Profit, Revenues Fall) Message-ID: <3F201DA4.C14A0038@istop.com>o   Fred Kleinsorge wrote:K > The amusing thing is that you can replace "Sun" in the above with DEC/VAX.* > and you pretty much have the same story.  J Except Sun isn't affraid to market its own products, and Sun still has itsG independance and is not remotely managed by Bill Gates and Andy Groves.i  J > Sparc has been running on vapours for a long time now, much like the VAXI > did.  The question is if Sun can somehow make the breakthrough in Sparcs7 > performance that will leapfrog it back into the game,e  M Funny how not too long ago, there was a discussion on just how fast VAX couldiM have been made if DEC had stuck to it. DEC's problems weren't that VAX wasn'teN the fastest in the world, it was that DEC refused to price VAXes competitively with its new competitors.s  K When you consider the pretty amazing work Intel was able to do to transfordrN its 8086 toy controller into a respectable chip that will outlive both VAX andM Alpha, I really don't see why you would dismiss SPARC's future so readily. IftL the architecture is sound and easy to upgrade, then sparc might even outlive	 the IA64.o  D Just because Alpha died of a premeditated murder and VAX died due to? mismanagement doesn't mean that its peers will have to die too.i  E Let me ask you this: if you had to place a bet, would you bet on IA64g' outliving the 8086 industry standard  ?l  I > into the sunset.  Sun needs to bite the bullet and come up with a Sparcp3 > transition plan.  Itanium would be the way to go,   J At this point in time, Itanium is just marketing gobledeegook.  Only a fewI vendors pretend that IA64 is industry standard etc etc etc.  If/when IA64eM proves its worth, then perhaps other vendors can consider it. But for as longfK as the 8086 remains the industry standard, there is no reason for anyone tom bet their business on IA64.   N There is nobody left to save Digital. Compaq died, and if HP fails, it will be& Sun against IBM for serious computing.  ( > Sun is conflicted.  Solaris vs Linux.   K HP is even more conflicted. HP-UX vs Linux vs Windows. And consider Carly'soM relationship with Gates and Groves and one realises that HP has a far greateroM conflict to deal with.  HP could never dare start selling Linux desktops in ahT large volume at low prices because it woudl jeoperdize its position with Bill Gates.  K And HP is stuck in a lengthy transition period where it must cater to Tru64eS customers by integrating those bits into HP-UX *and* manage the transition to IA64.P   ------------------------------    Date: 24 Jul 2003 16:04:41 -0700( From: bob@instantwhip.com (Bob Ceculski), Subject: Re: Sun Micro Profit, Revenues Fall= Message-ID: <d7791aa1.0307241504.4991dd9d@posting.google.com>   n "Rudolf Wingert c/o WAT017" <win@fom.fgan.de> wrote in message news:<000001c351ac$612252d0$114614ac@wat017>... > Hello, > F > how do Sun get top? Here in Germany they do sell two servers fort heJ > price of one. What do you do, if you have to buy a Sun? Yes you will buyF > two (the same price). In case of this Sun will become the top serverG > seller and the customers a lot of Suns just for fun, not real needed.  >  > Best regards Rudolf Wingert   ( you need two suns to equal one alpha ...   ------------------------------    Date: 24 Jul 2003 14:52:30 -0700, From: JimStrehlow@data911.com (Jim Strehlow) Subject: Re: Sybase client= Message-ID: <4b6ec350.0307241352.154d338b@posting.google.com>y  D dm <dimez@gmx.ch> wrote in message news:<3F1F73F9.8070700@gmx.ch>...# > Thank you, this brings some hope!N$ ...running the old UCX , Ver 4.2-21)! > so, updating the IP could help?a >  > thank you and regards. > Dieter Meier  C We have been running Sybase version 11 on OpenVMS v6.x through v7.3sE with tcp/ip stacks from MultiNet to Compaq TCP/IP Services for years.g  & We currently have running on one node:   $ ISQL > select @@version > goC  SQL Server/11.0.3.3/P/DEC AXP/OpenVMS V6.2/SWR 8706/OPT/3-NOV-1999  17:47:34.58    $ tcpip show version?   Compaq TCP/IP Services for OpenVMS Alpha Version V5.1 - ECO 4e6   on a AlphaServer DS20 500 MHz running OpenVMS V7.3    A I spend most of my time on Oracle 8i and 9i work; but we maintain D Sybase for legacy customers and for converting from Sybase to Oracle
 planning etc.m  & Jim Strehlow, Systems Manager, Data911 Alameda, CA, USA   ------------------------------  # Date: Fri, 25 Jul 2003 04:07:54 GMTs From: Rob.Buxton@wcc.govt.nz Subject: Re: Sybase client% Message-ID: <3f20ac00.348430468@news>e  < On Thu, 24 Jul 2003 07:51:53 +0200, dm <dimez@gmx.ch> wrote:   >Thank you,i >n >this brings some hope!r > ! >we have the following situation:c >nG >after a query with the ISQL command and having onbly a IP connection, h >nothing is happening! >m; >after 2 minites (timeout) we get following error messages:J >  >g" >         No such device available >and! >         Unable to create socketm >r4 >the creation of the IP device (socket) 5001 failed!J >there must be some incompatible shared libraries or problems with the IP  >stack >F" >(actual running: TCP-IP V5.1-15 ) >F2 >the sybase client seems to be the same Ver 10.0.3 > ! >but you TCP is a higher version!t8 >have you modified any TCP settings (buffer size or ...) >c@ >(I have to mention, that the same program is running on an old * >standalone machine, )AXP, running the old >UCX , Ver 4.2-21) >a  >so, updating the IP could help? >e >thank you and regards
 >Dieter Meier  >n >r3 Our version of the Sybase Client seems to be 10.0.1   0 I'd check that your Interfaces file is correct. D A 2 minute time out seems like it cannot connect to the target host.  B Maybe post a portion of the interfaces file and I'll compare that.  A Someone else has mentioned the inet driver, but the Sybase ClienteA startup  should start that up. Although I recall problems when wel moved from UCX 4 to TCPIP 5.       >e >Rob.Buxton@wcc.govt.nz wrote: >r> >>On Wed, 23 Jul 2003 12:03:45 +0200, dm <dimez@gmx.ch> wrote: >> >>   >>L >>>Has any body a running sybase client (license locally installed) for VMS J >>>7.1-2 (or higher) which can query with the IP stack (TCP-IP V5.1-15 or  >>>higher) ?6 >>>(Decnet is working, but on PC only TCP is working!) >>>:  >>>thanks for any ideas or helps >>>e >>>Dieter Meier9 >>>F >>>a >>>    n >>>cE >>Yes we're running Sybase Open Client 10.0 (something) under OpenVMS83 >>7.2 (VAX) and 7.2-2 (Alpha) with TCPIP 5.3 ECO 2.> >>1 >>Is there a problem you're not telling us about?y >>% >>We only use TCPIP for connectivity.s >>   >> >    ------------------------------  % Date: Fri, 25 Jul 2003 10:17:17 +1200 / From: Martin Hunt <martin.hunt@fairfaxnz.co.nz>a( Subject: TFTP timeout on tcp/ip services8 Message-ID: <ibm0ivotkabos3cerofaivsmqcnuei0bqt@4ax.com>  C We recently upgraded to tcpip services V5.3 (from V4.2), and we areaE often getting timeouts transferring files via tftp to the VAX, mainlyo* with larger files (a few thousand blocks).> There is a timeout parameter on the TFTP service, which I haveE increased, but I can't find any documentation for what it does. Like,sE what units does it use - seconds or minutes? Does anyone know if I amg7 actually using this parameter for its intended purpose?e   Martin   ---l Martin Huntc Systems Administratora Fairfax New Zealand Limitedw
 Wellington New Zealande   ------------------------------   Date: 24 Jul 2003 23:55:38 GMT2 From: "Zane H. Healy" <healyzh@shell1.aracnet.com> Subject: Re: The PDP lives+ Message-ID: <bfprlq0g3i@enews3.newsguy.com>   < Bob Koehler <koehler@eisner.nospam.encompasserve.org> wrote:` > In article <OFEE1249C5.1D61F695-ON85256D6D.004D088C@metso.com>, norm.raphael@metso.com writes: >>  & >> ..the TECLA is no vacuum cleaner... >> eD >> Is this a proper device to be named a Programable Data Processor?  D >    Seeing as it uses an "86 Processor" and runs Windows CE I thinkF >    the current owners of the PDP trademark, if there are any, should% >    run their lawyers past this one.m  B The owners would either be HP or Mentec in the case of the PDP-11.  G >    A real Programmable Data Processor was designed by DEC and hasn't n >    been built in a decade.  J Hmmm...  So, a Mentec M11 isn't a real PDP-11 since it wasn't designed andC built by Dec?  Development of the PDP-11 family didn't end with ther@ PDP-11/93 /94, that was just the end of DEC's development of it.   			Zaney   ------------------------------    Date: 24 Jul 2003 14:52:39 -0500+ From: young_r@encompasserve.org (Rob Young)w/ Subject: Re: XFC, IO Size and Backup - Take Twol3 Message-ID: <z2l42RgTYYzX@eisner.encompasserve.org>,  a In article <adMANVMnKPtX@eisner.encompasserve.org>, young_r@encompasserve.org (Rob Young) writes:e    @         Getting back to what size IO for ReadAheads.  Here is a             XFC> show context/brief   (         during a copy of a largish file:  B Address          I/O Type         ID  Start VBN    Length      IRPB FFFFFFFF826FAF90 eiotReadThrough  82    1710337        64 81CB2580B FFFFFFFF8264BBD0 eiotReadAhead    82    3076217       120 82030700B FFFFFFFF826569D0 eiotReadThrough  82    1710401        64 81FFB980B FFFFFFFF81F054D0 eiotReadAhead    82    3076337       120 8202FB80B FFFFFFFF8222C010 eiotReadThrough  82    1710465        64 81B63900B FFFFFFFF82493E50 eiotReadAhead    82    3076457       120 824B3800B FFFFFFFF8254C7D0 eiotReadAhead    82    3076577       120 82030700B FFFFFFFF826D0A50 eiotReadThrough  82    1710529        64 81CB6C80B FFFFFFFF824EFAD0 eiotReadAhead    82    3076697       120 8202FB80B FFFFFFFF822F0810 eiotReadThrough  82    1710593        64 81D6D340B FFFFFFFF82466690 eiotReadAhead    82    3076817       120 82030700  8         XFC uses 120 blocks as a ReadAhead block size.  <         Interestingly, the ReadAheads above are far ahead of:         the blocks being copied.  Also, copy does 32 block<         IO on copy operations, our ReadThroughs are grabbing7         64 blocks at a time - a built-in 1 IOs worth oft0         "readahead".  On our largish IO example:  % List of All XFC Active Contexts (CTX)a% -------------------------------------oE Address          I/O Type        File ID  Start VBN    Length     IRP F FFFFFFFF8378A1D0 eiotReadThrough      48    2231809       512 81CC6A00F FFFFFFFF833C00D0 eiotReadAhead        48    2233345       120 837F01C0F FFFFFFFF8390B510 eiotReadThrough      48    2232321       512 81973D40F FFFFFFFF8397DE50 eiotReadAhead        48    2233465       120 837F01C0F FFFFFFFF836DDFD0 eiotReadThrough      48    2232833       512 81CC6A00F FFFFFFFF836DB290 eiotReadAhead        48    2233585       120 837F01C0F FFFFFFFF836D7C90 eiotReadAhead        48    2233705       120 837F01C0F FFFFFFFF832CC190 eiotReadThrough      48    2233345       512 81CC6A00F FFFFFFFF83848990 eiotReadThrough      48    2233857       512 81973D40F FFFFFFFF837CC510 eiotReadAhead        48    2235393       120 837F01C0F FFFFFFFF837DAF10 eiotReadThrough      48    2234369       512 81973D40F FFFFFFFF837D7CD0 eiotReadAhead        48    2235513       120 837F01C0F FFFFFFFF835F15D0 eiotReadThrough      48    2234881       512 81CC6A00F FFFFFFFF8389E3D0 eiotReadAhead        48    2235633       120 837F01C0F FFFFFFFF832D9FD0 eiotReadAhead        48    2235753       120 837F01C0F FFFFFFFF837892D0 eiotReadThrough      48    2235393       512 81CC6A00F FFFFFFFF839ADBD0 eiotReadThrough      48    2235905       512 81973D40    F         We aren't very far ahead at all on our ReadAheads (compare VBNF         numbers).  Leading me to speculate the ReadThroughs there are A         indeed overrunning the ReadAhead functionality.  But with.J         a copy operation, the ReadAheads are able to stay far ahead of the
         copy.   D         I wonder if it would help to make that ReadAhead buffer size<         dynamic like the ReadThrough buffer size and grow it9         up to 2x the ReadThrough size - to within reason.   C         At 1024 blocks ( 2 * 512 ), you wouldn't overrun a back-end B         on these newer disk subsystems and you can always throttleD         it back if IO response time becomes unbearable or worthless.  @         As it is, that 120 block IO for ReadAheads on large IO -         ain't cutting it.   #                                 Roba   ------------------------------    Date: 24 Jul 2003 15:31:28 -0500+ From: young_r@encompasserve.org (Rob Young)>/ Subject: Re: XFC, IO Size and Backup - Take Twoo3 Message-ID: <aw6wxtVYW3j6@eisner.encompasserve.org>o  a In article <z2l42RgTYYzX@eisner.encompasserve.org>, young_r@encompasserve.org (Rob Young) writes:mc > In article <adMANVMnKPtX@eisner.encompasserve.org>, young_r@encompasserve.org (Rob Young) writes:s >    	[snip]c  H > FFFFFFFF8389E3D0 eiotReadAhead        48    2235633       120 837F01C0H > FFFFFFFF832D9FD0 eiotReadAhead        48    2235753       120 837F01C0H > FFFFFFFF837892D0 eiotReadThrough      48    2235393       512 81CC6A00H > FFFFFFFF839ADBD0 eiotReadThrough      48    2235905       512 81973D40 >  > H >         We aren't very far ahead at all on our ReadAheads (compare VBNH >         numbers).  Leading me to speculate the ReadThroughs there are 9 >         indeed overrunning the ReadAhead functionality.h    ) 	Actually, looking at those last 4 lines:o  G  FFFFFFFF8389E3D0 eiotReadAhead        48    2235633       120 837F01C0tG  FFFFFFFF832D9FD0 eiotReadAhead        48    2235753       120 837F01C0nG  FFFFFFFF837892D0 eiotReadThrough      48    2235393       512 81CC6A00eG  FFFFFFFF839ADBD0 eiotReadThrough      48    2235905       512 81973D40s  ? 	We are ReadAhead VBN 2235753 and yet ReadThrough 2235905 so itm= 	does appear the ReadThroughs are overrunning the ReadAheads.>' 	Unless of course I am misinterpreting!l   				Robm   ------------------------------  % Date: Thu, 24 Jul 2003 19:22:28 -0500>1 From: "David J. Dachtera" <djesys.nospam@fsi.net>nJ Subject: Re: [ANNOUNCEMENT] Migrate obsolete VAX/VMS SW to OpenVMS Itanium' Message-ID: <3F207844.62B0642E@fsi.net>t   Fred Kleinsorge wrote: > I > Funny, sounds like advertising to me.  If you have a list of customers,gM > software, and concerns from those customers - feel free to send this to SuedG > Skonetski who will be able to redirect it to the appropriate customere > advocates attention. > L > I have a top level question:  Are their concerns with IA64 pretty much theL > same concerns as Alpha?  That is, are they still on VAX because they could, > not or were blocked from porting to Alpha?  D I think you'll find that the answers will range from "Yes" to "We'reE still running software that was compiled/linked on V4.x" (or earlier)iC because the ISV disappeared or dropped VMS support, but the app. isa* business critical, so they nurse it along.  G Charon-VAX for OpenVMS-I64 (when/if it appears) may be of some value toa some (but likely not all).  B A VAX -> I64 (direct) binary translator may have an audience also, albeit somewhat limited.   My $0.02....   --   David J. Dachtera  dba DJE Systemsr http://www.djesys.com/  ( Unofficial Affordable OpenVMS Home Page: http://www.djesys.com/vms/soho/    ------------------------------   End of INFO-VAX 2003.407 ************************