https://twitter.com/BrianRoemmele/status/1257832168455208963
6 Likes
Marvellous! He mentions the turnkey system emulator in question, by JĂźrgen Winkelmann:
The MVS 3.8j Tur(n)key System (Tur(n)key Level 4-)
**********************************************************************
* *
* MVS 3.8j Tur(n)key 4- ("TK4-") is a ready to use OS/VS2 MVS 3.8j *
* system built specifically to run under the Hercules System/370, *
* ESA/390, and z/Architecture Emulator. It is an extension of the *
* original MVS Tur(n)key Version 3 System ("TK3") created by Volker *
* Bandke in 2002. See the Userâs Manual for credits and copyrights. *
* *
* Note: TK4- does not claim to be a new release of the original TK3 *
* system, hence its name is TK4-, not TK4. *
* *
**********************************************************************
Where do you get a PI for $ 5?
That I find that hard to believe. A IBM 370 was the same as good 386 PC.Not having windows makes a machine
fast.
I donât think the $5 is the important bit. A pi zero is nominally ÂŁ5 here in the UK, but of course thereâs postage and other overheads. Itâs really not the important point.
What this thread introduces is the existence of a very capable emulator which can run on very cheap hardware.
What it introduces is simply how powerful modern hardware is and, honestly, how little power commercial back room data processing really consumes.
Back then we had more effcient (for the computers) data formats. We used data structures designed to be as efficient as practical on really bad hardware.
All of those efficiencies add up to make legacy software âextraâ fast on modern hardware.
But in the end, itâs still legacy software. Software today is hard enough to use as it is, back then it was even more difficult because much of the constraints of the hardware leaked out to the end user experience. But as a rule, we simply put up with it back then.
Now, of course, we only have as anecdotal evidence that the software actually runs properly with no real data on itâs overall performance. How long does compiling all of the source code take on the Pi compared the mainframe it came from. On a legacy mainframe, it wouldnât surprise me if the Pi was notably faster. But on a modern mainframe, Iâm sure the Pi lags behind. But that doesnât mean its unusable.
But, having replaced many a âminiâ mainframe (no bigger than 3 or 4 washing machines lined up in a row) with machines that were either pizza box sized or mini-towers, itâs always nice to see them reduced to something the size of an index card.
And I thought the RAID USB drives was a nice touch.
Now you can put a 360 in your old washing machine
and emulate a IBM 360 with a Hard Drive. Card Punch/Reader optional at extra $$$. I/O is what ,made the old machines interesting. Does the PI emulate 360/370 microde? A important point for 360/370 emulation.
In the US, MicroCenter retail stores. The Pi Zero W is $5.
oldben wrote:
Does the PI emulate 360/370 microde? A important point for 360/370 emulation.
Well, no (to the former, and no to the latter too).
Hercules is an instruction-level emulator.
There are folks who are working on projects to emulate
System/360 at the microcode (or lower) level â a particular
model in each case (because different models of 360s or
370s were vastly different below the the uniform instruction-set
architecture level).
Lawrence Wilkinson is (or was) one, and Camiel Vanderhoeven
is another. E.g.,
Log In
Camiel Vanderhoeven wrote (on Mar 4, 2021)
( Log In ):
My project is building an IBM 360 model 65 cpu on an FPGA
(well, really, a 9020E cpu) to drive an original 360/65 console
panel. . . [My] hybrid debugging system. . . is Hercules with
the original cpu code in there, but it also has the FPGA code
translated into C. . . Itâs about 10,000 times slower than the
real thing. . . [M]y FPGA-derived CPU behaves well enough to
boot OS/360 MFT all the way to executing the first JCL jobs to
prepare for the OS/360 MVT system generation. The only reason
I havenât gotten any further, is because the software emulation
of the FPGA is so slow that it takes a few days to get there. . .
2 Likes
Great info, thanks Jim!
I read that Twitter thread and I donât think the claim is right. At the end, the author admits running some own small COBOL code on TK4 and then recompiling it and running it without issues on a proper mainframe later. This is certainly possible due to COBOL backwards-compatibility.
Folk asking about running DB2 and other things are right - one might a bit more than pretty limited MVS 3.8j to run a ârealâ application.
I am a big fan of Hercules and what it can do but the claim might be bogus.
Folk asking about running DB2 and other things are right - one might
a bit more than pretty limited MVS 3.8j to run a ârealâ application.
Of course, thatâs not a limitation of Hercules itself, itâs a limitation
of what itâs legal to run on Hercules. MVS 3.8j is legal; OS/390,
z/OS, âDB2 and other thingsâ are not . However, that catâs been
out of the bag for many years now, and widely known to be out of the bag
(presumably by IBM as well).
It used to be the case that mentioning this fact on a Hercules forum
would get you flamed to a crisp; these days, you might just get a
glass of warm water thrown in your face. ;-> The politics surrounding
the very existence of Hercules has been an ongoing drama for more
than two decades now; its users were once a good deal more paranoid
that IBM might exercise its legal and political muscle to have the
emulator expunged from the public internet. Thatâs never happened,
though at one time (and maybe still today) the better part of discretion
has been never to mention the H word at work if you happen to be an
IBM employee (ironic, since â allegedly â many of the most
enthusiastic Hercules users have always been themselves IBMers).
Still, if thatâs something youâre doing (i.e., running something newer
than MVS 3.8j) , itâs probably prudent not to advertise it on
a public forum like Twitter. Multiply so if youâre actually an
IBM employee or customer (or the employee of an IBM customer).
1 Like
The cat is out the bag, thatâs true. I am less concerned about licensing issues here (itâs IBMâs blind policy and problem, not mine).
What wonders me is rather the technical feat of ârunning entireâs bank COBOLâ on Raspberry Pi. I think it is plausible to claim to compile some source code and get it running on MVS. I think it is plausible to claim that some even relatively modern COBOL code compiles on the COB compiler shipping with 3.8j⌠One can run KICKS instead of CICS and play with it.
I somehow doubt that all datasets or volumes have been moved from the bank to someoneâs Raspberry Pi to run âall of itâ there.
But I know of professional mainframe consultants who keep running customerâs stuff on Hercules, but likely not on MVS - those guys are assembler die-hards who even wrote their own database from scratch.
Real die hards, still punch cards. 
What about IBM university software on punch cards?
I can find the fictional book Programing languge algol 69/360 ported the IBM 1130 using a macros
but the source or paper lisings seems long gone.
That is a sad loss of knowlage. Ben.
oldben wrote:
What about IBM university software on punch cards? . . .
I can find the fictional book Programing languge algol 69/360 ported
the IBM 1130 using a macros but the source or paper lisings seems long gone.
That is a sad loss of knowlage. Ben.
Itâs worth keeping in mind when exploring the world of old-computer simulators
that itâs nothing short of a miracle when any old piece of system software
is discovered to have survived for one of these ancient machines, especially as
far back as the batch-monitor era.
As much work as was put into these software constructs, they were thought
of by the people who used them and the companies that sold them as
ephemeral tools, to be discarded and forgotten as soon as the machines
that ran them became obsolete and no longer needed to be supported.
They were certainly not thought of as âcultural artifactsâ in decades past
(just as, perhaps even more scandalously, motion pictures were treated as throwaway
ephemera by the companies that made them, until quite recently).
There are miracles of preservation that turn up from time to time,
but they certainly canât be counted on.
Hereâs a litany of woe, from the retrocomputing trenches:
groups dot google dot com /forum/#!topic/comp.sys.xerox/ONU0QUUgSMc
[Want to find:] Tymshare SuperBasic Source Code
Stephen M. Jones
1/7/17. . .Actually, any software, binary, source for the SDS-940 run by Tymshare.
Rumour has it that all software associated with this system is completely lost,
in fact, âthe packs and tapes destroyedâ. Is this true? . . .Is all these years of work lost forever? If so, that is sad.
Al Kossow
1/7/17On 1/6/17 11:10 PM, Stephen M. Jones wrote:
Is all these years of work lost forever?
Yes, along with much of the other work done by other companies in the
past 50 years. Welcome to my world.Once a computer is obsolete, there is no monetary value to the companies
that made them or the companies that used them, so it was discarded. In
fact, time has demonstrated that leaving behind a paper trail for lawyers
to use in discovery can be a bad thing.What survives mostly comes out of peoples garages that actually cared about
the machines either for nostalgia or maintenance for the few that were kept
running. Be glad, for example, that Tim Litt cared enough about DEC LCGâs
tapes and disks that he had them donated to CHM when HP was going to throw
them out, and why you have the disk packs today.Believe me, Iâve been looking for 30+ years for old software, and there is no magical
archive of tapes that normal people can get to. Rumors exist of them in
support of military computers, but I canât imagine even those will be around
for anything from the 60s today. Outside of Tymshare, the only people I know
of that ran their version of the monitor was Harvard and what I have found
from that are a few newsletters and fragments from teletype printouts.The only possibility is if someone inside Tymshare saved a tape, and Iâve talked
to a LOT of people about it. This was a long time ago, there werenât many people
working on it, and again, after they switched to PDP-10s there wasnât a whole
lot of motivation for them to be interested in the 940s any more.jmfbahciv
1/8/17Wouldnât U of Michigan have had tapes? Sometimes, schools donât throw
away things like companies do.Al Kossow
1/8/17University archives have other problems, fundamentally paranoia about releasing
any personal information about anyone still alive. Apparently, from what I was
told by our archivists, you lose your right to privacy when you die.Stanford, MIT, and CMU have PDP-10 related archives but other than SAILDART, which
isnât part of Stanford special collections AFAIK, little is really available, or
even in a public catalog.The attempts to get 133-tenex from a SUMEX backup tape by the people doing
the github PDP-10 project will be an interesting test casegithub dot com /PDP-10
mentioned in
web dot stanford dot edu /group/htgg/cgi-bin/mediawiki/index.php?title=Adventure
web dot stanford dot edu /group/htgg/cgi-bin/mediawiki/index.php?title=Arcfiles.740418-760326.filteredLast I heard, no one was willing to accept the SAILDART archive because of the
creators restrictions that it not be made public for something like 80 years.Rich Alderson
1/9/17Sometimes, schools donât throw away things like companies do.
Barb, as soon as I was laid off from Stanford in November, 1991, every single
LOTS DEC-20 related magnetic tape went into a great big dumpster behind CERAS.
That was 25 years of quarterly backups, along with distribution tapes from DEC,
SPSS, Software House (System/1022 database), etc. Things were lost which were
never archived on any other Stanford DEC-20.
groups dot google dot com /forum/#!topic/alt.sys.pdp10/J3EkcwrFH5E
Al Kossow
6/21/01Barb, it looks like PDP-10 sofware wasnât the only victim
of the [DEC] Software LibraryâŚJust received this message from Bruce Mitchell as Iâve
been trying to find a copy of RSX-15 for Bobâs [Supnikâs]
simulator efforts:âThe RSX-11 development group
lost all their baseline software circa 1986 when they
sent it off to DEC Software Distribution for archival,
which threw it all away two years later because there
had been no requests for it.âAl Kossow
6/21/01It isnât limited to companies.
Try contacting CMU, Stanford, MIT, UCB, and see how much software
they still have from the 60âs or early 70âs.SAILâs archives go back to 1972.
As has been pointed out previously, CMU has a magtape library, but
they lost the index.Guess it just comes down to the fact that storage space is finite
and no one felt it was important to save any of this.jmfbahciv
6/22/01I take responsibility for some the loss. I didnât think it
was honest to take the stuff home (it was specically against
company policy so I considered that all dishonest). Today
I regret being honest.Mike Ross
6/22/01There are still a few operational PDP-15âs in private hands, they
would be the last hope for finding -15 RSX.Iâve got two of those; regret to report no evidence of RSX anywhere
near them. Both came from defence-related establishments and removal
of magnetic media was out of the question. . .[Thatâs undoubtedly a large part of why Cray software
disappeared so thoroughly. Youâd have to have been James Bond
to get it out of the building.]
groups dot google dot com /forum/#!topic/comp.sys.unisys/pudB5LBYOBs
Exec8 emulator
. . .
Al Kossow
4/22/04âDavid W. Schrothâ [wrote]:
. . .
[W]hile I can lay my hands on Exec sources going back to 38R5, I
wouldnât begin to know where to look for level 33 or level 35 sources.Sigh⌠John Dobnick at the UW-Milw had them as recently as a few
years ago in a tape rack in his office the last time I visited him
before he died. When I asked what had happened to the contents of
his office, it had all been thrown away.This discovery is one of the things that had prompted me the past few
years to intensify my efforts to try to find as much of the software
from the mid-70âs and earlier that still exists. While a LOT of DEC,
CDC, and IBM software survives, I have not heard of much from either
Burroughs or Univac that has⌠survived.
compgroups dot net /comp.sys.unisys/old-copies-of-mcp/1521958
scott1
8/20/2007does anybody know of an archive old Burroughs OS/software?
. . .
The Pasadena plant closed in 1991. In 1989, we were cleaning out the basement
in preparation for moving to the newly built offices on the old manufacturing
floor. We discarded all the old 7-track tapes because we had no way to read
them. I regret that to this day. We did dup the 16mm films to video, and
I have the video (I think the films went to CBI). Amongst the 7-track tapes
were the sources for most of the CP series (pre MCP) operating systems for the
B-[23]xx, B-[234]xxx and I think B-5000.scott [Scott Lurndal]
aek2
8/21/2007Sadly, the only hope now is if someone has saved tapes or listings from an
installation.CHM has a pile of listings, some of which have been scanned, and a set of 7-track
source tapes which were saved from the 5500 system at UC Santa Cruz. There are no
binary tapes, which will make bootstrapping a simulator difficult.[Al Kossow]
groups dot google dot com /forum/#!topic/comp.sys.cdc/t5TInJ2jRnk
Dave Mausner
8/19/03Unless you have a reel containing MACE 5 [a CDC 6600 operating system],
then the last working copy I know of has been lost.It was obtained in May 1970 from MACE author Greg Mansfield, handed to
Northwestern U programmer Larry Atkin, probably in exchange for a copy of
CHESS. It was then conferred upon NU programmer Chuck Filstead, who adapted
small parts of MACE, including its influential coding standard, into NUâs
SCOPE. In 1973 the tape was conferred to me.Then a programmer at Michigan State U, I re-assembled the entire system and
re-built the deadstart for use in research projects. At the time, MACE was
the only general-purpose CDC system that could be configured on-the-fly for
different cpu and storage arrangements.The tape, sealed in its original plastic cannister, and a padded cardboard
travel container, found its way to my fatherâs office building, where he
stored it for twenty-nine years. It continued to exist during the entire
time Tom Hunter and others were searching for early CDC systems to test the
Desktop Cyber project.My father sold the building and in July 2003 the new owner cleared the
storage room. When I mentioned the MACE tape to my dad, he recalled
possessing the container, but alas, the new ownerâs massive trash truck had
been removed to the landfill the previous day.Tom Hunter
8/20/03What a sad waste. You knew about the tape and you knew about many
people hunting for MACE 5. I donât know what to say!Dave Mausner
8/22/03To close the loop with the group, I had not seen the tape in 29 years, so
nobody knew it existed.Anyway, the lesson is: DONâT LET IT HAPPEN TO YOU!
But, on a more upbeat note, a quote from
âThe Restoration of Early UNIX Artifactsâ
Warren Toomey
( usenix dot org /legacy/event/usenix09/tech/full_papers/toomey/toomey.pdf )
âNever underestimate the âpackratâ nature of computer
enthusiasts. Artifacts that appear to be lost are
often safely tucked away in a box in someoneâs basement.
The art is to find the individual who has that box.
The formation of a loose group of interested enthusiasts,
TUHS, has helped immensely to unearth many hidden
treasures.â
;->
1 Like
Sadly, even the best packrats get thwarted. The Henry Spencer UTZOO Usenet Archive is now mostly unavailable because a poster from the 1980s didnât want their opinions of the time published, so theyâve asserted copyright and DMCAâd the archive from most mirrors.
BTW, itâs totally okay to include the links: editing them as host dot tld etc means youâve broken them for anyone whoâd want to read up
(Jim might have defanged the links because heâs still a very fresh member, and thereâs a limit. Itâs lifted when a person has a bit of history.)
I think itâs the case that the Internet Archive keeps things, even if it canât show things, when thereâs some claim or dispute. So in the distant future, the information they have might be available again. I donât suppose that fixes all cases though.