Talk:IBM System/360 Model 50

memory size terminology

I have edited the description of memory sizes to correspond to the description in the functional characteristics manual, referenced in the article. At the time of the model 50, it was customary within IBM to describe memory sizes by writing out the number; for example, 65,536 bytes. Elsewhere in the industry this would be called 64K bytes. Later there was recognition that this terminology was ambiguous: does 64K mean 65,536 or 64,000? Today there is a movement to use Ki instead of K when meaning a power of 2; for example 64 KiB. That transition is not complete, and there are still people who insist on using K to mean two to the 16th power instead of 10 to the 3rd power. To avoid that controversy I did not want to say that the model 50F had 64 KiB, so I fell back on the unambiguou text in the functional characteristics manual: 65,536 bytes. John Sauter (talk) 16:06, 27 September 2015 (UTC)

The MOS recommends K instead of Ki, etc., with which I wholeheartedly concur. However, you're right that your edit corresponds to the values given in Functional Charactetistics. Peter Flass (talk) 18:50, 27 September 2015 (UTC)

MFT

The system software section claims that the model 50 generally ran OS/360 MFT. Actually, the system software chosen depended much more on the amount of main storage than the model number. The reference to the storage estimates manual is not persuasive--the manual is as hard to read as a tax form. The reference to www.os390-mvs.freesurf.fr/mvs360.htm returns no data.

I suggest removing the whole paragraph, and making Call OS a higher-level topic. If there is no objection I will do this. John Sauter (talk) 14:28, 20 September 2016 (UTC)

I suspect most /50 shops did run MFT, but it requires a reference. The /50 was considered a mid-range 360. Peter Flass (talk) 16:13, 20 September 2016 (UTC)

The model 50 that I used ran MFT, but I agree that a reference is needed. I have removed the unsupported assertion. John Sauter (talk) 03:43, 23 September 2016 (UTC)

again

User Pi314m has restored the assertion about the Model 50 running MFT, with four footnotes. One is a bare assertion about ads in ComputerWorld from 1971 to 1973. Since it has no references it is not verifiable. Another footnote again references the Storage Estimates manual from 1973. That manual is 294 pages long, but the footnote does not say which page supports the assertion that MVT requires 256 KiB of memory. I searched through the manual but was unable to find the statement. The third footnote again references http://www.os390-mvs.freesurf.fr/mvs360.htm, which still does not return any data. The fourth footnote refers to an IBM publication from 1981 that is behind a paywall.

The new text is not much better than the old text. If there is no objection I will remove it, again. John Sauter (talk) 14:59, 10 October 2016 (UTC)

Having seen no objections, I have again removed the unsupported claim that the model 50 usually ran MFT. I did find a reference to the need for at least 262,144 bytes of main memory to run MVT, but that same reference also claims that MVT will run on both the model 40 and model 50! Also, even if MVT didn't run on the model 50, some of the machines might have run DOS. John Sauter (talk) 18:27, 12 October 2016 (UTC)

yet again

I am going to revert the assertion about the model 50 usually running MFT again. The claim is statistical in nature ("most likely") and requires a citation to an authoritative source which makes the same claim, not to evidence. Even if evidence citation were acceptable, a posting in the folklore news group of somebody's speculation isn't authoritative. Even if it were, the claim is that the model 50 would not run MVT well. That says nothing about the prevalence of MFT over, for example, DOS. The reference to a sample of ads in ComputerWorld seems like original research, and is not verifiable since the ads are not listed. The reference to CPU power, at http://www.os390-mvs.freesurf.fr/mvs360.htm, still does not return any data. That the model 65 was more powerful than the model 50 is not in dispute, but is not relavent to what software was used on the model 50. John Sauter (talk) 14:18, 1 November 2016 (UTC)

and once more

I have been trying to avoid an edit war by describing here in the talk page why I think the assertion that the model 50 usually ran MFT is inappropriate because it is unsourced. The editor who disagrees with me has recently restored the assertion, with this comment: “since a claim of REFimprove is being marked as UNsourced: US Army report comment re "a healthy compliment of core" is being ignored. */)”. I have no idea what this means. My concerns above have not been answered. I invite the editor to explain why the claim that the model 50 usually ran MFT is well-supported here, in the talk page, rather than just in a reversion comment. I will remove the text again unless the editor is willing to discuss it. John Sauter (talk) 16:52, 8 November 2016 (UTC)

More information re MFT over MVT

Among the statements by a group of (former) IBMers in https://groups.google.com/forum/#!topic/bit.listserv.ibm-main/zqAApnIeXec%5B26-50%5D are "We got blood out of the turnip back in those days." The author of those words doesn't deny that "there were a lot of folks that ran MVT on a /50" and neither do I.

These IBMers knew quite well what was in use, and it was neither MVT nor DOS as the most likely choice. In those days, rigidity was "in" sufficiently so that partition sizes were set, rather than use CPU cycles on a 50 to handle REGIONs.

Read the Army SysAdmin's report: they kept plowing through people's REGION choices.

As for flexibility, one option for MFT sites was to "collapse" partitions at a given time of day, so that a 52K Partition and a 128K Partition allowed running jobs that needed 180K. When the clock struck a later time, it was back to 52K & 128K, with the other partitions unchanged.

Re DOS on a 50? One site that ordered two 360/50 systems received one 50 and one 30, until the second 50 was delivered. Perhaps THEY kept DOS for longer than planned, but pre-POWER (spooling), and with DOS lacking a Relocateable/Relocating Loader, DOS was an unlikely choice, even before what some called MFT-II.

I hope the above is satisfactory Pi314m (talk) 05:15, 15 November 2016 (UTC)

My quesrion is: is the importance of including this information worth all the argument? Peter Flass (talk) 11:50, 15 November 2016 (UTC)

This is interesting, and the page you linked to has some valuable memories, but none of it adds up to a claim that most people ran MFT on their Model 50s. I found a claim on that page that it was possible to run MVT in 128 KiB, and another that there were a lot of folks that ran MVT on a Model 50 with 384 KiB, but not even speculation on the percentage of Model 50 MFT users. Do you have a URL for the Army SysAdmin's report? I would like to read it. John Sauter (talk) 15:00, 15 November 2016 (UTC)

The URL is https://books.google.com/books?id=_BGDRGWzVw4C&pg=PA33&lpg=PA33 Pi314m (talk) 23:51, 15 November 2016 (UTC)

Thank you for that reference. It describes a successful attempt to improve the core and disk utilization of a 3 MiB Model 65 with LCS and 80 disk drives running MVT, HASP, RJE and TSO. I recommend it to all old codgers like myself who read this talk page. It will bring back fond memories of the old optimization procedures, though doubtless most of us worked with smaller machines. John Sauter (talk) 15:43, 16 November 2016 (UTC)

You're welcome.

It is precisely these sites with less of the then-very-expensive core, along with those blessed with LCS who preferred to avoid the joys of MVT's ROLLIN/ROLLOUT, and possibly even offer the lower-overhead(than TSO) CALL/OS, who comprised the numbers to which I can't offer a precise or even close number.

There's no shame in admitting that MOST are in a group, even if that number is as low as 51% (which is how elections are decided). Requiring a super-majority (e.g. 60%) is not called for in this case.

The provable fact that many companies were selling non-IBM memory at that time is because sites hoped to improve their situation - not by going MVT but by simply getting better use of their CPU. Larger memory size as an enhancer of performance is how SYNCSORT et al built the non-IBM software industry, and it was the extra memory that made this possible. YES, you could ask me for references on these 3 paragraphs, but please consider the alternative: future readers won't know about what many of us experienced in trying to make do with what was available. The "blood out of the turnip" quote is real, and many talented systems programmers made their MFT sites dance, without going to MVT.

Those who spent time reviewing SMF records were a (dare I say major) part of this. If you insist on a Citation-needed tag, I can understand, but Wiki's writeup notes that it is . . . (my term): an eye-sore. Pi314m (talk) 08:17, 17 November 2016 (UTC)

Since I will include an actual citation with the words "most did not run MVT" (which amounts to saying most ran MFT) I will proceed, per "requires a citation." Pi314m (talk) 18:23, 18 November 2016 (UTC)

Just to throw monkey wrench, I suppose some could have run DOS. Peter Flass (talk) 00:19, 19 November 2016 (UTC)

Indeed, it is quite reasonable to run DOS on a 64 KiB Model 50. I see there have been a lot of recent edits, so I will wait before doing more. However, it is my feeling that “Most did not run MVT” does not equate to “most ran MFT”. If there is no objection I will change the claim to “Most did not run MVT”. John Sauter (talk) 04:18, 19 November 2016 (UTC)

My edits were pretty much just cleaning up the references (and flagging one reference, used twice, as a dead link, as the site doesn't respond for me), so there's no need to wait for me. The reference says "most did not run MVT", so I changed that paragraph to 1) mention DOS and 2) say "few ran MVT". I'll leave it up to others to find a source that indicates how many ran DOS and how many ran MFT. Guy Harris (talk) 05:17, 19 November 2016 (UTC)

In looking for information about DOS/360, OS/MFT and OS/MVT I learned that the minimum memory for MFT is 128 KiB, so the 64 KiB Model 50 couldn't have run MFT. John Sauter (talk) 05:26, 19 November 2016 (UTC)

That's worth adding to the "System software" section. Guy Harris (talk) 05:53, 19 November 2016 (UTC)

Good thought, I have added it. Feel free to improve my somewhat stilted wording. John Sauter (talk) 15:17, 19 November 2016 (UTC)

AAEC and DATAWAY

When I started at AAEC in February 1974 we had an IBM 360/50G running MVT/HASP quite happily. But we didn't use TSO as such, we had a homegrown ASCII network instead. Andrewa (talk) 04:17, 3 May 2020 (UTC)

How did you manage to get MVT to run in 128 KiB? By 1974 the official minimum memory for MVT was 256 KiB. Might you have been using an older version? I tried to squeeze MFT onto a 64 KiB model 40 around that time--I was able to get NIP to run by doing major surgery on it, but nothing worked after that. John Sauter (talk) 11:53, 3 May 2020 (UTC)

I was at that stage a "trainee operator/programmer" writing Fortran IV which I'd learned at Uni... so when I dropped out I was overqualified to be a computer operator but there were only two sites outside of the universities running Fortran in all of Sydney, AAEC and Forestry. Thanks for nothing Macquarie! So I just kept them both informed that I was keen and available and thank God a job came up. And learning Assembler, I'd taught myself Autocoder On Tape on the 1460 at Macquarie.

Anyway, point being that I didn't have much to do with the Sysgens at that stage, but that's before IPO so they took a whole weekend, and there was much scope for optimisation and I'd guess that was the answer.

We had some pretty bright people.

The division chief, Dr D J Richardson ("Don"), had been dissatisfied with the original tape-based Fortran system and had rewritten it to reduce the number of passes, I think that was on an IBM 1620 with a 1401 doing the unit record I/O, long before my time anyway so I may be wrong on the hardware.

The 360s (a 50 and then a 65) supported a home-grown ASCII network every terminal of which was defined to the host 360 as a 7-track tape unit but there was a patch needed to MVT to support that... they were "smart" and could raise attention. When we went to the 3031 and MVS, Don wrote an access method rather than use a patch (I don't know whether MVS supported 7-track tapes anyway). It was written in a weekend, single-handedly, on coding sheets and punched up on the IBM 029 punches, and with no macro expansion was 200 pages of print on the 1403 printer. It compiled, it ran for more than ten years, and it was never patched.

Later AEMOVE was marketted as an alternative to IEHMOVE, another of the local geniuses wrote it, it was an order of magnitude faster. He was using I/O buffers that were "circular", the channel programs chased each other around the buffer and he missed very few revs of the disk drives. It sold well worldwide. I sometimes wonder whether he worked on FAVER later, it had similar magic and by then he like I had left AAEC for private industry and we lost touch.

The third local genius ported a Pascal "trunk" compiler to the IBM mainframe. The second edition was "Pascal 2000", and again sold well worldwide. Niklaus Wirth came out to Australia specifically to meet and congratulate the team, of which I was part (I wrote and idiot tested the installation JCL for Pascal 2000, and it was trouble-free unlike the first edition, that's my claim to fame).

See my personal website for more on that team! Andrewa (talk) 01:58, 4 May 2020 (UTC)

That's interesting; thank you for the background. Can you say more about the ASCII terminals that appeared to MVT as mag tapes? Were they teletypes? Also, what is Pascal 2000? I was involved with Algol W when Professor Wirth was at Stanford, but I've never heard of Pascal 2000. John Sauter (talk) 12:43, 4 May 2020 (UTC)

The ASCII network was called the DATAWAY and interfaced to a PDP-9L computer that was linked in turn to a channel of the 360. It had two honest 7-track tape units which were also accessible to the 360. The DATAWAY was connected to the PDP9L via AAEC-designed circuit boards manufactured by DEC.

Many of the terminals were teletypes, but others were small PDP-11s and PDP-8s running experiments particularly at HIFAR, there were a couple of DECwriter II terminals and original DECwriters, both serial and parallel versions. Many of these terminals were connected via two NOVA computers which each ran a home-grown network called SMUT (serial multiple terminal) and could each communicate with the 360 throught that. There were applications running on the 360, which you typed in a semi-secret (need to know) string too access... I could for example go to a SMUT teletype and delete, rename or ZAP files on the 360s disk drives.

We also made extensive use of the 8-track paper tape punches on the teletypes. If the DATAWAY was down for any reason many experimental rigs (neutron diffraction for example) would instead dump the data to paper tape. But the DEC machines spoke a subtly different dialect of paper tape to the IBM 360. The IBM machine as built did not recognise X ""00" (no holes punched) as a character, it just skipped these. DEC tapes of course used all 2**8 possible punchings but the IBM reader was only designed to recognise 2**8-1 of them. IBM had modified our high-speed tape reader to recogise all 2**8 codes, it had a unique button labelled "FEED HOLES" to put it into our own special PDP and teletype compatible mode. The addition to the manual describing its use was hand-typed, according to our IBM CEs it was the only one in the world.

Pascal (programming language) was of course a hypothetical teaching language designed by Wirth, but someone else implemented it as a working language and it just took off. A Japanese professor then wrote the "trunk" compiler in Pascal (I forget his name) which Jeff Tobias at AAEC implemented on the IBM mainframe and Pascal 2000 was the second distribution of this implementation.

To me the most fascinating thing was that the 360/50 had in the PSW an "ASCII" bit. I do not know why this wasn't used at AAEC, but it was a good call as the 3031 and all other 370s used this bit instead to invoke extended addressing mode. The one flaw with Gene Amdahl's design was he left no spare PSW bits for such expansions. Of course with the dominance of ASCII on the Internet, most IBM mainframes today probably use the ASCII support which has now been restored in a different way.

I actually like EBCDIC but that's probably because my first bit-level machine, the 1460, was BCD, so my brain just got wired for a byte with zone and numeric bits from an early age. Sixteen bit words are worse... there are exactly four ways to logically number the bits of a sixteen bit word, depending on whether you have a bit 0 and at which end of the word you have the lowest numbered bit. At AAEC, with the PDP11, PDP9 and one other (I think it was the PDP8 but I didn't have a lot to do with them), we had three of those four from that one manufacturer alone. But back to BCD... you could actually read a 1460 object deck without needing a reference card, the opcodes for branch were B and V for example and the others all made some sort of poetical or graphical sense too! Memories... Andrewa (talk) 00:00, 5 May 2020 (UTC)

The ASCII bit affects only sign and zone codes for decimal arithmetic and ED/EDMK instructions - and was designed around "USASCII-8", which added a zero bit to 7-bit ASCII, but not at the top of the byte - it was in the middle of the byte, so, for example, "+" wasn't 0x2b, as it is in regular ASCII, it was 0x4b. That, plus the continued use of EBCDIC as the primary character code in IBM OSes and software running on them, may have rendered "ASCII-8 mode" no longer useful, so S/370 reused the bit for EC mode.

Software running on IBM mainframes now presumably just translates between EBCDIC and ASCII or ASCII-based character encodings as necessary or just runs in ASCII (especially on OSes whose native character encoding is ASCII). Guy Harris (talk) 03:04, 5 May 2020 (UTC)

Fascinating! That does explain a lot. I never used the 360 ASCII mode as it had been made unnecessary long before I arrived at AAEC, I just noted it was there and what it had become when the 3031 arrived. But it still seems to me that Gene Amdahl (who had left IBM by that stage) probably had a better idea of future directions than those who removed ASCII support rather than maintaining it. I confess to some hero worship there... the addressing modes of the original 360 seemed a pain when first encountered, but the more I used them (we wrote a lot of reentrant code) the more I liked them. But compared to the failure to exploit the protected mode of the x86... words fail me there. Andrewa (talk) 15:14, 5 May 2020 (UTC)

External links modified

Hello fellow Wikipedians,

I have just modified one external link on IBM System/360 Model 50. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:

• Added archive https://web.archive.org/web/20071220191855/http://www.os390-mvs.freesurf.fr/mvs360.htm to http://www.os390-mvs.freesurf.fr/mvs360.htm

When you have finished reviewing my changes, you may follow the instructions on the template below to fix any issues with the URLs.

Y An editor has reviewed this edit and fixed any errors that were found.

• If you have discovered URLs which were erroneously considered dead by the bot, you can report them with this tool.
• If you found an error with any archives or the URLs themselves, you can fix them with this tool.

Cheers.—InternetArchiveBot (Report bug) 20:04, 7 April 2017 (UTC)

It worked, but it doesn't actually support the claims for which it's used as a reference, so I just nuked both references to that long-dead page. Guy Harris (talk) 21:01, 7 April 2017 (UTC)