Talk:IBM Future Systems project

Documents on FS?

I can't find any references for this in a cursory Google search. Can anyone provide some cites? -- The Anome 11:26, 10 Sep 2004 (UTC)

I'd be very surprised if you found any.

All IBM documents relating to FS had to have the highest security classification, entailing an elaborate system of storage and access. After the project collapsed in 1975, at around the time the daily IPL (Initial Program Load) of the prototype was approaching 24 hours to complete, documents were systematically destroyed by everybody signed-off to access them.

I'd love to see Wikipedia publish an authoritative overview of this spectral project. It was by no means small: by 1975 around half of IBM's vast R&D expenditure was reputedly tied-up in it. Needs to be based however on personal reminiscences by participants like John F Sowa. Hard to meet WP's criteria for verifiability.

Quacksalber 04:46, 7 May 2007 (UTC)

Some documents that might be interesting:

A task report from 1971 - the description sounds a bit System/38-ish, with the "New Machine Interface" (NMI) being like IMPI (the actual S/370-ish machine language whose instructions get fetched and executed), the "Execution Discipline Interface" (EDI) being like IMPI plus low-level OS calls, with a one-level store, and the "Application Development Interface" (ADI) being like (TI)MI, the instruction set generated by application-language compilers, translated to IMPI (see, for example, the comment "The criterion is: can we build an optimizing compiler down to EDI so that the executing program runs as well as it would, had we run it on an S/370 built from the same technology", S/380 and AS/400 being systems that translate MI code into IMPI code and executing the resulting code).

Some other IBM documents that have recently appeared on bitsavers.org. Guy Harris (talk) 08:29, 23 November 2017 (UTC)

Updates

I've just edited this page.

I joined IBM as a regular employee in the summer of 1974, just as the crisis over FS was coming to a head.

I'm not sure that the motivation given in the article is correct. Rather than wanting to make software easier to develop, I think that IBM's real motivation was more driven by wanting to develop a new architecture which other companies would be unable to copy. At that time, several large companies, e.g. RCA had computers which were reverse-engineered versions of the S/370.

Security on the project was intense. As pointed out in the John F. Sowa memo, which I've added as an external link, the specification was divided up into over a dozen highly classified documents, each with a separate "need-to-know." This was intended to keep a competitor from stealing the design, but what it really did was keep anyone, inside IBM or not, from understanding it.

I'm also not sure about the relationship between FS and relational database technology. Although Ted Codd was starting to work on relational databases around the same time at IBM, as I recall the database target for FS was concentrated on heirarchical databases, as IMS/DB, and plain ISAM were the predominant DB technologies back then. — Preceding unsigned comment added by RickDeNatale (talk • contribs) 11:41, 16 October 2006‎ (UTC)

I can confirm the RDB link. I was recruited by IBM Peterlee Scientific Centre early in 1973 to work on the FS project, supplying a prototype relational DBMS (IS/1) to IBM ASDD (Advanced Systems Development Division) which was set up (or reconstituted?) to pioneer a range of application software to accompany the FS launch. ASDD Mohansic prototyped these product offerings mainly in the language APL. Many research projects endured the asphyxiating security and re-badged themselves "FS" to take advantage of liberal funding and protection from Sales & Marketing stamping on them as counter-strategic. The end-result however was to drop out of history when FS was cancelled. Thus visitors to Mohansic were shown one of the star exhibits: "Intelligent Grid", a spreadsheet ante-dating Visicalc by some 7 years. The end-user operated it via a transparent touch-panel taped over the screen. I challenge someone to furnish a published reference to that!

I can furnish references to IBM Peterlee RDB-related external publications 1972-75, though. IS/1 was actually sold to a couple of UK customers under the guise of a specialised data analysis package (OLAP it would be called nowadays) named BIS (Business Information System) if memory serves. There should be some mention of that in the open literature. The fact that BIS used a pukka RDB under-the-covers (Edgar F. Codd knew about it and approved) had to be kept secret for internal political reasons. Sales Division had it in for Ted Codd because he was undermining customer confidence in the flagship data base management system IMS by going around promoting the relational view of data.

ASDD Mohansic appears to be mentioned nowhere else in Wikipedia. I think it should be recorded alongside Poughkeepsie, Endicott & Rochester as a major plank of FS.

Quacksalber 04:46, 7 May 2007 (UTC)

New updates

I had alerted John Sowa, who was one of the key figures in IBM ASDD architecture in the mid-1970s about this article. He contributed some valuable background information to the article, for some reason his contribution was mistaken for vandalism.

I've reverted to his last edit, and cleaned it up.

I can vouch for John's contributions because I was there at the time. — Preceding unsigned comment added by RickDeNatale (talk • contribs) 14:05, 18 October 2006‎ (UTC)

I have started new edits. At that time, I was leading the IBM World Trade representation in the FS project, and I still have a few memories. I am the author of the article referenced in the external links. --Dambernac (talk) 13:08, 21 November 2007 (UTC)

Would it be possible to have a sentence in the article remarking of the possibility of the project coming primarily from the sales side of the business rather than the engineering side - the project was clearly designed to destroy IBM's competitors - their mainframe-compatible rivals and the entire nascent software industry. The IBM software engineers reaction to the instruction set as "incomprehensible" also hints at a possible impossible target being pushed by the sales team. I say all this just from reading the article - I have no specific knowledge of the project. Apologies if it's not correct to post this here. 90.216.34.136 (talk) 23:08, 17 December 2008 (UTC)amr

Horst reference

Too bad this isn't a real book. I got bit by this too, once. It would be nice to see a good history. Peter Flass (talk) 20:03, 9 May 2013 (UTC)

About the overestimation of clock rates

IBM was very secrecy about clock rates of their Midrange and Mainframe machines, because how much of the intended (transactional) work can be done only partly relates to the processor clock rate. Since the Wintel world was all about clock speed, IBM chose to be silent about it to appear more competitive. I'd opt to remove the statement in the last paragraph in the Processor part of the article.

While RISC was cheaper to build, both IBM i and z/OS hardware still rely heavily on microcode. On which level is not known to me. I think this should be mentioned in the last paragraph also, because the implied sole reliance on RISC is inconsistent with the second paragraph, and also questions the ongoing backwards compatibility of Midrange and Mainframe, spanning decades.

Both should be backed by appropriate sources, since what I'm providing is from my memory.

--Poc (talk) 11:26, 3 June 2020 (UTC)

Presumably the "was" in "IBM was very secrecy about clock rates of their Midrange and Mainframe machines" means "was, but is no longer", given that IBM are no longer silent about clock rates (the z14 runs at 5.2 GHz and various POWER processors run at speeds that get up to and above 4 GHz.

"While RISC was cheaper to build, both IBM i and z/OS hardware still rely heavily on microcode." IBM i hardware relies on RISC processors, so either 1) the POWER processors are RISC processors that rely heavily on microcode or 2) the IBM i hardware no longer relies heavily on microcode. (Note also that much of the so-called "microcode" in the midrange systems going back to IBM System/38 was only called that for legal reasons. See, for example, Frank Soltis' "Inside the AS/400" - the "vertical microcode" was just IMPI machine code, called that to try to make sure that IBM wouldn't be required to provide it to competitors as a result of antitrust legislation. The S/38 was, at the bottom, an S/370-ish processor, running the IMPI instruction set and running system software compiled into that instruction set, including a binary-to-binary translator that translated MI application code into IMPI code and ran it, rather than directly interpreting the MI code in a microcoded interpreter. The IMPI processors were microcoded - that's what the so-called "horizontal microcode" was - but the various PowerAS/Power ISA processors in the "RISC" {AS/400, System i} systems and in the IBM Power Systems have little if any microcode, and run MI code by re-translating it into Power ISA code.)

As for the mainframes, most instructions are implemented in hardware, and many others are implemented by traps into millicode, which is similar to, for example, DEC Alpha PALcode, in that it's a mix of regular machine code and code using some additional processor-specific instructions and resources available only in the special processor mode to which millicoded instructions trap. Guy Harris (talk) 17:24, 3 June 2020 (UTC)