9

I've been reading Exhibit 14971 from US vs. IBM (Parts 1, 2, 3) which seems to give a very good overview of the history of the computer industry up to 1980, particularly the way IBM handled its product planning beginning in the '60s.

One thing that puzzles me though is the "gap" in the mainframe naming conventions. We have "System/360", "System/370", and "System/390", but between 1977 and 1990 IBM used the 30XX/43XX/93XX model numbers as designations.

Why didn't IBM market these systems as "System/380"?

I can sort of understand 1977 being a little early for announcing a series "for the 1980s" so the dates don't line up as well as with the /370 and /390 families, but were there other considerations in play?

At the same time, the 30XX/43XX series was not advertised as "System/370" per se but as "System/370 Compatible". This choice puts a deliberate (albeit small) amount of separation between these machines and the rest of the series. Why create that distinction?

Are there any sources available that point to IBM's reasoning for these choices?

Wildcat Matt
  • 651
  • 4
  • 9
  • 8
    As someone who worked for IBM and lived through the "what product shall we rename this week?" mentality, this doesn't surprise me at all :-) – paxdiablo Apr 26 '22 at 20:16
  • 1
    Maybe they thought that it might be confused with the System/38, which was something of a turkey. – Mick Apr 26 '22 at 22:06
  • @Mick The S/38 is sort of alive today. The AS/400 was pretty much a better S/38. The AS/400 has been through a number of name changes but its latest incarnation is the IBM i. Being pedantic, IBM i is the latest version of its operating system. The hardware is IBM Power and is shared with Aix. – badjohn Apr 26 '22 at 22:15
  • Maybe because someone asked "why do we use slashes in system names?" – dave Apr 26 '22 at 22:28
  • I believe the reasoning between the System/360 naming is : we have you covered the whole way around the circle ie 360 degrees. So, System/370 can maybe be explained with 70s. – Stefan Skoglund Apr 26 '22 at 22:57
  • 1
    (No data at hand, so just memory bits which ma be wrong) In the 1980 it was a turn marketing wise to model numbers as the idea was that compatibility was king, Just take the XA addition, offering an (user side) address space up to 31 bit, a fundamental improvement, programs had to adapt for. Still it was only marketed as 'enhancement' Compatibility was the holy gral used by IBM as main marketing device. This fad went so far that manufacturers of compatible machines dropped their proprietary enhancements for new CPUs or at least their usage as major sales argument to stay within. – Raffzahn Apr 28 '22 at 02:37
  • @Raffzahn so basically, "we don't want to use /380 which might suggest incompatibility and increase churn from the installed /370 base but we also don't want to explicitly continue using /370 which might suggest obsolescence since it's a 10 year old name"? – Wildcat Matt Apr 28 '22 at 13:11
  • @WildcatMatt I guess a bit of both.Also, /370 was quite well settled - essentially meaning mainframe, as if everything else was somtehing less worthy :) – Raffzahn Apr 29 '22 at 13:54

1 Answers1

12

If you have a look at the announcements for the various System/XXX machines you mention, you'll notice a pattern:

Comments from the latter two of those press releases (with my emphasis) indicate the reason why they were so named:

  • We are confident that the performance of System/370, its compatibility, its engineering and its programming will make it stand out as the landmark for the 1970s that System/360 was for the Sixties.
  • System/390 -- with its broad array of product options -- is designed to satisfy computing needs for the Nineties as defined by IBM's customers who want to manage their information system resources better and integrate them with strategic business processes for competitive advantage.

The other thing you should remember is that IBM distinguishes between architectures and implementations. That's the most likely reason why the other models don't get a System/380 moniker (despite some being released in the 80s) - IBM considers them to be merely System/370-compatible (part of the System/370 architecture), not enough of a major change to warrant a new line. You can see that in IBM's mainframe timeline.

I believe the majority of these were to cater for (relatively) minor changes such as processor improvements, not massive improvements to the mainframe architecture as can be seen (for example) in the System/390 release(1). In any case, the 30xx/43xx ones started arriving in the 70s rather than the 80s, and were very much still of the current (at the time) architecture.

I have no direct knowledge of the 93xx series, at least in terms of the IBM mainframes. However, a bit of research turns up the fact that they may not have been really considered as a mainframe, given their "baby" status, described as a VAX-killer so targeting the mid-range space rather than mainframe.

(1) And even more so for the System/360 which was IBM's "bet the farm" release to try and get away from the previous situation of having way too many disparate architectures.

paxdiablo
  • 4,742
  • 20
  • 47
  • 1
    From comments at the time in the press and by both DEC and IBM staff, I understood that the AS/400 was meant to be the "VAX killer". –  Apr 27 '22 at 07:19
  • @Martin, it looks like both 9370 and AS/400 came out of the failed Fort Knox project, so they probably had similar goals, the same goal as Fort Knox itself, which was to stop DEC and other midranges from eating IBM's lunch :-) – paxdiablo Apr 27 '22 at 07:39
  • 3
    I appreciate the quotes from the press releases for the marketing angle of /370 and /390. With the advertisement of 30XX/43XX as /370-Compatible (as opposed to simply /370) it looks like there was some attempt to set them apart from the older models in a "have your cake and eat it too" move. – Wildcat Matt Apr 27 '22 at 13:09
  • 2
    From my memory the 9370 family was a brainchild of an 1980 idea to cover a lower end market (like many compatible manufacturers did at the time) without cutting the margins with existing machines. Mainframe departments at manufacturer had a very strong sense that there can't be any real reason against everything done by /370 - except the price, so a lower end version with considerable lower price was the answer - but at he same time these families were quite unwelcome at sales due lower margins. so they ended up having rather sub standard sales effort. – Raffzahn Apr 28 '22 at 02:44
  • 1
    the 9370 was essentially kept away from all but experienced mainframe users. At the same time they were as well unwelcome by the mid range people, as they of course wanted to keep their customers growing with larger /38 (et.al.) machines. The right idea but doomed by company reality :) – Raffzahn Apr 28 '22 at 02:53
  • 2
    @Raffzahn - a reality seen over and over again across many industries, but particularly obvious in computers and other fast-moving fields. It is really hard for a business to obsolete a major product line until somebody else does it for them. – Jon Custer Apr 28 '22 at 14:55
  • There's also a long history of users trying to get things cheaper and big blue accommodating that. Think of the specialist db2 and java cpus that aren't included to the same extent as the general purpose cpus when calculating lease costs. This could be where the 93xx series came from. – paxdiablo Apr 29 '22 at 02:21
  • 1
    "That's the most likely reason why the other models don't get a System/380 moniker (despite some being released in the 80s) - IBM considers them to be merely System/370-compatible (part of the System/370 architecture), not enough of a major change to warrant a new line". Counterargument: S/370-XA (released 1983) was a huge change from the original S/370 architecture – 24-bit to 31-bit addressing, fundamental changes in the IO architecture too. 370 to 370-XA was a massive jump. By contrast, ESA/370 to ESA/390 was a far smaller jump (even smaller than 370-XA to ESA/370 was) – Simon Kissane Oct 17 '22 at 05:53
  • I thought the 9370's were supposed to be mainly for local computing at remote branches (remote from HQ) of companies that were already IBM mainframe customers. The hardware price was low but the the software rental/maintenance agreements weren't ... nor was the effort required to get one running and keep it running. If you were already an IBM shop with your own expert IT people and with a maintenance agreement with IBM's worldwide maintenance people you'd consider it as a more flexible way to keep doing things the way you've been doing them.... – davidbak Dec 26 '22 at 17:45