Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!umich!samsung!xylogics!world!madd
From: madd@world.std.com (jim frost)
Newsgroups: comp.arch
Subject: Re: Why IBM went for the 8088 -- The Untold Story
Message-ID: <1990May26.225542.22864@world.std.com>
Date: 26 May 90 22:55:42 GMT
References: <1WWjqN#0SBCWN5sLFYH9X6tz009zLSK=eric@snark.uu.net>
Organization: Saber Software
Lines: 63

eric@snark.uu.net (Eric S. Raymond) writes:
>Unfortunately for us all, the "next machine farther up" was a doomed turkey
>called the System/23, a dedicated word-processing machine like the old Lanier
>boxes that ran a customized version of Z-80 CP/M off of *slow* 8-inch floppies.
>IBM chose the 8088 design to be *less capable* than the System/23.

No, no, it must have been something else.  The System/23 Datamaster is
an obvious predecessor to the IBM PC (they keyboard for the PC and the
Datamaster are *identical* except for the keycaps on the function
keys) but the /23 was already out of production before the PC was
around.  The program was a terrible failure.

The /23 was junk.  EBCDIC, 128k, 2 8" floppies (loud as hell), a
printer that was so integral to the system that it wouldn't boot
without it being on, with keyboard, CRT, and floppies all in the same
unit.  It looked a lot like an h-19 terminal but squarer and larger.

The only language the thing ran was BASIC, by far the strangest
dialect I've ever seen.  It had an awful lot of fortran-style I/O
functionality in it so it was trivial to write report generators.

The operating system seemed to be based around the BASIC interpreter,
much like many micros (eg Apple).  It was definitely custom.

All system and programming errors were indicated by a four-digit
number in the status bar at the bottom of the screen (very 3270-ish
but the line didn't have any funky symbols, just text).  There was a
seven-volume documentation set, one volume gave you plain english
translations for the numbers.  Paired with the error was an action
code which told you what you could do to correct the error.

Not a lot of people bought the /23 since it was very expensive, had no
software or compilers, was fairly unreliable, and extremely slow.  I
believe the program was canned within a year of its release.

I modified a record-keeping system on a /23 in around 1983.  They'd
had the machine for about 5 years, I think, although the details are
kind of fuzzy.  That would put its release a year or two ahead of the
IBM-PC.  The owner of the machine said that IBM had dropped support
for the box almost immediately, although there was support available
from at least one third-party vendor.

The PC was a natural follow-up to this thing -- if the original
keyboards weren't manufactured from leftover /23 keyboards, I'll eat
my vegetables.

The internal competition decision (pick the slow one so that it
doesn't hurt higher-end sales) sounds right, it's affected most of
IBM's new "personal" machines right up to the PS/2 line and (if my
rumors are correct) was a hotly debated point in the RIOS design (you
might note that the RIOS can potentially be made quite a bit faster
with a simple field-upgrade, so I think they pushed the current market
MIPS envelope and left room for growth, probably a smart idea).  I
just don't think the /23 had anything to do with the decision.

In my mind, hell would be forced to write code for the /23 again.  I
hope never to see one of the retched things again.

Happy hacking,

jim frost
saber software
jimf@saber.com