Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utcs!mnetor!seismo!gatech!akgua!usl!elg
From: elg@usl.UUCP
Newsgroups: net.arch
Subject: Re: Missionary Position .vs. 69
Message-ID: <823@usl.UUCP>
Date: Sat, 19-Jul-86 13:21:36 EDT
Article-I.D.: usl.823
Posted: Sat Jul 19 13:21:36 1986
Date-Received: Sun, 20-Jul-86 06:40:51 EDT
References: <2900019@ztivax.UUCP>
Reply-To: elg@usl.UUCP (Eric Lee Green)
Organization: USL, Lafayette, LA
Lines: 42
Keywords: generalized, specialized hardware/software

In article <2900019@ztivax.UUCP> david@ztivax.UUCP writes:
>>> jdg@elmgate.UUCP (Jeff Gortatowsky) writes:
>>>It was always
>>>my feelings that, if a CPU manufacturer were to write the language compilers
>>>first, THEN generate a CPU design to run it, we'd all be alot happier.
>I know of one system which was completely developed like this.  Some
>software people wrote "the perfect language" and the "perfect OS
>concepts" and then some smart HW folks developed the hardware to
>support it.  It has some really neat features, but (of course) it has
>some problems too.  More good things than bad, but there is one BIG BAD
>PROBLEM.  It has got to be the world's most un-portable system.  Since
>the whole world is making advances in chips, the state of the art
>tends to advance faster (over a long time) than any one company can, no
>matter who they are.  If a system is un-portable, then it may be great
>for awhile, but over time, it will fail to keep up with the state of
>the art, and will end up getting tossed in the trash can of history.
>

For a somewhat example, see Multics (Honeywell Level 68). They used
special hardware to implement the segmented/ringed architecture of the
operating system, and added instructions to the processor to support
PL/1 (their choice of "perfect" language at the time, much like "C"
might be today). The example is less than complete, because the basic
architecture of the processor dates back to some ancient GE processor
that was particularly brain-damaged, but the net result is the same --
by 1976, a few years after Honeywell introduced it, Multics was for
all intents and purposes obsolete, being too slow, too expensive, and
too inefficient. Honeywell made a belated attempt in 1980 to update
the hardware enough to run at a decent speed, but it was too little,
too late -- the window of technology had passed them by.
   Interesting enough, Prime and Stratus have re-implemented the basic
Multics OS on their computers (Prime with minicomputers, Stratus with
68000-family super-microcomputers designed for fault-tolerant
systems), with no special hardware beyond a MMU. The 68020-based
Stratus handles almost as many users as a Multics unit... pretty much
proving the point that generalized technology is going to move much
faster than you can develop the technology you need to build a
specialized system around a language and operating system.
-- 
-- Computing from the Bayous, --
      Eric Green {akgua,ut-sally}!usl!elg
         (Snail Mail P.O. Box 92191, Lafayette, LA 70509)