Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!linus!philabs!cmcl2!harvard!bu-cs!bzs
From: bzs@bu-cs.UUCP (Barry Shein)
Newsgroups: net.unix-wizards
Subject: Re: ULTRIX futures?
Message-ID: <186@bu-cs.UUCP>
Date: Fri, 21-Feb-86 11:39:49 EST
Article-I.D.: bu-cs.186
Posted: Fri Feb 21 11:39:49 1986
Date-Received: Mon, 24-Feb-86 07:16:27 EST
Organization: Boston Univ Comp. Sci.
Lines: 53


Steve Summit makes some good points about convergence/divergence. One
thought that has bothered me is the "unstoppable force meets immovable
object" effect [concept stolen from a Superman comic of my childhood.]

An excellent example is the csh/sh 'divergence' flamage. Almost everyone
who gets to describe UNIX in broad terms (such as a text) speaks with
pride that UNIX' shell is just a user level program that is easily replaced
to suit needs, it really should be a source of pride and is an excellent
example of some of the generalities that permeate the system's design.
Yet, often in the same text, the incompatibilites of the two shells are
mourned.

The unstoppable force, technological development, and the immovable object,
standards/compatibility, are constantly at odds in these issues. One can
argue that csh just wasn't progress over sh, but that borders on playing
the art critic, it's largely a matter of taste, both are widely enough
in use that their success is self-evident.

The point is, the shell was made replaceable, exactly one major variant
shows up, and wars start. Something is wrong, either with the original
concept or with people's sense of humor. (I know there are other variants,
consider that statement for the purpose of discussion, I am well aware of
uucico.)

Unix lends itself to this, even without sources one can extensively
customize the system and explore alternatives. With the sources you
can start a computing revolution from your own garage (maybe even w/o.)

We must have standards, we do have standards, they're just not perfect
(and may never be.) The syntax and semantics of an operating system are
a little more complicated than FORTRAN66 and we just don't understand
things like distributed operating systems at the level that we do, say,
arithmetic. Everyone agrees we can do better in the standards area,
probably the compromise will be an extension of the standard that has
always been there: There will be a core of very capable features which
if adhered to will provide massive portability. From that secure point
the pioneers may head out to find out where the next standards need to
be, and will wave shouting /* Sorry, only runs on XYZ Unix Ports */.

Put little Venn diagrams in your head, draw the universal set, draw
SYSV, SYSIII, BSD4.2, BSD4.1, V7, V6...

Now add MVS, VMS, AOS/VS, PRIMOS, CP/M, RSX, CMS, MS/DOS...

I think you get my point, enough.

	-Barry Shein, Boston University

P.S. INFO-FUTURES, where this discussion will probably travel to, is
being set up, the requests are coming in by the scores (30 yesterday
alone) and I am waiting for things to settle down in a few days (I
hope it settles down!)