From debra@att.ARPA Mon Oct 10 07:51:07 1988
Received: from kth.se by majestix.liu.se; Mon, 10 Oct 88 07:51:02 +0100
Received: from enea.se by kth.se (5.57+IDA+KTH/3.0)
	id AA16946; Mon, 10 Oct 88 07:48:26 +0100
Received: by enea.se (5.57++/1.85)
	id AA10861; Mon, 10 Oct 88 07:49:47 +0100 (MET)
Received: from RELAY.CS.NET by uunet.UU.NET (5.59/1.14) 
	id AA21707; Sun, 9 Oct 88 20:43:32 EDT
Message-Id: <8810100043.AA21707@uunet.UU.NET>
Received: from att.arpa by RELAY.CS.NET id aa06844; 9 Oct 88 20:39 EDT
From: debra@ATT.ARPA
Date: Sun, 9 Oct 88 20:26:38 EDT
To: lar@IDA.LiU.SE
Subject: Re: what does "UNIX" really stand for? (also Plan 9)
Status: R

Hi Lars !

The 2 things you know about version 9 UNIX are correct. AT&T does not
sell it, but a number of Universities have some version of it, mostly
"left behind" from a visiting research person from Bell Labs.

Alice runs version 9, and so do many other Vaxen en Microvaxen, too many
to count them (several 8550's, 785's 780's 750's Microvax II's and III's)
Someone once ported the thing to a Sun but that was not too succesful.
Most of version 9 also runs on a Cray.
(Alice is a Vax 785. During office hours it serves 40 to 50 users, with
very reasonable performance. But most researchers now either share 8550's
or have their own Microvaxen.)

There are plenty publications about details of Version 9. Source code is
"classified" in a sense, but the general description of the system isn't.

I believe there are 2 reasons for having a different research Unix than
the commercial one:
1) The research version changes all the time. Everyone is changing Version 9
   and an "automatic software distribution" program keeps all the different
   machines up to date. There is no "old" version around to go back to, as
   needed for a commercial Unix. (While one is fixing bugs and (re)writing
   routines and utilities one must keep a "released" version around all the
   time.)
2) The research version contains non AT&T code (mostly from Berkeley), which
   is more complicated to sell.

However, the sad reality is that the new features that are being created
in research take anywere between 3 years and infinity before they become
part of the commercial Unix. Which leaves the customers in the cold and
encourages the competition to reinvent the tools research already has an
market them.

Some of the features Version has for quite a while are:
- upas (the mailer that will become standard in System V, 4.0, and which has
  understandable configuration files)
- streams (long before Berkeley and System V)
- /proc for easier debugging of running processes (also to appear in 4.0)
- system calls and utilities that work with long filenames, and the ability
  to use long file names on Berkeley file systems that are used over the
  network. (V9 itself does not have long file names, but its network file
  system interfaces to Suns that do have long file names.)
- true job control (from BSD)
- windowing software for blits 5620's and 630 terminals
- easy backup over the network, onto worms.
lots and lots of goodies (you name it, we have it).

I compared V9 with Ultrix 2.0 in a number of different tests, and V9 is
clearly more performant.

But there are drawbacks, which are hard to swallow outside research:
- large number of known bugs which will not be fixed (ever)
- very unstable platform: any utility which works can be automatically
  replaced by a broken one in 5 minutes.
- the complete V9 with all utilities probably requires a whole RA81 for
  itself, source not included. most systems do not have all utilities for
  obvious reasons.

All in all, V9 is a nice research environment, but it is not ready for general
use. If only the development area's could turn the new developments in
research into a part of System V in one year instead of an indefinite number
of years the whole user-community would benefit.

[Sigh]

Paul.