Subject: BSD vs. SysV vs. "unix"
To: NetBSD Userlevel Technical Discussion List <>
From: Greg A. Woods <>
List: tech-userlevel
Date: 03/19/1999 20:19:15
[ On Friday, March 19, 1999 at 10:31:53 (-0800), [This is my bacque pas, this is my faux pas] wrote: ]
> Subject: Re: Changing root's shell to /bin/sh
> But even worse, you're coming along and saying "This thing in ***
> wasn't well thought out either, SysV did it better and just because
> *** looks like SysV or was derived from it doesn't mean it is SysV".

I'm not sure what you mean here -- but sure, I'll defend SysV when
someone's trying to claim some other fubar'ed idea is from SysV when
it's not.  If I picked a feature that's in some OS, say AIX, and then
said that it's a *BSD feature just because AIX has a lot of *BSD
features, wouldn't the *BSD crowd jump all over me?????
> I think I have tried to smooth enough feathers that I personally have
> ruffled on this issue, and I'm done with it.  It's clear that what you
> (collective, not singular, in this case!) want IS SYSTEM V.  Well, if you
> want it so badly, why don't you split your own source branch off of NetBSD
> and call it NetSVR4 and be done with it?  Obviously you've got sufficient
> experience by which to do so.

I think that's completely off base.

It's not clear to me at all that anyone arguing for "radical" new
features in NetBSD wants them to be identical to their origins, wether
those origins are in SysV or not.

> NetBSD is a Berkeley thing.  There are some concepts which make sense, but
> frankly SysV is NOT the Goddess' gift to the computing world, and THEY
> FSCKED UP IN THEIR IMPLEMENTATION of the concepts!  If the functionality
> of the good concepts could be implemented in a not-so-ugly way, I'll
> support it, for what my end-user vote is worth.  If not, then not.
> Simple.

Remember we're talking about different generations of software here, as
well as goals meeting entirely different requirements.

> System V does NOT have that many things worth even mentioning on
> technical merit, let alone points which excel over Berkeley, and
> as I mentioned above, those points have either been covered by
> POSIX, or we've already implemented them.

SysV's administrative functionality has always been trying to better
meet requirements that one might find in operational systems -- eg. 7x24
lights-out database servers, etc.  They're designed to be run by real
operators who wouldn't know a shell script if you showed them one.
Maybe they didn't fully meet those requirements, but that was the goal
-- that's where they were headed (and they did meet those requirements
well enough to keep the big telco switches running).  BSD's
administrative features are much more oriented towards programmers.

SysV's networking was also headed to meet very different requirements
(remember GOSIP?).

Many of SysV's programming features were refinements (or at least
attempts at refinements) to ideas that came out of the Research Unix
systems.  BSD was a branch off the Research Unix tree far before SysV's
roots branched off, and of course the SysV guys continued to work for
the same company as the Research guys for quite some time while at the
same time meeting the needs of voice network operators who were using
SysV in critical operating environments such as telephone switches, and
it shows.

Most of the absolutely ludicrous and most frustrating noise from the NIH
crowds on both sides has to do with extremely minor differences in the
systems -- things so simple and mundane as startup scripts, filesystem
layout, etc., whereas the remaining really important differences are
indeed far more interesting and important to consider.  What about a
really good and modern implementation of the streams idea (would that
make ipv6 easier to absorb)?  What about Bellovin's session management
ideas (that never even made it into SysV)?  What about implementing a
new stateful network filesystem that's truely transparent to all
filesystem semantics?

People who bundle every idea that came out of AT&T and Bell Labs into
one big ball and call it "SysV : a fscked implementation of Unix" really
do frustrate me.  Even you Greywolf claim you'll pay attention to useful
concepts, features, and so on, but you paint everything with an "ugly
brush", which immeadiately puts everyone on the defensive and gives a
totally different impression of what you're really thinking.

It's pretty hard to sit on the fence like I do when *everyone* assumes
I'm on one side of the fence or the other.  People in this crowd are
sure that I'm a SysV bigot.  My SysV friends think I've defected and
gone soft in the head.

But I don't really care.  I have specific requirements and goals and
those are what I'm really interested in.  If it were pragmatic to do I'd
kick every kind of Unix off all of my machines and run something better,
like maybe Plan 9....

							Greg A. Woods

+1 416 218-0098      VE3TCP      <>      <robohack!woods>
Planix, Inc. <>; Secrets of the Weird <>