Date: Sun Jan 10 00:40:31 1982
Subject: Unix 3.0 vs V7
There is a nontrivial amount of evidence that Unix 3.0 (I've never seen
anything from Bell that calls it "III", although I may not be up to date)
is not in fact a direct derivative of V7, but split off from the "Mother
Unix"'s line of development somewhat earlier. I have no access to 3.0
sources, but I have seen and studied a copy of the manual (warning: there
is no guarantee that the manual I saw exactly matches the system that will
be released). There are a number of decidedly peculiar things, like:
- There are no multiplexed files. At all.
- Ditto no packet driver.
- There are occasional archaic things that are gone from V7.
- The function of V7 dup2() is accomplished in a completely
different and much more cumbersome way.
- There is internal evidence that ioctl arrived late (there is
a separate system call to set things like the close-on-
exec bits on file descriptors).
A plausible hypothesis is that 3.0 is the descendant of a system that
split off from the "main line" of development shortly after the
32-bit filesystem cutover but before a number of other changes that
preceded the V7 release.
What to do about it? Well, one can go with Berkeley. I don't plan to
do that partly because I'm running 11s and not Vaxen, and partly because
I'm not very happy about some of the things Berkeley has done to Unix.
My own plan is to stick with V7 as the base system and to add in various
useful things from 3.0 as the need and/or inclination arises, while avoiding
some of the more awful things. (I haven't made up my mind about the horrid-
but-versatile tty interface yet.)
Date: Sun Jan 10 10:44:51 1982
Subject: UNIX 3.0
Whether we like it or not, UNIX 3.0 is likely to become the most common
version of UNIX outside of academe. Very simply, Western's licensing
arrangements and fees make it certain that all the UNIX OEM types will use
3.0 as the base for their products -- and these are the folks who will
supply UNIX for micros.
Date: Tue Jan 19 19:29:20 1982
Subject: Re: UNIX 3.0
The only reason that VAX UNIXes have been unanimously 4.1bsd is that
32V was little more than a proof that V7 could be ported to the VAX.
I am glad Berkeley has undertaken to make 4.*bsd available, but now
that System III is out I know of one site that is DEFINITELY going to
run it on their VAX (I work there). Many of our reasons for doing so
would not apply to university sites, especially since you guys can
afford to "buy" UNIX licenses for everything (PWB, V7, 32V). Some of
the rest of us, however, have to make careful tradeoffs and there are
several valid reasons for going with System III for production use.
Date: Wed Jan 20 13:53:27 1982
Subject: Re: UNIX 3.0
Buying a 3.0 license automatically gets you permission to run all the
earlier versions of UNIX (32V, V6, V7, PWB) on your machine, so the
claim by gwyn@utexas is suspect. He gives no other reasons for preferring
3.0 over 4.1BSD, but claims there are several. I'd be very interested
to hear what these reasons are.
Date: Thu Jan 21 10:03:04 1982
Subject: Re: UNIX 3.0
UNIX 3.0 and 4.0 indeed do not page. 5.0 probably won't either.
They are looking into it but can't decide how to do it.
Note also that 3.0 and 4.0 will not run on a VAX 11/750, they only
work on a 780. 5.0 will work on the 750, I think.
Anyone considering the change should carefully examine the blurb
that Western puts out advertising System III (I have no idea why
they decided to call it System III, since internally it's UNIX 3.0).
There is a list of new features since V7/PWB. This list contains
a whopping seven items - depressingly short. Three of these seven
new features are new device drivers (KMC, synchronous terminals,
and a parallel communications link driver). They also tout their
new tty driver (better than V7, for the most part, but also totally
incompatible, requiring lots of ifdefs). The other features they
mention are named pipes (should take any good UNIX hacker half a
day to put this into V7 without peeking at 3.0), a new accounting
package, and generally newer versions of everything.
Now compare this to what Berkeley has done since V7.
There really are some good things in 3.0 that haven't been released
before. Aside from the device drivers, they are mostly in user
programs that can plug into V7 or nBSD very cleanly, such as SCCS
and a much newer and better uucp, nroff, and make. Also, consider
that what appears to be the same system runs on both the 11 and the
VAX - something that can't be sneezed at. Finally, the licensing was
done really attractively. There can be no doubt that everyone should
get a System III license, since that lets you run whatever you want
from PWB, V7, Berkeley, etc. I hear that an educational license is
about to spring into existence, too.
Date: Sun Jun 27 01:47:24 1982
Subject: Re: System III
Posted: Tue Jun 22 09:35:51 1982
Received: Sun Jun 27 01:47:24 1982
Has anyone compiled a list of things that are in V7 but missing from system
III? If not, allow me to start one - please feel free to mail me additions.
the ms macros
In addition, the compatibility mode in the tty driver (which is upward
compatible with 2.0, not V7) does not implement CBREAK, TIOCSETN, and
every time stty is called, it clears the ECHOE bit. However, the
functionality of CBREAK and TIOCSETN are in the 3.0 driver, just not
in compatibility mode.
By the way, it's been about 6 months since 3.0 was released - does anyone
on this list use it or have tried it (not counting people inside the
Bell System) and do you have anything good, bad, or otherwise to say about it?