Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 5/3/83; site utcsrgv.UUCP
Path: utzoo!utcsrgv!mason
From: ma...@utcsrgv.UUCP (Dave Mason)
Newsgroups: net.usenix
Subject: Usenix day 1 review
Message-ID: <1750@utcsrgv.UUCP>
Date: Thu, 14-Jul-83 01:01:58 EDT
Article-I.D.: utcsrgv.1750
Posted: Thu Jul 14 01:01:58 1983
Date-Received: Thu, 14-Jul-83 01:44:30 EDT
Organization: CSRG, University of Toronto
Lines: 78
Just to keep you poor people who aren't attending informed:
- The opening remarks were the usual sort of thing.
- Michael Lesk gave an interesting keynote address centred around
whether Unix was really user friendly or not. The talk was good overall
but one of the major points was based on a comparison between a "command
oriented" system vs. a menu oriented system. I quote "command oriented"
because in fact it wasn't, in fact it was a keyword matching system which
was clearly superior for his application. Overall though the points were
well made.
- This was followed by an un-adulterated sales pitch from WEco. This was my
first Usenix conference, so I don't know if it was typical, but judging
from the hissing from the audience I assume not. If DEC made such a strong
sales pitch at a DECUS there would be a substantial uproar. I nominate
that if double scheduling of sessions is neccessary next year that this
session be opposite something dull like Ritchie, Thompson, Joy, etc.
discussing UN*X futures!
- I attended part of each of the A & B sessions thereafter.
- The Holt, Mendell & Perelgut (from UofToronto) paper on TUNIS
was well presented (though I may be
biased as I already had seen most of the material & am at UofToronto)
They were talking about redesigning the Unix nucleus and writing it
in a high level language with concurrency features (Concurrent Euclid)
and with potential for proofs of correctness. The design is very clear.
- Michel Gien from France then talked about SOL. It adheres closer to the
original design of Unix, but was done in basically ISO standard Pascal,
which neccessitated substantial changes.
- I moved to the programming tools session:
- Nakumara & Murai from Tokyo were talking about an extension to the man
system to include on line help from within programs and general sub-
stantial increase in speed. The slides for this were very full and
moved too fast, but it seemed interesting.
- Farley, Kunkel & Thompson from Mark Williams then described a very
advanced C source level debugger that they currently have running on
Concurrent on the DEC Rainbow. It sounds like it is probably slow
right now if you are doing complex things, but it is very ambitious and
allows a lot of possibilities.
- I then got lost in the Exhibition area for a while and missed some
reportedly interesting approaches to the "'make' problem".
- Wilens from COSI described a port of Unix (called Serix) to the IBM Series
1 computer, an act that deserves an award for bravery ( and possibly
something else). He wasn't very forthcoming about features, but the
explanation of the architecture from the viewpoint of a Unix-like OS
was quite interesting.
- Dunietz & Powell from Microsoft talked about how the Z80 is interfaced
to the 68000 in the TRS16 to perform the I/O for TRS-XENIX. Well presented
and fairly interesting.
- Laura Neff from National Semiconducter then talked about the port of Unix
(called Genix) to the 16000 with particular reference to the memory
management, and the mapping of files to address spaces.
- There was a BOF session about the Teletype 5620 a.k.a. (although not totally
correctly) the BLIT. This is a 1024x800 bit mapped terminal with a
Belmacs 32 ,256kb and no graphics processor in it. The hardware is nice
but nothing spectacular. The software on the other hand is quite interest-
ing. In the most commonly used mode, the terminal can have up to 6 user
windows simultaneously on the screen. Each of these can be connected to
a Unix process on the host, and the creation and destruction of these is
very nicely handled. The bad news comes in 2 parts: $6000 and binary only
support for System 5 only. The price didn't come in for much discussion,
but nobody in the room expressed much interest in the binaries only. As
this was partly a fishing expedition, hopefully the response will change
this policy. Note that all this is preliminary, delivery is December,
although the price sounded pretty firm.
- This was followed by a reception at the Ontario Science Centre where all
and sundry walked in slanted rooms, figured out digital logic, tried to
make a Votrax say "UNIX", and generally acted like big kids. A fine
time was had by all.
....whew!...I thought I was jotting down a couple lines...oh well,
hope this is helpful, and hope it doesn't steal all the wind out of
the sails of the people that did attend from your site.
UNIX is a trademark of ...
XENIX is a trademark...
...
-- Gandalf's flunky Hobbit -- Dave Mason, U. Toronto CSRG,
{cornell,watmath,ihnp4,floyd,allegra,utzoo,uw-beaver}!utcsrgv!mason
or {cwruecmp,duke,linus,lsuc,research}!utzoo!utcsrgv!mason (UUCP)
Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 5/3/83; site utcsrgv.UUCP
Path: utzoo!utcsrgv!mason
From: ma...@utcsrgv.UUCP (Dave Mason)
Newsgroups: net.usenix
Subject: Day 2 of Usenix
Message-ID: <1765@utcsrgv.UUCP>
Date: Fri, 15-Jul-83 09:07:22 EDT
Article-I.D.: utcsrgv.1765
Posted: Fri Jul 15 09:07:22 1983
Date-Received: Fri, 15-Jul-83 09:29:32 EDT
Organization: CSRG, University of Toronto
Lines: 63
This is going to be short. As mentioned in another note, 3 hours of typing
disappeared....sigh...
- Steffan & Veach (Bell Labs) EMACS editor shell.
- Dave Korn (Bell Labs) Bourne shell extension. Uses editor above or vi for
history editing. Job control where OS supports. Only problem:
the accronym..as someone pointed out: Real Programmers don't use
KiScHe!
- Zelitzky & Srivastava (National Semi) talked about EPascal and C compilers
for 16000. Not as polished as others but good info content. Included
info on 16000 module hardware.
- Ross (ANDYNE) & Taylor (DCIEM) talked about Unix extensions to provide real-
time support. Unlike the standard approach (particularly on RSX etc.)
the program & OS make a contract of the form (from the program view):
"I promise not to use more than X seconds of CPU time in the next Y
seconds if you (the OS) promise to make sure that I will get X sec of
CPU in the next Y seconds." The scheduler at any point runs the process
that is the closest to exceeding its deadline, and the OS contract
negotiator only accepts contracts that it knows it can meet.
- Cordy & Holt (UofToronto) talked about Turing, a new Pascal derivative now
being used for introductory CS courses. It is designed toward several
goals: easy to teach/learn, formal verification, reliability,
efficiency, interaction. The surprising thing is that it seems to meet
the goals (yes I am from UofT, but I AM impressed).
- Tanenbaum, van Staveren & Keizer (Free University, Amsterdam) talked about
portable compilers. This is a revival of the UNCOL ideas where with
L languages & M machines, one writes L front ends to produce
intermediate code and M code generators to read the intermediate code.
The system looks interesting (producing C & Pascal compilers) but slow.
The optimizations can be disabled to speed things up, and there is
an interpretter for the intermediate language (called EM).
- Then we got to the Unix directions session.
- Rob Pike (Bell Labs) gave a good talk on creeping featurism and how it was
bogging down Unix. 4.2 kernel is ~10 times the size of the V5 kernel,
and Rob sees this as a symptom.
- Jim Balter (INTERACTIVE Systems) talked about differences between System III
and System V. Seemed like a lot of incompatibility for not much gain.
- Chambers & Quarterman (UofTexas-Austin) compared 4.1c & System V. It is
rumoured that these systems both evolved from something called V7. If
this is indeed the case, it must have happened sometime before we came
down from the trees, cuz there sure are a lot of differences now.
- Mike O'Dell (UCBerkeley) talked about when 4.2 is coming out...a year or so,
except it's called 4.3, and a subset called 4.2 is already on the way,
or at least the licenses agreements are. There have been several
personnel changes and it will take a while to get everything straight.
- Then there was a Unix Futures Panel chaired by Mike Tilson (UCR) and with
Pike, Balter, O'Dell and Denis Ritchie (Bell Labs).
There was a lot of discussion and I am a little short of time, but
my overall impression was not very positive. Everybody seemed to agree
that Unix is THE system of the present, and the base of the systems of
the future, and that Unix is going to continue to gain features (tho
as mentioned above that isn't neccessarily good) but that it has reached
maturity and will slowly slip into senility. Yet more evidence that
maturity is here is the fact that there were about 1500 people at the
conference. One possible future that Unix is related to is the use of
things like the BLIT/5620, but the future will be held off until the
price starts to become more reasonable.
Well I gotta go, so I remain, with asbestos sheet :-)
-- Gandalf's flunky Hobbit -- Dave Mason, U. Toronto CSRG,
{cornell,watmath,ihnp4,floyd,allegra,utzoo,uw-beaver}!utcsrgv!mason
or {cwruecmp,duke,linus,lsuc,research}!utzoo!utcsrgv!mason (UUCP)
Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 5/3/83; site watmath.UUCP
Path: utzoo!watmath!bstempleton
From: bstemple...@watmath.UUCP (Brad Templeton)
Newsgroups: net.usenix
Subject: On the floor of summer USENIX 83
Message-ID: <5520@watmath.UUCP>
Date: Fri, 15-Jul-83 19:36:59 EDT
Article-I.D.: watmath.5520
Posted: Fri Jul 15 19:36:59 1983
Date-Received: Sat, 16-Jul-83 00:54:01 EDT
Organization: U of Waterloo, Ontario
Lines: 50
Moments from USENIX for those who could not make it:
(probably not in the procedings)
- A nasty session with the AT&T rep. Hissing whenever he said that
some new system was "truly a sophisticated ....", which he did,
several times. Alex White asked why they didn't send anybody real,
he got a round of applause.
- "Sex, Drugs and UNIX(TM)" buttons were a hot item. You did not get
these from people at HCR, and they were not worn by non-hcr high-ups
whose initials are not M.T. or R.B.
- Also popular were the old Waterloo "There's only 1 TRUE brace style"
buttons and ones that said "If SOURCE is outlawed, then only outlaws
will have SOURCE". A person very similar to me had on a button that
said "Nuke Lady Arwen for Christ". No offense to L.A.
- A good reception at the Science Centre. With this great fancy place
to explore, several people could think of nothing better to do than
to look at the Science Centre's Vax. A rather interesting laser display,
where Usenauts, pretending they were from the great unwashed, suggested
you would make a laser lens out of wood, very small rocks or a duck.
- Two Andy Tannenbaums, which caused some confusion. If the one from
Vrije (sp) university had not worn a BTL badge, it might have helped
clear things up. I made this mistake myself.
- Not at all a complete set of unix box vendors at the show.
- B NEWS SUCKS buttons, even one for Mark Horton, although I'm not sure
he put it on.
- Good IMAX show at the Cinesphere. All the standard ones plus a few
I had not seen. Hail Columbia was quite good. David Yost of Fortune
gave alb@harpo a Hail Columbia button at the show. Many die-hards stayed
to 3 am. I hope they did not show up at the 8:30 am session!
Thanks to Fortune systems. Renting the Cinesphere must have cost them one.
IMAX may become a regular USENIX feature.
We learned that Americans and Canadians are both capable of making
strongly moralistic IMAX documentaries.
- The balance of the sexes in the UNIX community is getting better, but
only slightly. Only a couple of women were at Gred Wood's net.singles
party.
- Only one vicious session, that on "cat -v considered harmful" by
Rob Pike of BTL. The UCB vs. BTL split in the unix community seems to
be getting nastier!
--
Brad Templeton - Waterloo, Ont. (519) 886-7304
Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site ucbvax.UUCP
Path: utzoo!linus!wivax!decvax!tektronix!ucbcad!ucbvax!mo
From: m...@ucbvax.UUCP
Newsgroups: net.usenix
Subject: Re: On the floor of summer USENIX 83
Message-ID: <237@ucbvax.UUCP>
Date: Wed, 20-Jul-83 11:26:35 EDT
Article-I.D.: ucbvax.237
Posted: Wed Jul 20 11:26:35 1983
Date-Received: Thu, 21-Jul-83 19:37:43 EDT
References: <5520@watmath.UUCP>
Organization: U. C. Berkeley Computer Science
Lines: 32
On the contrary - I suspect and hope that over time, people will see some
convergence between BTL and UCB systems. There will probably
always be incompatabilities, but the order of magnitude can
be dramatically reduced, and the parties involved are all
interested in seeing that happen.
As for Rob's talk being viscious, I heartily disagree (even being
on the receiving end of things). Most of his points are truly right
and large systems do suffer from "excessive creativity" over large
time. Rob used his usual flair for hyperbole to make a strong point,
but I certainly agree much more than I disagree. I do disagree with
the abstract included in the schedule (for my money, autoconfig
alll by itself is a factor of 10 improvement, but arguing that
is another day), but the real content of his talk hit several
nails squarely on the head. From looking at other "computer science
mainstreams" (dare I say ADA?), I am glad we have people strong
enough in conviction and intellectual integrity to stand up and
point out "truths which may be unpleasant". Whether they are
truths may be the subject of vigorous debate, but I thank Rob
for the service he did.
-Mike O'Dell
PS - the quote from the Lion books was priceless.
PSS - my old boss had a saying well worth repeating in a world where
some people seem to find it easier to pogram than to think about what
they are trying to do:
"Creativity is no substitute for knowing what you are doing."
Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 5/3/83; site wivax.UUCP
Path: utzoo!linus!wivax!dyer
From: d...@wivax.UUCP (Stephen Dyer)
Newsgroups: net.usenix
Subject: Re: On the floor of summer USENIX 83
Message-ID: <18458@wivax.UUCP>
Date: Thu, 21-Jul-83 19:47:56 EDT
Article-I.D.: wivax.18458
Posted: Thu Jul 21 19:47:56 1983
Date-Received: Fri, 22-Jul-83 03:29:24 EDT
References: <5520@watmath.UUCP> <237@ucbvax.UUCP>
Organization: Wang Institute, Tyngsboro, Ma.
Lines: 5
Indeed, the only vicious event at USENIX was the stand-up commentary
attacking Rob Pike that someone in the audience made after the "future of UNIX"
talk. I was amazed how personally this guy (couldn't see who he was) took
Pike's intentionally overstated comments. I must remember to add UNIX
to my list of taboos along with sex and politics.
Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 5/3/83; site cbosgd.UUCP
Path: utzoo!linus!decvax!tektronix!ucbcad!ucbvax!cbosgd!mark
From: m...@cbosgd.UUCP
Newsgroups: net.usenix
Subject: Re: On the floor of summer USENIX 83
Message-ID: <136@cbosgd.UUCP>
Date: Fri, 22-Jul-83 16:29:36 EDT
Article-I.D.: cbosgd.136
Posted: Fri Jul 22 16:29:36 1983
Date-Received: Sun, 24-Jul-83 01:47:59 EDT
References: <5520@watmath.UUCP>, <237@ucbvax.UUCP>
Organization: Bell Labs, Columbus
Lines: 34
I suppose we could all find some inconsistency in V6 UNIX to
counter Rob (e.g. ed regular expressions are different from
shell regular expressions), and I do wish I had had my set
of 6x9 Berkeley manuals there so I could have pulled them
from my briefcase as a counterexample, but he did make many
good points, and the gist of his talk was well worthwhile.
(I suppose I should point out that, even though he considers
cat -v to be wrong, I consider his Blit code to be wrong.
It's all in the eye of the beholder.)
I would like to see someone go through the manuals of system V
and 4.2BSD and heartlessly throw out as many options as possible.
Many of the options to, say, ls are just burying the poor user.
Notice I said options, not features. Many features can be turned
on, with no way to turn them off, and almost nobody would mind.
(Of course, some zealot will complain about every one no matter
what you do.) For example, the 4.2BSD "ls -g" option could be
always turned on, showing both owner and group. The -i option
could be part of -l. The -F option could be on unless -1 is on.
Similar arguments could be made for the tty driver - while it's
good to have all those strange bits to control at the ioctl level,
the user stty command needs only a few options. Things like ctlecho,
crterase (or echoe for system V), should probably always be on.
I do strongly agree with Rob that page mode in the tty driver (or
in the terminal, if at least one manufacturer could be made to
do it right) would obviate the need for every program in the world
to filter its output through more. Even the author of more felt the
code really belonged in the tty driver (although certainly not as fancy
as the more command is), but neither Berkeley nor USG will consider it.
Too bad - a half dozen UNIX systems have it (including the Fortune)
and it's a real win.
Mark Horton