[icus.general] 1989 Summer USENIX AT&T UNIX pc BOF Minutes

lenny@icus.islp.ny.us (Lenny Tropiano) (06/19/89)

                   1989 Summer USENIX: AT&T UNIX pc BOF Minutes

                        Sponsored by ICUS Software Systems
 	              Lenny Tropiano (lenny@icus.islp.ny.us)
                      Gil Kloepfer, Jr. (...icus!limbic!gil)

			     Tuesday, June 13, 1989
			       Hyatt Regency Hotel
			         Frederick Room

   For all those who were unable to attend, I want to summarize the events so
   the "net-community" can benefit from what was discussed.   Granted I won't
   be able to give explicit detailed account of what happened, because one, I
   was running the whole thing (and speaking a lot of the time), and two,  we
   had probably one of the longest BOF's at the 1989 Summer USENIX.   The BOF
   was  scheduled  for the 6PM to 8PM slot, on Tuesday, June 13th  (the first 
   slot for BOF's).  As it turned out we needed slightly more time than  what
   I first anticipated.   The meeting ran over till about 10:30PM, luckily no
   one had the room reserved from 8PM to 10PM.  

   Some thanks are in order: first I want to thank all of those who attended.
   Without your  support and input, the BOF wouldn't have been what it turned
   out to be.  I want to thank specifically, Alex Crain (alex@umbc3.umbc.edu)
   and Howard Motteler (motteler@umbc3.umbc.edu) for bringing their equipment
   to the BOF  so that we could use it afterwards.   Thanks  are in order for
   Gil Kloepfer, Jr. (...!icus!limbic!gil) for speaking about the ICUS 2nd HD
   plans,  showing us some of the technical aspects, answering questions, and
   taking some  notes  while I spoke.  Marc Mengel (mmengel@cuuxb.att.com) of 
   AT&T  deserves  special thanks for giving us all some  insight on the AT&T
   support procedures, as well as some excellent technical information.

   [=] What really did go on?  

   Well I introduced myself to the crowd at the BOF. I put up my fancy slides
   on the overhead projector, made using "Paint Power." (Oooo...)  I outlined
   some  topics that seemed  to  be of interest to  the UNIX-pc  community at
   large, which were:

       [I]   Hardware Realm
	     a.   Additional Hard Drives on the UNIX pc (ICUS plans, 
		  John Milton's board)
	     b.   What is meant by P5.1?
	     c.   SCSI board work ("California Project")
	     d.   "Mondo Combo" board (John Lydic's work)
	     e.   3.5" (800Kb) Floppy Drives
	     f.   Voice Power Boards
	     g.   IBM PC-bus bridge
	     h.   Tape Drives
	     i.   Packet Radio
	     j.   "Cheap networking"

       [II]  Software Realm
	     a.   Socket Library (Alex Crain)
	     b.   Internetworking   TCP/IP, SLIP
	     c.   Security issues on the UNIX pc (lack thereof)
	     d.   Things delaying FixDisk 2.0 (3.51c, 3.51d, etc..)
	     e.   Various Mailers, Sendmail, smail, elm
	     f.   UNIX pc Device Drivers
	     g.   X Windows port
	     h.   BSD UNIX port
	     i.   SVR3.0 port
       [III] Miscellaneous Realm
	     a.   AT&T Support of the UNIX pc (or lack of it...)
	     b.   Convergent Technologies & AT&T nostalgia
	     c.   How do we stand as an "user group" than can support
		  this machine?
	     d.   (800) Hotline experiences ...

   I started the ball rolling by going through these topics, hopefully firing
   up a means of discussion  among the group, which had grown to 45 people by
   that time.  Putting Marc Mengel on the spot, I asked if he would start the
   BOF off with a discussion on AT&T, from his professional point of view.

   Marc spoke about the TIER system at AT&T,  how one is greeted when calling
   the AT&T HOTLINE, and the modes of travel for your ticket. Marc worked for
   a while  on the TIER 4 support team for the UNIX pc, and still on occasion
   offers his assistance to the group (or shall we say person,  still running
   the UNIX pc dept.)  He explains, that the NSSC TIER 1 (the normal engineer
   that you'll speak with, unless the call is escalated) get rated on the how
   fast they resolve the "ticket" they are working on, and how many they will
   complete in one day.   I'm sure some of you have been put through the NSSC
   system where  the  engineer will close out a call,  even though you aren't 
   completely satisfied;  then you call back and start the complete cycle all
   over again.  Marc says, you'll just feeding them what they want,  the best
   recourse  you have is to  tell them you'll call them back,  and agree on a
   time before they close the ticket officially.  Of course, the method which
   your call escalates is when you call several times with the same  problem,
   or the ticket cannot be closed by the TIER 1 engineer.    I mentioned that
   I  will sometimes,  "confuse, dazzle, or bend the truth," after going in a
   complete circle for quite some time.  If I've been down the same path once
   before, and I  get the same  answer to  aid in engineer to close his call,
   I will explicitly say  that,  "I've tried that several times,  without any 
   good results."   By this time, the call should be escalated to the  higher
   TIERs within AT&T.   Of course, writing letters to the Data Systems  group
   president doesn't hurt. It generally gets results, he mentioned.

   By this time,  the  group discussed porting BSD UNIX to the UNIX pc.  Marc
   mentioned that there are a lot of BSD-isms in the UNIX pc version of UNIX.
   Among these were BSD 4.0 paging, SVR2.0, System III, SVR3.0 terminfo, etc.
   In other words, the version of UNIX in the UNIX pc is a conglomeration  of
   a long history of UNIXs.  More on UNIX ports to come ...

   Marc is going to mail me some information he had when he was at the TIER 4
   group  for the UNIX pc:  some  different  "panics," what to do about them,
   etc.   An example of a "panic" he mentioned was some sort of address fault
   when  the pc=0x150c (in 3.51) and 0x1494 (in 3.50).   This actually turned 
   out to be a hardware change to the motherboard.  Two  capacitors added  to
   pins on the 68010 CPU solved the problem.  This  had the TIER 4 team going 
   nuts for  quite  some time.   I started talking about the FixDisk 2.0, and
   the rumors that were circulating.  Basically I spoke about the problems we
   were having with a  bug in the  tty driver causing clists to just be lost.
   Marc mentioned how nasty that code was, and what it's original origin was:
   the DZ-11 tty driver that everyone  sees  when learning about tty drivers.
   Just as  a side note,  I am  running a  version of the kernel, UNIX3.51dD,
   that is very stable and hasn't lost a clist yet.  It  looks  like  the new
   FixDisk (2.0) will be moving on its way.

   After Marc, I handed the floor over to Gil Kloepfer, Jr.  There seemed  to
   be a big interest in the hardware modifications Gil and I worked on.   Gil
   explained the history behind the 2nd drive modification, what P5.1  really
   is, and how the upgrade actually multiplexes the two hard drives.  He  had
   a overhead slide of the schematic that we distribute with our plans.   The 
   board that is built on a perfboard has a 5 chip count: 3 TTL chips,  and 2
   disk driver chips.  No  special PAL  programming was  necessary,  which is
   unlike the P5.1 4th select line field modification that Convergent had. He 
   explained what  the board  gave them,  what was necessary,  how much tech-
   nical  expertise  was necessary,  and how it compared with the effort that
   John Milton has done.  Since John Milton was unable to attend,  Gil  spoke
   informally on the topic of the Mondo Combo board, since we heard him speak
   at the Trenton Computer Festival a few months back.  Unfortunately we were
   unable to give any specifics on how things are progressing but  it sparked
   an interest nevertheless.  Gil also spurred some interest in a serial port
   idea, about doing most of the work at the hardware level, and keeping  the
   software drivers simple.  The line discipline at the hardware  level  will
   dramatically increase the performance on the serial lines.

   A few attendees asked about tape drives. No one really had a good solution
   on how to tackle the tape problem.  The Floppy Tape is too slow, and  does 
   not have the storage capacity it should.  Writing drivers for QIC-40  tape
   drives isn't an easy task. Maybe if the SCSI idea ever gets off the ground
   that will be the way to go.  Some mentioned the serial line tape backup. I
   don't  know  of any  vendors that  sell such a device,  but in theory, one
   could write the  data out  to the serial port.   Granted it would be slow,
   even at 19.2K.   Everyone considered doing the 3.5" floppy drive a step in
   the right direction (as far as backups are concerned),  and  the  hardware
   implications are much simpler.

   After Gil we moved into the "Software Realm,"  and the interest in  Alex's
   socket code that he wrote.  Alex spoke about  serial line networking,  and
   that he is going to release his socket driver library to the  public.   It
   is purely beta, and he said, "it is assured to panic, one way or another."
   Although Alex wanted to implement AppleTalk, he opted for SLIP since  that
   was the more popular protocol.   For those interested  in beta testing the
   socket code,   contact  Alex at the  University of  Maryland in  Baltimore

   X-windows was a popular topic of discussion, since a lot of the exhibition
   this year was on the X-windows, Open Look(TM), and other  graphical inter-
   faces.  Unfortunately X-windows is quite large,  as many of the  attendees
   mentioned.  Alex said maybe we can work on a small implementation  of  the
   X-windows code, and not have to deal with the 50MB of source code.  Again,
   Alex's socket library will help in doing the networking stuff.

   Some small discussions were brought up about how  good GNU CC is,  and how
   it compares to the stock pcc-compiler.  The majority of people agreed that
   the GNU CC compiler was far superior, and it generated much more optimized
   code over the stock compiler.   Alex  is working on getting GAS and GNU LD
   working on the UNIX pc.  We look  forward to that in the future.  Although
   the GNU CC compiler generated better code, it also takes quite a bit  more
   time to compile and optimize (much to do with the binary size ~400K).

   The most surprising topic of discussion came up when we all thought the BOF 
   was about to wind down.  I knew  about this previously, but was not really 
   at liberty  to discuss the  topic  to anyone, unless it was to be publicly 
   known.   A small company, AGA, Inc.,  and  ASI, Inc. has bought the SVR3.0 
   $100,000 source code license from AT&T and is planning on working to  port
   to the, yes you've guessed it, UNIX pc.   A team is being developed  right
   now to  work on all  aspects of the project including coding, maintaining,
   etc. If anyone is sincerely interested, please mail a short letter telling
   how you can help with  this project.  E-mail to svr3-port@icus.islp.ny.us, 
   this  will automatically be forwarded to the proper people who are working
   on developing this project.

   Porting SVR3  seemed like an enormous and impossible project to undertake,
   especially since no one had access  to the UNIX pc source code, and to the 
   hardware specific code like the driver(s),  for example.   Then out of the
   clear blue sky, a person from AT&T UNIX Source Licensing spoke up. He then
   mentioned that he was fairly new to the Greensboro, NC group, but now that
   he was there they would "rock and roll!"  Most of the licensing group does
   not  know how to write one piece of code, he said.  He, on the other hand,
   has an engineering background.   He owns a UNIX pc, and wants to see some-
   thing come of it.   His reasoning is  that  AT&T might be receptive to his
   idea of  releasing the UNIX pc source code to the proper licensees.   Even
   more important,  he  mentioned  wthe possibility  of an inexpensive way of
   enabling us to be able to purchase the source product.  The chatter didn't
   stop,  everyone was talking about the possibilities.   Another  road being
   explored now is if we can  convince AT&T that the source code for the UNIX
   pc (3.51x) is  of  a lesser  version than  SVR3.0, then we should be  able
   under some provision in source licensing, to get the previous release. But
   since the UNIX pc product wasn't developed from one particular release, it
   might be hard to convince the right people.  But since he wants to "rock-n
   roll", we might have a good chance.   Of course we intend to keep the SVR3
   product (if it ever  gets off  the  ground),  compatible with the existing
   UNIX pc software.   Basically by keeping the binary compatibility, as well
   as the loadable device driver scheme, the concerns about having the things 
   like the DOS-73 (dc73) driver, Voice Power (voice) driver, Ethernet, etc..
   working, could still be maintained.

   Well that's about it, hopefully we'll see you all there in January at  the
   1990 Winter USENIX in Washington, DC. I'm sure there were many other  sub- 
   topics that came up.  Hopefully the individuals who where there can add to
   my summary.   Again, thanks for making it what it was ...

Lenny Tropiano             ICUS Software Systems         [w] +1 (516) 589-7930
lenny@icus.islp.ny.us      Telex; 154232428 ICUS         [h] +1 (516) 968-8576
{ames,talcott,decuac,hombre,pacbell,sbcs}!icus!lenny     attmail!icus!lenny
        ICUS Software Systems -- PO Box 1; Islip Terrace, NY  11752