[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

trip report

The following is my trip report for my management
in TEX format.  Enjoy.

Don Kaiser 
Boeing Advanced Systems

\headsep 0 in
\headheight 0 in
\topmargin .2 in
\footskip .5 in
\footheight .5 in
\oddsidemargin .2 in
\evensidemargin .2 in
\textheight 8.8 in
\textwidth 6.3 in

TO: \=  W. Whitehouse \= 33-22 \kill
TO: \>  W. Whitehouse \> 33-22\\
\> \\
cc: \>  D. Gilmour\> 33-22\\
\> \\
SUBJECT: \= \bf Trip Report for the 1988 National Users Group\\
\> \bf Conference for Symbolics Lisp Machines \\


The National Users Group Conference for Symbolics Lisp Machines (SNUG) was held
at University of Pennsylvania, Philadelphia, June 6-10, 1988.  Tutorials
were held the first two days and the plenary session the last three days.
Randy Groves from BCS Advanced Technology Center was the only
Boeing employee who attended the tutorial sessions.  

Before the first plenary session there were informal user discussions 
about the financial health of Symbolics.  Symbolics' sales have not been
good and they recently laid off 200 employees with approximately
400 employees remaining.  The situation seemed more serious than I 
originally thought
because of conference attendees' stories of companies not ordering machines 
or sending newly ordered machines back because
of the questionable viability of the company.

The first day's morning session involved organizational business
matters: welcome address, state of SLUG, and library status.
I made a copy of submitted library ``hacks'' which I will be installing
at the SEATAC and ITDL sites.  Randy Groves is also getting TEX through
the SLUG library which includes a free runtime Pascal that
I will be installing.  Jay Wurts, the new CEO and chairman of Symbolics,
spoke about Symbolics' future at the first afternoon session.
Jay  attended M.I.T. in the late 1960's and early 1970's
for a technical degree and the Sloan School of Business Management program.  
He started his own software company, Management
Decision Systems, and sold it for \$50 million after
selling \$100 million in software.  
Jay commented that because he has experienced all of the
problems that Symbolics is currently having, including a $1/3$ 
revenue loss in his own company,
he can readily propose solutions to the problems.
Jay emphasized that Symbolics will remain a viable company because
\item They have no debts.
\item They have good borrowing potential.
\item They have \$11 million in cash.
\item Their Ivory chip will be ready for shipping shortly.
Some of the goals of the company are
\item Ivory Chip
\item No more words or promises - working Ivory chip products will be shipped
this summer providing a delivery vehicle for users.
\item A plug-in Ivory chip board providing the Symbolics 
software environment, Genera, will be available for several workstations
including the Mac-II, Sun, IBM-PC, and DEC equipment.
This will allow customers to use standard hardware
yet have access to Genera.
\item After chip delivery, the company will focus its efforts
on layered software products for the chip.
\item Genera will continue to evolve to retain its status
as a premier software development environment.
\item Symbolics will be managed to respond to the customer's request
\item Symbolics will have beta and alpha test sites for making sure
future products will be aligned with customer needs.

In the next session the board of directors answered questions from the
conference attendees.  The main idea that came across was that
Symbolics can endure financially until the Ivory chip products sell.
Their comment was that companies can go bankrupt even though they are
making a profit, but no company that has ready cash is forced into
bankrupcy.  They have every intention of being around for many years
and heavily  interact with the users through SLUG.  After the
session, I had the opportunity to personally talk to Jay Wurts.  He is
very personable but wants people to not waste his time and to get straight
to the point.  He commented that he is enjoying the challenge of his
new position.  I feel that he has the right combination of managerial
and technical skills for his position and that he has the potential
for improving Symbolics' marketplace.

The late afternoon session had parallel sessions.  The 
session I attended was on setting up
and running Dialnet, a national and international
network.  I will be installing Dialnet at SEATAC.
This will allow us the ability not only to send electronic
mail to other Symbolics and non-Symbolics sites, but
also electronically mail software bugs and receive patches. 

The second day Symbolics gave details about their products: Joshua,
Concordia, Statice, and the Ivory Chip.  Joshua is an expert system
tool for building and delivering expert system applications.  It
provides a uniform access to heterogeneous knowledge bases.  Joshua
has two major advantages: performance and integration.  Programs can
be easily changed to use more efficient data structures without
changing the rule set or any other knowledge-level structures.  In one
case, changing the data structures sped up an application by a factor
of three.  Integrating an existing tool with Joshua can be done
without modifying the tool.  The example given was replacing Joshua's
default truth maintenaince system with deKleer's truth maintenaince
system in three days.  In addition, Lisp and Joshua are highly
integrated and can be easily mixed.  Future features for Joshua are to
add models for
\item all the standard search programs (A$^*$, B$^*$, etc.)
\item a NOAH-class planner
\item a diagnostic planner

Concordia is a development and management system 
for maintaining and delivering very large volumes of textual
and graphical documentation.  A hypertext approach is used
where documentation is written as reusable records or modules that
are stored in a database.  These modules are linked together
to form parts of a document such as
paragraphs, subsections, sections, and chapters of a book.
Online documentation access is possible using the Document
Examiner which is part of the standard Genera enviroment.
In addition, the documentation can be printed for paper delivery.
Cross-references within the online documentation can be ``called up''
by selecting it with the mouse.  Also, searches for particular
topics can be performed.  Concordia has been very successful in-house
at Symbolics for maintaining their own software documentation
(roughly 15,000 pages) and 
at Alcoa for managing documentation for a rolling mill (roughly 40,000 pages).
The comment emphasized at the session was that Concordia is
very different from desktop publishing tools.  Desktop publishing tools
are intended for small documents (five to fifty pages) produced 
by one person.  Concordia is intended for large documents (hundreds 
to thousands of pages) with very long development and life cycles, produced
by a team of writers.  Symbolics may release a less capable hypertext
tool known as LEWT (low end writers tools) for users wanting a desktop
publishing tool.

Statice is an object-oriented database tool which is based on the 
DAPLEX model.  Database objects are referred to as entities that 
have attributes.  Entities can essentially be viewed as the database 
counterpart of Lisp Flavors. Entities are shared persistant objects that can
co-exist between several Symbolics machines.  Normal Lisp objects stored
in memory are volatile and disappear when the machine crashes or is rebooted.
Each entity has an entity handle which is a Lisp
representation of an entity that points to the database object.
Statice provides concurrency control of the database objects.
Also, the user can query the database by associative and direct access
techniques.  Statice allows recovery by continually storing
database updates to magnetic tape.   A sophisticated error correcting
code was developed for the tape in order to recover from
corrupted data.  The developers tested it by scrapping the oxide
off of part of the magnetic tape and reading back the data.
Statice offers distributed computing so clients
can process simultaneously with the servers.  I am interested
in ordering Statice because our AI avionics 
demonstrations involve multiple Symbolics machines that share
the same information.  A Statice demonstration was given using a database
of Symbolics employees.  A small record of the employee was shown
where the operator could select an item from the employee's record
to view a digitized image
of the employee.  Currently, the largest Statice database (all of the 
Symbolics sites in the world) is five Megabytes,
but there is no theoretical 
limitation to the size of the database because it is
stored in files that may span across several disks.
Future plans for Statice are
\item Further work on performance and reliability.
\item Interactive browser and schema designer.
\item New software tools based on Statice, possibly a new Symbolics 
namespace tool.

Not surprisingly, the Ivory chip talk was very generic.  The chip is
based on a third generation symbolic processor architecture that is
optimized for Common Lisp, object-oriented programming, and logic
programming.  The address space was increased from the current 3600
architecture to 4 Gigawords of virtual memory.  Every word in memory
contains either an object reference or is part of the object
representation of an object.  The memory word contains 40 bits: a 32
bit pointer and an 8 bit data type tag.  The Ivory machine will have a
CPU board and an I/O board.  There will be an assortment of CPU boards
for different processor boxes.  Symbolics would not say which
processor box would be first but the logical assumption is the Mac-II
because of Texas Instrument's CPU board for the Mac-II.  The Ivory
chip has been running for 1 $1/2$ years, however, it was not released
because Symbolics is not finished porting Genera to the chip.  There
will be source-level compatibility between the 3600 architecture and
the Ivory architecture, but the compile code will be different.  All
of the layered products will be available upon the chip's first
release except for Ada and Prolog which will be available later.

In addition to new product annnouncements the second conference day 
included Mike Hilgenberg, Director of Customer Services,
speaking about Symbolics' continuing efforts to improve
customer services.  The conference attendees were quite vocal
during this session because software and hardware support
is not always up to their expectations.  This is always an ongoing
topic at every SNUG conference
and unfortunately the attendees
do not always complain to the right people.  Richard Lamson,
a previous Symbolics support person, emphasized that
support people need to develop stronger ties to the code developers
in order to provide better support.  Symbolics had no solution
other than having the support people from California visiting 
the Cambridge office once in awhile.
The major changes in software support will be the shipment of 
software Engineering Change Order (ECO) tapes, i.e., software patches, 
and setting up Dialnet to handle electronic patches.
Those who attended the conference were able to get an advanced
copy of the ECO tape which I will be installing at SEATAC
and the ITDL.  In my opinion, Symbolics'
support division was so poor 1 $1/2$ years ago that I would not
even bother calling for help.  This has dramatically improved
to the point that  I call at least twice a week for help and normally
get an answer on the spot.

Jim Dumoulin spoke about ``The Co-Existance of Symbolics Systems on a
Large Multi-Site, Multi-Vendor, Multi-Protocol Local Area Network''
in the evening session.  He gave an excellent talk about the 
network at the Kennedy Space Center and the problems he had.
Jim commended Symbolics on the fine job they did on their networking products
and for always doing the right thing, but sometimes the
right thing did not always work.
The main problem was that non-Symbolics machines did not always correctly
follow  protocols and the Symbolics machines would have to 
be changed to make up
for these inconsistencies.  The reason the Symbolics machine was changed
is that they are much more flexible.
The following are some rules he realized after being on the network.
\item Whoever is on the network first, rules.
\item Whoever can change the easiest, changes.
\item Whoever has the greatest number of machines on the network, rules.

The third day Dan Weinreb and Scott McKay from Symbolics responded
to a wish list that the attendees had submitted.  The response
to every item was that Symbolics cannot get to that right now and
the only thing on their agenda is getting the Ivory products out.
They will take the items under future consideration when the
company is more stable.  A 40 minute videotape from video production
houses was shown of television commercials and SIGGRAPH submissions created
using software from Symbolics' graphic division.  The whole tape
was very impressive and enjoyable to watch especially when I
recognized the commercials.  Also, a new SNUG board was elected
with the previous president, Paul Pangaro, remaining for a second term.

In general, this is one of the few conferences I look forward to each year.
I always receive valuable information regarding current software problems
I have and future directions of Symbolics.  My programming skills
have improved from interacting with other attendees and handouts
I receive.  In addition, Symbolics has
implemented features that I have requested during these meetings.
I look forward to SNUG 1989.

\vskip 60pt
\hspace{3.2in}Don Kaiser

\vskip 60pt
P.S. I am maintaining a library of several documents and references
for each of the Symbolics products (Concordia, Joshua, Statice, and
Ivory chip), a software conversion overview for Ivory, a software
services technical bulletin, and an ECO tape bulletin.