From jbbrus at comcast.net  Fri Sep  4 14:25:30 2009
From: jbbrus at comcast.net (Bernice Barnett)
Date: Fri, 04 Sep 2009 14:25:30 -0600
Subject: [Abqlispscheme] I'm new to this list
Message-ID: <4AA177BA.30202@comcast.net>

My name is Jeff Barnett and I'm a life-long Lisp fan who has recently 
retired to Albuquerque. A mutual colleague introduced me to Jim Prewett 
who in turn introduced me to this list. As I'm sure you all know, Jim is 
very persuasive even using e-mail. The result is that I'm now scheduled 
to give the September 20 Sunday talk. The purpose of this post is 1) to 
tell you what the talk will be about and 2) to introduce my Lisping self.

The Talk; My talk will be about CRISP (Crunching Lisp), a Lisp-like 
system that was designed and implemented circa 1971 to support 
development of speech understanding systems. (In other words, DARPA paid 
for this bit of Lisp lore through their Speech Understanding Research 
program.) CRISP provided many interesting program constructs including 
the following: a variety of "spaghetti" stacks, unusual control 
structures including coroutines in the context of the "spaghetti", lots 
of data structures including cons nodes that had one, ..., or eight 
cells, data types, slick naming mechanisms, and flexible memory 
management techniques. CRISP was designed to be used by medium to large 
groups of implementors and many of the mechanisms were there to keep 
them out of each other's hair (and namespaces). The talk will provide 
some historic context as well as a technical description of one of the 
many freaks in the Lisp Family tree.

Me: In the 1960's DARPA, then called ARPA,sponsored the development of 
Lisp 2 on the IBM ANFS Q32 computer. Working on Lisp 2 was my 
introduction to Lisp. The project was joint between System Development 
Corp. (SDC), a think tank, and Information International Inc., Ed 
Fredkin's play pen. Since ARPA couldn't figure out how to manage a bunch 
of Lisp Kids, they cajoled Minsky at MIT and McCarthy at Stanford to be 
program God Fathers or Den Mothers as need be. The Q32 offered us 
approximately 48K 48-bit words and was a time-sharing machine with a 
typical load of about 15-20 users.

I designed and implemented, with Bob Long, a Lisp 1.5 dialect in the 
late 1960's. That Lisp was used for more than a decade to support 
research at SDC and several other palaces. It ran on IBM 360 time 
sharing and restricted its address space to 2^16 32-bit words so a cons 
node fit in 4 bytes. That address space included a chunk of "small 
integers". The most interesting fact about this Lisp was that only two 
errors where found in the compiler/assembler during its entire life 
time; it was actually two manifestations of the same bug. A special 
implementation technique was invented that produced the robustness. 
Maybe we will talk about that one of these Sundays.

For most of the 1970's I was involved with the DARPA SUR (Speech 
Understanding Research) Program. I managed to sneak in the development 
of two Lisp-like systems. One was CRISP, the subject of my talk (see 
above). CRISP ran under the IBM VM370 OS and we had around 4MB virtual 
memory-a vast amount for that time. Doug Pintar worked on CRISP with me. 
The other Lisp was an interpretive system that ran on a Raytheon 
mini-computer with 32k 16-bit words. We managed to implement an entire 
speech understanding system top-end (no signal processing) in that Lisp.

I'm sure you have noted the severe resource constraints imposed on all 
of the systems I've mentioned so far. Well that was going to change in 
the future. In the late 1970's I went to USC/ISI for five years. The 
only Lisp available to me was Interlisp, a dialect and implementation I 
didn't like very much, so I did non-Lisp things for a while. The reason 
Interlisp drove me bats was unclear semantics, a major sin for a program 
language system. When I say unclear semantics, I have in mind a system 
where eval, the compiler, and the block compiler could compute different 
results for the same Lisp expression. In the mid 1980's I went to the 
Northrop Research and Technology Center (NRTC) and ran into the 
Symbolics Lisp Machine. Over the years I accumulated a bunch and used 
them for research in distributed problem solving among other things. 
With the advent of Common Lisp and the superb development environment 
provided by the Lisp machines, I realized that my days as a Lisp 
implementor were mostly passed. I stepped down to join the user class 
where I now hang out. These days, I run Allegro on my home PC.
-- 
Jeff Barnett

PS My wife looked at this e-mail a while ago and comment that it sounded 
like I was going to stand up and confess at an AA meeting. I said perfect.