COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	.require "let.pub" source
C00008 ENDMK
C⊗;
.require "let.pub" source
∂AIL Dr. William Carlson↓Office of Information Processing Techniques
Defense Advanced Research Agency↓1400 Wilson Blvd.↓Arlington, VA 20009∞

Dear Bill:

	As I said, I thought your summary of the sense of the meeting
was excellent.  The meeting itself was useful in making known what
everyone was up to.  I think people's plans will be affected in the
long run, and a certain amount of duplicated effort will be avoided.
Of course, it was a meeting of the tops of a lot of icebergs, and
the bottom parts will be affected more slowly.

	As I mentioned, I would much like the opportunity to try out
one or more LISP machines at Stanford.  It seems to me that it is now
demonstrated that the machine can be an effective tool for developing
and running large symbolic computation programs and that it willl
provide an effective environment for a user who has sufficient access
to one.  It needs to be seen how many people can effectively share one,
since they are still quite expensive.

	Here are some comments on the contents of the meeting:

	1. I do not share the view that environment is everything,
and language is unimportant.  Xerox PARC has created an extremely
pleasant programming environment but at great expense per user.
This environment has influenced many attempts at emulation, prominently
at M.I.T.  It remains to be seen how important quality of facilities
is compared to the number of reasonable terminals and amount of
computer time.  An experiment to determine the effect of super facilities
on rate of program development would be worthwhile.

	Moreover, I think that LISP as a language for computing with
symbolic information has many advantages that have not been realizedd
in any other languages so far.  An attempt at formulating these was
included in my %2History of LISP%1 paper which I enclose.

	2. I think it would be worthwhile to try to bring up ADA in
a time-sharing environment and provide it with analogous facilities
to LISP.  These facilities should include

		a. A standard representation for the ADA language in
ADA data structures.  Moreover, lists of ADA expressions must also
be representable and computable with, so that the language
essentially has to contain LISP.  Primitives for composing and decomposing ADA
expressions must also be available and standard.

		b. It must be possible to run single ADA statements
in analogy with the ability to evaluate single just defined LISP
functions.  Since ADA has declarations and block structure (I presume)
while LISP is single level, there will be additional problems in
making a convenient system.  This problem is one reason why the
outcome of a "friendly ADA environment" experiment cannot be predicted
a priori.

		c. The full ability to transfer AI applications
will require the ability for ADA programs to create new ADA code,
attach it to whatever is used for property lists and to run it
from inside the program.  However, this facility of LISP is not used
in all AI programs.

		d. These facilities need to be a separable ADA package,
because their associated large run-time routines will not be wanted in
ordinary uses of ADA that often have to be space efficient.

	Thanks for inviting me to the meeting.

.reg