COMMENT ⊗   VALID 00003 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	∂05-Feb-77  1335	JMC  
C00125 00003	∂22-Jan-77  2039	JMC  	about messages
C00126 ENDMK
C⊗;
∂05-Feb-77  1335	JMC  
To:   RPG, MFB    
CGOLMA[w77,jmc] is the latest CGOL manual.

∂05-Feb-77  0127	JMC  	CGOL
To:   PRATT at MIT-AI  
I tried it out.  It took a while to learn how to define functions.
My first reaction is that the notation is not very pretty, but maybe
I'll like it better when I have tried it further.  Is there a more
recent writeup than that dated 4/17/76?

∂04-Feb-77  1322	JMC  
To:   pratt at MIT-AI  
I intend to try CGOL, but I haven't had time yet.  I would
welcome converting LC0 and LC4 to MACLISP.  The compiler will run on
MACLISP as soon as DE is replaced by DEFUN, but the output code may
not run.  I further want to modify LC4 in an understandable way to
produce JRST instead of CALL (hmm, CALL may not run in MACLSP) on
the top level, so that iterative functions will be handled well.
If you were to make that happen while you do the other conversions,
that would be very nice.  LC0 should not be improved unless the
improvement shortens it and makes it easier to understand.

	Our recent PhD student, Corky Cartwright, has shown how to
describe LISP functions by formulas of first order logic, so that
termination can be proved by structural induction, and I have added
an axiom schema of minimality that characterizes the least fixed point
to the same extent that the induction axiom schema characterizes the
integers; i.e. there are non-standard models, but they are really weird
and admit almost all of the same theorems as the standard model.  This
isn't really written up yet, but if you want to fight through incomplete
things, there is a file called THEORY[W77,JMC] that will be a revised
version of chapter 5 and will contain the theory, a file called SUBGOA[w77,jmc]
which relates my schema to subgoal induction and several messages from
a graduate student named Wolf Polak (WP) in my message file.  Well, I'll copy
these messages into SUBGOA so you won't have to root about.

∂04-Feb-77  1252	JMC  
To:   ef at MIT-AI
Ed: Are you still collecting names.

 ∂04-Feb-77  0832	FTP:Feigenbaum at SUMEX-AIM	ARPA DIRECTOR CANDIDATES   
Date:  4 FEB 1977 0832-PST
From: Feigenbaum at SUMEX-AIM
Subject: ARPA DIRECTOR CANDIDATES
To:   JMC at SU-AI

TWO OTHER CANDIDATES ARE:

1) (OBVIOUS) JERRY ELKIND OF XEROX PARC
2)(NOT SO OBVIOUS) @ROFESSOR ROBERT RITCHIE OF THE UNIVERSITY OF WASHINGTON,SEATTLE.
 PLEASE PASS THESE NAMES ON TO FREDKIN.
THANKS.
ED
-------

∂03-Feb-77  0006	JMC  
To:   CARL at MIT-AI   
Didn't I give you a copy of my Lisp notes.  Chapter 3 of those notes
gives two such compilers.  The source is LSPDOC[206,LSP].

∂02-Feb-77  1105	JMC  
To:   nilsson at SRI-AI
Mar. 21 is fine.  The title is Towards a General Theory of Patterns.

∂02-Feb-77  0451	JMC  	Getting a replacement.  
To:   minsky at MIT-AI 
	I have declared a desire to be replaced as Director of
Stanford AI Lab - mainly because I think I skimp the job of knowing
what everyone is doing here and elsewhere and getting best possible
people and bragging about their work.  The present candidate is
Nils Nilsson.  Do you have any comments, advice or potential
candidates?  I think it needs to be someone whom Stanford will
make a full professor.

∂02-Feb-77  0449	JMC  
To:   LES    
 ∂02-Feb-77  0247	MFB  
  IT LOOKS TO ME LIKE THE BEST TIME TO WORK (I.E. RUN LARGE LISP PROGRAMS)
IS IN THE EARLY MORNING (4 - 8 A.M.). I CANT TRAVEL TO THE LAB THEN
BECAUSE SOMEONE ELSE USES MY CAR TO GO TO WORK AND ITS TOO COLD TO
MOTORCYCLE. DO I STAND A CHNCE OF GETTING A HOME TERMINAL?
					MARTIN
					BROOKS
Les: How many equally worthy candidates are there? - JMC

∂02-Feb-77  0447	JMC  
To:   MFB    
Let me look into the magnitude of the home terminal problem.

∂01-Feb-77  2233	JMC  	previous memo request   
To:   oestreicher at USC-ISI
The address is John McCarthy, Artificial Intelligence Lab, Stanford CA 94305.

∂01-Feb-77  2225	JMC  	memos    
To:   oestreicher at USC-ISI
Can you send me ISI memos 74-23, 74-26, and 74-27 concerning the
proposed military message system.

∂01-Feb-77  1254	JMC  
To:   WP
Come and talk, because I don't really understand your message.

∂01-Feb-77  1250	JMC  
To:   JBR    
Random file in LET,JMC.

∂01-Feb-77  1059	JMC  
To:   JRA    
Regrettably you have a competitor, namely me.  I have already arranged
to teach CS206 again this Spring in order to debug my own text.  Because
of new developments in formalization of recursive functions, I
can now put strong emphasis on proving programs correct.

∂01-Feb-77  0302	JMC  
To:   JBR    
Why does it take 85 disk references to DI a single file?

∂30-Jan-77  1845	JMC  
To:   RWW    
The file is know.art[f75,jmc].

∂30-Jan-77  1226	JMC  
To:   PMF    
Yes, no bad characters now.

∂30-Jan-77  1208	JMC  	XEROX MACHINE SIGN 
To:   PAT    
Please put a DYMO label on the lid of the Xerox saying, "If you leave
this lid up, light will damage the drum, and the copies will get cruddy".

∂30-Jan-77  0956	JMC  
To:   DSB    
I visualize the first stages of running things out of the editor as
follows:

	1. File control.  examining directory, copying, deleting, renaming,
file switching as in E.

	2. Treating the page printer as a "file" to be edited.  Output
can be deleted to save space on the screen, rediverted to be input
after editing (like <ctrl><cr> at AI but with the whole dialog accessible),
parts of the page printer can be put into files and taken out.

	3. TEE should be able to support DDT and incremental
compilers, but they shouldn't be part of it.  Otherwise the project
gets too unwieldy.

Remarks:

	1. The project must start with a draft user's manual.

	2. The inexperienced and infrequent user must take first place.
This needn't mean verbosity, but requires avoiding mysteries.  I
especially want to provide for the person who never writes a program
but uses the system to prepare documents, e.g. term papers.  The idea
is that convenient facilities here will allow Stanford to require
a higher standard of writing.

	3. Please phone me, and we'll get together and discuss it.

∂29-Jan-77  2305	JMC  
To:   JAB    
I think his name is Blaine.

∂29-Jan-77  2256	JMC  
To:   JAB    
Please talk to Dick Gabriel about taking MACLSP.  ILISP, which
I also want should come from IMSSS.  You can also talk to
Mike Farmwald who will TA CS206 in the Spring.

∂29-Jan-77  1144	JMC  
To:   RWW    
Ashok Chandra is coming to Stanford Feb. 24 and 25.

∂29-Jan-77  1143	JMC  
To:   ef at MIT-AI
 ∂29-Jan-77  1032	DEK  
the largest known prime is 2↑19937 - 1, it has 6002 decimal digits.
reference: vol2 p356, see index under prime number,large.

Why don't you compute how much computation would be required to
enlarge this one?

∂29-Jan-77  0153	JMC  
To:   TTP    
How about Monday at 1pm?

∂28-Jan-77  2309	JMC  
To:   boyer at SRI-AI, moore at SRI-AI
I wasn't miffed at what you said.  It's just that I think the
considerations in our proof about equality of state vectors
modulo "don't-care-registers" will be necessary in many practical
cases so that it is unfortunate that both you and Milner and
Weyhrauch in their LCF proof avoided them - whether from choice
or necessity.

∂27-Jan-77  0044	JMC  
To:   CLT    
You may be interested in subgoa[w77,jmc].

∂27-Jan-77  0037	JMC  
To:   WP, ZM 
Here is more of the previous note.
	Notice that the Cartwright idea of saying indomain(x) for saying
that x is defined is required in order to express subgoal induction
as an axiom schema in logic.  This suggests that other methods of proving
partial correctness can also be described as axiom schemata.

	In fact consider the mutually recursive definitions

	f1 x ← if p1 x then x else f2 h1 x

	f2 x ← if p2 x then x else f1 h2 x

which give rise to the axioms

	∀x.(f1 x = if p1 x then x else f2 h1 x)

and

	∀x.(f2 x = if p2 x then x else f1 h2 x),

and a minimization schema involving two function variables which I won't
write down.

	The inductive assertion method is then expressed by the schema

	∀x.(P(x) ⊃ q1(x,x)) ∧
	∀x u.(q1(x,u) ⊃ if p1 u then Q(x,u) else q2(x,h1 u)) ∧
	∀x u.(q2(x,u) ⊃ if p2 u then Q(x,u) else q1(x,h2 u)) ∧
	⊃
	∀x.[P(x) ∧ indomain(f1(x)) ⊃ Q(x,f1(x))]

where P, q1, q2 and Q are all predicate variables taking the indicated
numbers of arguments.  The fact that indomain occurs only once in the
schema and not at all in the equations may seem odd, but it must occur
in some induction schema in the axiomatization of the basic functions
and predicates p1, p2, h1 and h2 or you won't be able to prove anything
non-trivial.

Questions:

	1. What other valid schemata are there?

	2. How do we express the inductive assertion method as a single
higher order schema independent of the number of functions being mutually
defined and with the specific recursion schemes expressed by instantiating
higher order function parameters?

∂27-Jan-77  0002	JMC  
To:   WP, ZM 
Just as ZM and WP were telling me, my minimal fixed point schema is
related to subgoal induction.  Namely, it is a subcase of subgoal
induction, and subgoal induction also takes the form of an axiom schema.
This note is a reaction to WP's 1/26 handwritten note.

	Suppose, as in that note, we have the recursive definition

	f(x) ← if t x then  h x else g1 f g2 x.

This gives rise to the first order logic formula

	∀x.(f x = if t x then h x else g1 f g2 x)

and the minimal fixed point schema

	∀x.(indomain(if t x then h x else g1 g g2 x) ⊃
		g x = if t x then h x else g1 g g2 x)
	⊃ ∀x.(indomain(f(x)) ⊃ f(x) = g(x)).

(I hope omitting parentheses somewhat inconsistently after one argument
functions and predicates hasn't caused confusion.)

	However, the general rule of subgoal induction for this function
can also be expressed as an axiom schema, namely

∀x.(t x ⊃ q(x,h x)) ∧
∀x z.(¬t(x) ∧ q(g2 x,z) ⊃ q(x,g1 z)) ∧
∀x.(P(x) ∧ q(x,z) ⊃ Q(x,z))
⊃
∀x.(P(x) ∧ indomain(f(x)) ⊃ Q(x,f(x)))

where q, P, and Q are considered free predicate variables of the appropriate
numbers of arguments.

	Now the minimal fixed point schema is an instance of the
subgoal induction schema - namely we take

q(x,y) to be y = g(x)
P(x) to be always true
Q(x,y) to be y = g(x).

	This seems to be a simpler proof of the main result of WP's note.

∂26-Jan-77  2328	JMC  
CC:   WP
zm/a

∂26-Jan-77  2325	JMC  
CC:   WP
zm

∂26-Jan-77  1226	JMC  
To:   ef at MIT-AI
 ∂26-Jan-77  1129	FTP:Taynai at SUMEX-AIM	DDRE Candidate  
Date: 26 JAN 1977 1130-PST
From: Taynai at SUMEX-AIM
Subject: DDRE Candidate
To:   JMC at SU-AI
cc:   Feigenbaum, Taynai


A possible  candidate for the DDRE position would be
Barney Oliver of Hewlett Packard.  I believe that his title is
Vice President for research, but I am not sure.
Ed
-------

∂26-Jan-77  1224	JMC  
To:   JAB    
The relevant official at NSF is sick for 2 weeks, but D.E.C. promises
co-operation.

∂26-Jan-77  0145	JMC  
To:   Berliner at CMU-10A   
This is a test preparatory to investing in the real message.

∂26-Jan-77  0143	JMC  
To:   fxxx at CMU-10A  
foo

∂25-Jan-77  1116	JMC  
To:   JAB    
Ralph will no doubt do the right thing.

∂24-Jan-77  1540	JMC  
To:   ef at MIT-AI
Feigenbaum suggests William F. Miller, Stanford Provost for DDR%E.

∂24-Jan-77  1148	JMC  
To:   friendshuh at SUMEX-AIM    
I'll be there.

∂23-Jan-77  2112	JMC  
To:   ZM
Do you want a terminal at home?

∂23-Jan-77  1959	JMC  	primes   
To:   ef at MIT-AI
Why can't one quickly get even bigger primes by testing
2p**2+1,6p**2+1, etc., or even a higher power?

∂22-Jan-77  2149	JMC  
To:   DEK    
How many digits does the largest known prime have?

∂22-Jan-77  2037	JMC  
To:   pratt at MIT-AI  
We already have such a program called BBPP for
"Blackboard pretty print" since my publication notation was
called blackboard notation.  Its writeup is BB.DOC[206,LSP],
and it can produce files directly for the XGP, for PUB, for
our line printer and for teletype according to taste.  I also
have PUB text responses in MEMO.PUB[LET,JMC] that make it easy to write
publication language in documents.  Unfortunately, BBPP produces
PUB incompatible with what I use for memos.

∂22-Jan-77  2017	JMC  	Text preparation   
To:   ROZ    
	I have been pursuing Xerox about getting one or more of their
500 dpi printers for Stanford.  There seems to be some possibility, but
the man in charge is rather vague about Xerox's decision making
capabilities in this direction, muttering about lawyers.  It wouldn't be
for LOTS unless (a) Xerox gave it away and (b) Massy and Miller were
to agree to expand LOTS' capability so that it could serve a lot of
text processing people in addition to the students in courses that
use computing.  Therefore, I have been thinking about a central facility
that had the communication capability to receive files from various
computers.  Such a thing would pretty well need an operator to take the
paper out and therefore would probably have to be operated by SCIP,
since it would be too expensive for anyone else to maintain 24 hour
service.  The eventual solution is to have at least as many XGPs as
the campus has Xerox machines so that they could be ubiquitously located
and not require operators.

	Phototypesetting is another direction to be considered.  M.I.T.
AI Lab produced drafts of Pat Winston's book on AI using the XGP and
the final copy using an Information International Comp80 phototypesetter.
It would be very desirable to have XGP quality proofs and typeset
quality final documents in a compatible system.  I would be glad
to discuss the campus-wide editing and typesetting problem with
you and Gene.

∂22-Jan-77  1528	JMC  
To:   pratt at MIT-AI  
On Friday, hopefully, Patte mailed you a copy of my LISP notes which
make up a few chapters of a book.  The source for the book is
LSPDOC[206,LSP], but it won't be easy to get much impression of
what the notation looks like from that unless you can PUB it.
Because the publication language uses boldface and italics, it
is not suitable as an input language.  It seems to me that the
language most convenient to write, the language most convenient
to read, and the language most convenient as an object for programs
to manipulate are all different, and the expert is best off using
all three.  Perhaps with a Xerox style fancy display and an ALTO
to do the work, one could have publication language appear on
one's screen.

∂21-Jan-77  1148	JMC  
To:   RWW    
Gurkewitz use seems fine to me, and she can start when she
wants to, but I would like a letter from her or from Martin requesting
it, giving expected amount of use and duration of project.  If the
present use is preliminary with a decision about further use to
be made later, this should be stated.  Also full co-ordinates:
address, telephone, ARPANET address.

∂20-Jan-77  1622	JMC  
To:   pratt at MIT-AI  
You might send LSPLUG to SIGPLAN; they print discussions of the relative
virtues of programming languages.

∂20-Jan-77  1216	JMC  
To:   pratt at MIT-AI  
We have converted LSPLUG to PUB and printed a nice-appearing version,
so that's what we'll distribute.

∂20-Jan-77  1213	JMC  
To:   TW
Is Pressburger doing anything for you?

∂19-Jan-77  0252	JMC  	McCarthy-Painter compiler.   
To:   boyer at SRI-AI, moore at PARC-MAXC, RWW, RSC  
	As I thought, your idea can be implemented so as to avoid
having to reverse the list.
We have
	compile(e,t,u) = 
if isconst e then cons(mkli val e,u)
else if isvar e then cons(mkload adr e,u)
else if issum e then compile(s1 e,t,cons(mksto t,compile(s2 e,t+1,cons(mkadd e,
u))))

Here u is object language code that follows the expression being compiled.
The correctness statement is

	outcome(compile(e,t,u),eta) = outcome(u,eta4)

where

	eta4 =(sub t) a(ac,value(e,xi),eta).

	Because of the =(sub t), the correctness statement can't be
written as a single equation, but the proof seems to go about as in my
paper, but the change obviates the need for the lemma about append.

	Incidentally, do you use =(sub t), or do you avoid it by using
a stack as Weyhrauch's LCF proof did?  If so, you prove less.

∂19-Jan-77  0214	JMC  	10 chars/sec vs. 30.    
To:   TED    
How soon can you make the dial-out line 5505 or else make the
speed-switchable line 5505 or freeze the switch to 30 chars/sec?

∂18-Jan-77  2110	JMC  
To:   PMF    
Dick Gabriel has agreed to help move MACLISP to LOTS.

∂18-Jan-77  2025	JMC  
To:   JB
We have looked, couldn't find it, but I'm sure it will turn up.

∂18-Jan-77  1405	JMC  
To:   pratt at MIT-AI  
I assume it's ok to circulate LSPLUG in the CS Department.

∂17-Jan-77  2336	JMC  
CC:   LES, REG    
Note the attached message from Glenn Ricart and my reply at
the bottom.

 ∂17-Jan-77  2017	DGR   via NBST	Dialnet   
Thanks for the writeup on DIALNET.  Your proposal that DECUS
co-operate in implementing Dialnet is accepted.  I'm the DECUS
Networking and Communications Working Group Chairman when
I'm not wearing other hats, so that makes it official.

I didn't bother to consult anybody else yet (although I am sure there will
be great enthusiasm from some) because I am personally interesetted
in such a project.

At the National Institutes of Health, we have found a need for 
occasional PDP-11 to DEC-10 communications of relatively low volume
if it were to be achieved at low cost.  Therefore, Keith Gorlen and I
have written what we call CLINK (Communications LINK)--a set of software
in BLISS that implements the DDCMP error detection/retransmission
protocol and an ad hoc version of DAP, a data access protocol.
The programs run in user mode on both machines, and hence
do not have to be resident.  The hardware used is simple, inexpensive
asynchronous lines.  The PDP-11 user dials any DEC-10
timesharing port, uses a virtual termilnal mode to LOGIN, and then
invokes the DEC-10 program which in turn sends a DDCMP START message
to the 11 to get it fired up running DDCMP.  They then transfer
files between the two machines.

One of the things in the back of my mind has been that CLINK can
be adopted for DEC-10 to DEC-10 or PDP-11 to PDP-11 communications without
any major work since the source for the two ends is mostly in common.
Most people use CLINK with Vadic 3405 modems (1200/1200 baud) and
get about 700+ useful bits of data per second over the link.
I have also arranged for a dialer on a Bell 100-series line for attempting
DEC-=10 to DEC-10 communications but have not had the time to get the
dialer working.

I envision DECUS discussions taking place through teleconferencing and mail
messages over a network with single links that come at night and disappear
after the day's traffic is done.  I see users submitting problems to DEC
online.  I see software distribution taking place to your machine overnight.
I see the DECUS library as being accessible on-line.  Etc. of course.

My thinking follows very closely the lines that you have written in 
Section 1. "The DIALNET Idea".  Some of the ideas in section 2. "Scenario"
go beyond what I had considered (those are (a) terminal linking, but this
is a good idea (b) footnoting, I have often wondered what a good way would
be to do this, but had always given up as to ohard to implement, and (c)
running programs on a remote computer, because of the dissiculty in
charging for the time).  A comment on 3, standardization.  If you
would like, DECUS has a liason to the ANSI x group and could be used to
help promolgate a standard.

Issues: What error correction facilities are required? DDCMP

What is the trade-off between buffer size and compute time? My first reaction
is not to worry about it, but measure it later and then use the
best size for the machine you are on.  The network should accpt varying sizes.

Can dial-up communications rates meet most of the needs for computers
belonging to different research organizations?  An imponderable question, but
we have found that it suffices for a large fraction of people with their "own"
PDP-11s who want to talk to our 10.

What is the best way to handle the fact that different modem speeds have
different prices.  ?  Select lowest common demoninator and allow higher speeds as options.  The higher speed formats are not widely agreed upon, and much
as I like the Vadic modem (we have several hundered here) you have 202-type
higher speed modems.  (and there is a third entrant in this field now, as well as ell,
who will come out with their 212 modem).

What style of interaction is convenient?  We can draw on the ARPAnet work,
and copy, or strike out on our own.


What is the status of the NSF proposal?  What kind of SAIL support were
you envisioning for the project?  There are some questions about the nature
of contributions that can be made by DECUS volunteers, and a positive
role needs to be defined.  If I were going about it, I would suggest that
(1) some overall design work be done, (2) CLINK extended in several ways
to rpovide the DDCMP, (3) User interaction programs be written, (4)
Performance measurement and evaluation in a test environemnt, (5) Rework
based on four, and (6) global promulgation.
				. . . .. Glenn

The NSF proposal has been submitted, and we expect a reaction soon.
As to modems, we need to evaluate price-performance of the modems,
and compare it with the cost of the rest of the project.  Conceivably
we could work with a modem maker, if a new design would make it all
work much better, and it were judged that it wouldn't be relatively
too expensive for users.  Les Earnest and/or I will react further to
your remarks, but needless to say, we welcome your co-operation.  I
just received a letter from Bill Kiesewetter of D.E.C. offering their
help also.

∂17-Jan-77  2335	JMC  
To:   DGR    
The NSF proposal has been submitted, and we expect a reaction soon.
As to modems, we need to evaluate price-performance of the modems,
and compare it with the cost of the rest of the project.  Conceivably
we could work with a modem maker, if a new design would make it all
work much better, and it were judged that it wouldn't be relatively
too expensive for users.  Les Earnest and/or I will react further to
your remarks, but needless to say, we welcome your co-operation.  I
just received a letter from Bill Kiesewetter of D.E.C. offering their
help also.

∂16-Jan-77  2323	JMC  
To:   day at MIT-AI    
The AI Lab has had psychology students.  Please telephone.

∂16-Jan-77  0228	JMC  
To:   REG    
 ∂15-Jan-77  2214	DEK  
the service on lots is so much slower than at ai, i don't know how
to tell my students to use it when they have accounts e.g. in the
music program. 
i have 100 students trying to use lots, perhaps more when you count the
3o or so watching on tv. there are lots of other classes using lots too,
and lots of volunteer hackers. the students have to wait hours to get
a console, and then the response time is so slow it sometimes types
only one character at a time; one student told me it takes
three minutes just to log off. what do you plan to do when MORE courses
use lots? this quarter it is just "pilot" runs... I know that the
memory hasn't all arrived yet, nor have all the display
terminals, but even so i can't see how the lots system can
be anything but swamped from now on.  net result is that the
students will be happier using punch cards at scip, but they wont
be able to get accounts any more...   if lots service were better,
i wouldn't hesitate to insist that cs144 students run on it. but as it
is, i don't want to, since it is surely no worse this year than in
the last 4 years cs144 students have been using mix up at ai. in factt,
the burden on you should be lighter now, since you now have a kl-10.

Well, we'll have to do our own chasing then.
When the memory comes, we'll have to evaluate LOTS and see how its
capacity compares with the requirements.

∂16-Jan-77  0227	JMC  
To:   DEK    
Well, we'll have to do our own chasing then.
When the memory comes, we'll have to evaluate LOTS and see how its
capacity compares with the requirements.

∂16-Jan-77  0221	JMC  
To:   LES    
OK, I guess on Feinler.

∂16-Jan-77  0220	JMC  
To:   LES    
For that I'll double your fee.

∂16-Jan-77  0218	JMC  
To:   pratt at MIT-AI  
LOTS now has 32 ports, and about 29 of these are available to users.
Our order was for 48, and the others will be delivered in March.
We intend to order  another 16.

∂14-Jan-77  1403	JMC  	ICJAI paper   
To:   reddy at CMU-10A 
This is to accept the invitation to give a paper.
If you are at all interested in being Director here,
I would like to discuss it with you.

∂14-Jan-77  0226	JMC  
To:   HPM    
Must that monstrous job run for such a long time?

∂14-Jan-77  0216	JMC  
To:   LES    
Is PN one of the 3 new musicians we agreed to?

∂14-Jan-77  0215	JMC  
To:   JC
Did you mail me a message about new users?  I can't find it in mail file.

∂14-Jan-77  0208	JMC  
To:   LES    
I have been chasing MIX users off.  CS144 is on LOTS.

∂14-Jan-77  0202	JMC  	CS144    
To:   DEK    
I notice that some students are running MIX on this machine.
Could you announce that they should do CS144 on LOTS.  For unknown
reasons, this machine has been especially loaded recently.

∂14-Jan-77  0201	JMC  
To:   KIP    
CS144 is on LOTS - not on this machine.

∂13-Jan-77  2340	JMC  	CSDDIS   
To:   LES    
Budget looks ok for first try.  Can you get proposal to Feigenbaum
tomorrow?  Call him?  11am to discuss it would be great for me, because
I will be at Zohar's class tomorrow from 10 to 11.  Incidentally, I
got a letter today from Kiesewetter expressing interest in Dialnet.
I tried to call Weingarten, but he is out for a couple weeks.

∂13-Jan-77  2327	JMC  
To:   RXM    
CS144 runs on LOTS not here.

∂13-Jan-77  1444	JMC  	Future Study  
To:   Dertouzos at MIT-MULTICS   
I was talking to Joel about it.  I have finished more commentaries than
I have yet sent you.  He said you would be more active on the volume soon.
He also told me that there exists a bound volume of Xeroxes of the papers.
Can you send me one posthaste?

∂13-Jan-77  0144	JMC  	MIX 
To:   CLT    
Are you using MIX here for a course that should be using LOTS?

∂13-Jan-77  0134	JMC  
To:   BTH    
You are "UNKNOWN"; use is heavy.  What are you doing?

∂13-Jan-77  0132	JMC  
To:   REM    
Use is heavy tonight.  Is your use to AILAB advantage?

∂12-Jan-77  2256	JMC  	cs206    
To:   DSB    
Yes, I will teach it in Spring.  If you know LISP and are
working with proof-checker, then there may not be great
point in your taking it - especially if you read the course
notes.  I will be glad to discuss it with you if you want
to.

∂12-Jan-77  1543	JMC  
To:   RWW    
Clobbered again by not being allowed to represent NIL by NIL.  Goddamn!

∂11-Jan-77  2009	JMC  
To:   pratt at MIT-AI  
I have indeed switched to MACLISP for my own work.  I'll use it
in class too if I can get a version that works in TOPS-20 alias TENEX.
RPG can tell you the state of CGOL here; I haven't tried to use it,
because MACLISP was enough learning effort for the moment, but I
have one question about defining syntaxes.  What about error handling?
That's what really killed the FOL use of MLISP's syntax reader.
I'm sending you my publication language and am also a bit reluctant to
teach three languages.  That's why I dropped using RLISP and MLISP
in class.  Also when we switched to ILISP, there were just too many
manuals.  By the way, what's the MACLISP manual situation?

	Corky Cartwright has just finished a thesis about proving
"Typed Lisp" programs correct.  He has a prover, but what interested
me was that he really got proofs of total correctness entirely into
first order logic, and  I have been able to extend his method to
proving non-termination.  I need to look at your modal approach again,
but now it seems to me that vanilla-flavored first order logic is
just fine.

∂11-Jan-77  1949	JMC  
CC:   boyer at SRI-AI, moore at PARC-MAXC, DCL, RSC  
I would like an informal discussion of the theorem prover.  It sounds
like an advance all right.  Apart from that, Cartwright's thesis gets
total correctness of LISP nicely into first order logic, and I have
an extension of his results that put proving non-termination into the
same framework as well as what I think is a slight cleanup of his
results.  My ideas concern the logical formalism - not heuristics,
but now for the first time, I understand a good first order theory of
partial functions.  I do not exclude the possibility that others have
understood it previously.

	Cartwright is here for a week or so, and we should meet while
he is here.

∂11-Jan-77  1425	JMC  
To:   RSC    
Please phone me.

∂11-Jan-77  0053	JMC  	The bug. 
To:   RWW    
The step that fails on corky.dmp[w77,jmc] is "taut nil APP v = v 2 3 4;".

∂10-Jan-77  1917	JMC  	PUB Problem   
To:   LES    
Could you look at LSPLUG.PUB[W77,JMC] and perhaps at LSPLUG[W77,JMC].
The former is a file produced by a program TJ6PUB from the latter
which is an M.I.T. memo of Vaughan Pratt's praising LISP.  I want
to print the memo for local distribution.  Dick Gabriel used
the program written by Mitch Model two years ago.  What I would like
to know is what is garbage in LSPLUG.PUB so I can edit it out.

∂10-Jan-77  1505	JMC  
To:   * 
Please do not log in here on Datamedias and then go out on Telnet.
Use the TIP directly.  Telnet ties up resources non-trivially.

∂10-Jan-77  0203	JMC  
To:   LES, RWW, TW
Carlstrom is coming Tuesday morning.

∂09-Jan-77  2119	JMC  
To:   PMF, TED    
Disregard previous message.

∂09-Jan-77  2102	JMC  
To:   PMF, TED    
After a power failure my imlac wonIt go on.

∂09-Jan-77  1754	JMC  	Please print  
To:   PAT    
Please make 5 copies of LSPDOC.XGP[206,LSP] and send one to Barbara
Partee at the Behavioral Sciences Center and another to
Vaughan Pratt at the M.I.T. AI Laboratory.

∂09-Jan-77  1658	JMC  	LSPLUG   
To:   RPG    
Vaughan Pratt has produced a brief for LISP as a departmental programming
language that I have FTPed as LSPLUG[W77,JMC].  It is in some M.I.T.
PUB like form.  Do you know how to print such things?

∂09-Jan-77  1657	JMC  	LSPLUG   
To:   pratt at MIT-AI  
	Thanks for the pointer to LSPLUG.  I shall print it out when
I get someone to tell me how to do the font control and circulate it
in the department.

	You omitted one virtue - the fact that the internal form of
LISP programs are the same as LISP data - a feature it shares only
with machine language and which is responsible for many of the other
virtues.

	I will send you a copy of my notes on LISP which contains a
proposed publication language.  The language is complete for pure
LISP but is not yet fully defined for PROG, RPLACA, etc.

∂09-Jan-77  0133	JMC  
To:   RWW    
Thanks for the fix.  I'll write up my suggestion tomorrow.

∂08-Jan-77  1528	JMC  	bug in distrib
To:   RWW    
In trying to use Corky's method, I met a bug in dstrib that
eventually took the following trivial form.
"distrib λX.P(X) IF TRUE THEN Y ELSE Z" works as in INFO.FOL, but
"distrib λX.(P(X)) IF TRUE THEN Y ELSE Z" gets a meaningless error message.
Please fix it.  The example I really want is
distrib λu.(u = nil APP v) IF nil = nil THEN v ELSE CONS(CAR nil,CDR nil APP v)
which is part of a proof that append (called APP) is total.
The step should work in a bare FOL after the commands fetch corky.ax[w77,jmc];
∀e APPEND nil v;.

∂07-Jan-77  1342	JMC  
To:   PMF    
Thanks.

∂07-Jan-77  0019	JMC  	Telnet from datamedias  
To:   JBR, ME
I notice some people come in on Datamedias and go out on Telnet.
How expensive is this for us in system load?  Shouldn't we make
them use the TIP?

∂06-Jan-77  1550	JMC  
To:   LES    
Good memo.

∂06-Jan-77  1445	JMC  
To:   PMF    
Scott Daniel is the man at Suppes who knows about LISPs.

∂04-Jan-77  2309	JMC  
To:   PMF    
See bb.doc[206,lsp]

∂04-Jan-77  2006	JMC  	Grumble about notices   
To:   TED    
There is a deplorable tendency to totally uninformative X is down,
Y is dead notices.  The users would be less anxious if a bit
more effort to be informative were made.


∂03-Jan-77  1252	JMC  
To:   REM    
You were the largest single user in December.  Unless you can
convince me that the the AI Lab gets a commensurate benefit, you
must stop.

∂03-Jan-77  1119	JMC  
To:   PMF    
There has been a last minute change.  CS258 will be taught by
Zohar Mannna.  I had planned to teach it without a TA, but he
may want one.  I will teach CS206 again in the Spring and will
need a TA.

∂02-Jan-77  1712	JMC  
To:   VEW    
Call Jane Bunin at 849-0546.

∂01-Jan-77  2245	JMC  
To:   feigenbaum at SUMEX-AIM    
Zohar agrees.  I think it will make him better known to students.

∂01-Jan-77  1210	JMC  
To:   LES    
Do you understand what JAB has done and what we should do now?

∂01-Jan-77  0220	JMC  
To:   PMF    
I have forgotten precise procedure for reloading.

∂01-Jan-77  0211	JMC  	<ctrl><break> 
To:   PMF    
It doesn't work on my terminal after ty\dig.  This repeats a messaged SENDed
in case you didn't get it.

∂30-Dec-76  2034	JMC  
To:   WP
Friday 3pm if that's ok.

∂29-Dec-76  1110	JMC  
To:   PMF    
How does that helpp?  Is there any way of making it think it's talkin
to an Imlac when it is in a user progrr prog ignoring <call>?

∂28-Dec-76  1447	JMC  
CC:   TOG, LES, JBR    
I'm sorry, but there is no justification for making you a user.
Please don't use the machine unless and until you have something
useful to the Lab to do and can get someone to request an account
on your behalf with a convincing argument that the Lab will benefit.
The request is not just a formality and will be looked at critically.
Please delete your files.

∂28-Dec-76  1414	JMC  
To:   RDR    
Please check with me before you forward things of mine.  I very much
wanted Ralph to get the facts absolutely straight before sending it
outside Stanford.  I fear I have been unfair to D.E.C. in some
respects and have neglected to criticize others.  If they see it in
a wrong form, they will be naturally inclined to seize on whatever
criticism is unfounded and ignore the well-founded criticism.

∂28-Dec-76  1408	JMC  
To:   DEW    
 ∂28-Dec-76  1059	FTP:Nilsson at SRI-AI	proposed seminar  
Date: 28 Dec 1976 1058-PST
From: Nilsson at SRI-AI
Subject: proposed seminar
To:   jmc at SU-AI, eaf at SUMEX-AIM, buchanan at SUMEX-AIM,
To:   davis at SUMEX-AIM, engelmore at SUMEX-AIM, ccg at SU-AI,
To:   tw at SU-AI, les at SU-AI, luckham at SU-AI,
To:   nmartin at SUMEX-AIM
cc:   nilsson, friedland at SUMEX-AIM, moore at SU-AI,
cc:   wilkins at SU-AI



     I am proposing to organize a seminar on automatic problem solving as
described below.  I would be interested in hearing about ideas for topics
from you and/or your students.


                           Nils Nilsson

                   SEMINAR ANNOUNCEMENT

                   WINTER QUARTER, 1977

			CS  320

		ARTIFICIAL INTELLIGENCE SEMINAR:
              RESEARCH IN AUTOMATIC PROBLEM SOLVING



ORGANIZER:  Nils J. Nilsson

TIME AND PLACE: Mondays 
                Serra House Conference Room 
                3:30 pm
		(First meeting January 10,1977)

	Several ongoing Ph.D. dissertations as well as other project work
in Artificial Intelligence involve research in Automatic Problem Solving.
Examples include work in automatic theorem proving, automatic programming,
expert systems and common sense reasoning systems.  The purpose of the
proposed seminar is to encourage discussion of the various ideas being 
pursued and to permit us to find out what each other is doing.

	The seminar will meet once a week, and can be taken for one unit of
credit. Each week a volunteer speaker will talk about his/her research. 
It is possible that some topics will take more than one meeting to complete.

	If you are interested in participating and would like to volunteer
to discuss your project, please let me know as soon as possible. We  
especially need someone willing to start the series off on Jan. 10.

						Nils Nilsson
				SRI Phone:  326-6200 ext. 2311
				ARPANET:    Nilsson @SRI-AI






-------

I think I would like to give a seminar entitled "Problem solving,
pattern recognition, and parsing", but not till the middle of the
quarter.  Dave Wilkins should give a seminar also.

∂28-Dec-76  1407	JMC  
To:   nilsson at SRI-AI
I think I would like to give a seminar entitled "Problem solving,
pattern recognition, and parsing", but not till the middle of the
quarter.  Dave Wilkins should give a seminar also.

∂28-Dec-76  0020	JMC  
To:   PMF    
Once in a while, the WHO line covers the current line.

∂28-Dec-76  0007	JMC  	Information, more thereof    
To:   TOG    
I need:
address and phone
connection with Stanford
What do you propose to do?
Are you working with someone in the
Lab who will say that good things
will come of giving you an account?

∂24-Dec-76  1655	JMC  
To:   JAB    
Sounds fine to me.  Talk to Jeff too.

∂24-Dec-76  1053	JMC  
To:   RWG    
ok

∂24-Dec-76  1039	JMC  
To:   RWG    
Are you coming this evening?

∂24-Dec-76  0041	JMC  	Imlac    
To:   PMF    
<ctrl><break> doesn't work to stop typeout and <break>W doesn't turn off
who line.  Is this a feature?

∂23-Dec-76  2109	JMC  
To:   HPM    
Yes, bring them.

∂23-Dec-76  2103	JMC  
To:   HPM    
Would you like to come and eat goose tomorrow (Fri.) evening?

∂23-Dec-76  0043	JMC  	Rosenschein paper  
To:   TW
Was it you who gave me How does a system know when to stop inferencing?
by Stan Rosenschein?  There seems to be some very interesting ideas in
it.

∂22-Dec-76  0057	JMC  	Pressburger   
To:   PAT    
Here is a letter to him.

Dear Mr. Pressburger:

	I have been unable to discover what you are doing for us
as a research assistant.  Please telephone me as soon as possible.

∂21-Dec-76  1442	JMC  	Partee   
To:   TW
Barbara Partee has three readings for "Sam thinks that Mary wants him
to have an affair with Bill's wife", and has challenged me to represent
them in my first order concepts system.

∂21-Dec-76  1238	JMC  
To:   PMF    
Well, I'm impressed and will try it out right away.

∂20-Dec-76  2321	JMC  
To:   LES    
Have you pursued Pressburger?

∂20-Dec-76  2311	JMC  
CC:   JAM, LES    
I am handling requests like that of extending MLB, but I need to talk

∂16-Dec-76  1155	JMC  
To:   CG
Are there any copies of the LISP notes left, and if so where?

∂16-Dec-76  0200	JMC  	Editor in Maclsp   
To:   RPG    
I would like to be able to change the CE to a certain function of it,
edit this altered expression, then apply the inverse transformation
to get a change CE as part of the function being edited.  Example:
if I can pseudoprint the CE so that the parentheses become the
atoms L and R, I can correct parenthesis misprints much more easily.
It isn't clear whether this is feasible.

∂15-Dec-76  1519	JMC  	samefringe    
To:   carl at MIT-AI   
SAMEFR[F76,JMC] contains my current thoughts on the subject.
The obvious algorithm is modified to treat the arguments more
independently, and a new analysis shows that the number of
cells used is only equal to the length of the car-chain
to the atom.

∂14-Dec-76  1056	JMC  
To:   REP    
I'm pretty sure something can be worked out.

∂13-Dec-76  0248	JMC  
To:   DCL    
Pas encore.  Demain, peut-etre.

∂13-Dec-76  0035	JMC  	samefr   
To:   MFB    
I am now doubting Hewitt's claim that it can be done with storage proportional
to the logarithm of the depth.  SAME.LSP[F76,JMC] has a solution with storage
proportional to the depth.  (ABOT U) prefixes the successive car's of (CAR U)
to U until an atom is reached interspersing the items with the atom A.
(NEXT U) takes a chain of successive subexpressions - beginning with an 
atom and interspersed with A's and D's and finds the next such chain.
SAMEFRINGE2 initializes SAME2 which goes through the two structures
indexing them with NEXT and comparing the atoms.  The A's and D's could
be omitted if it weren't for possible merging structures.  The amont of
storage required is proportional to the longest CAR-CDR chain in the arguments
which could be the length of the list in case of lists of atoms.  Note that
no storage is required for the items in U and V since they all pre-exist -
only for the top levels of these lists.
e$
$

∂12-Dec-76  2214	JMC  	samefringe    
To:   MFB    
See page 248 of my mail file for Carl Hewitt's commentary.  The solution
involving the logarithms presumably involves using a pointer into each
tree as a variable.  Since the pointers need specify only binary branching,
they could be stored in single words it the tree h depth less than 36
and the depths were stored in other variables.

∂12-Dec-76  1455	JMC  
CC:   future-net at OFFICE-1, LES, REG, JBR
COMPATIBILITY OF A FUTURE ARPANET AND A FUTURE DIALNET

	Here are some comments relevant to the future of the ARPAnet in
reply to the message from Elizabeth Feinler of SRI (FEINLER@OFFICE-1).

	We at SU-AI have been considering the possibility of using the
dial-up telephone network to achieve a subset of the capabilities of
ARPAnet.  The problem is that the ARPAnet is too expensive for many of the
people with whom we would like to communicate, and many of them have to
little interest to the Defense Department to justify connection to
ARPAnet.

	The idea is to equip computers with telephone dialers (SU-AI has a
dialer rented from the phone company) and to develop protocols allowing
communication among any computers equipped with telephone connections and
implementing the protocols.  We call this proposal Dialnet.  Like ARPAnet,
Dialnet would provide for person-to-person messages, file transfer, and
terminal access.  As with ARPAnet, the host could limit this access.
While the facilities would be similar, the data rates would be much
smaller, and one dialer would support only one interaction at a time.
Unlike ARPAnet, any installation that implemented the system would be
automatically a "member" and a user would only have to know the telephone
number of the installation, the login name of the recipient and a password
(if required) in order to communicate.

	We are considering two levels of Dialnet protocol capable of
communicating with each other.  The first is for time-sharing systems, and
the second is for single user systems down to the level of the hobbyist
computer.  The small computer would need to be capable of answering the
telephone when unattended for the message service to be useful.

	It has occured to us that the Defense Department also has
installations for which the cost of an ARPAnet connection is not justified
by the size of computer or its expected frequency of network use.  For
these installations, Dialnet would also be appropriate.  Moreover, it
seems that Dialnet and ARPAnet should be compatible, and this
compatibility is readily achieved if the need for it is taken into account
in the design of ARPAnet.

	The requirements for compatibility seem to be the following:

	1. The protocols are divided into high level protocols that know
about messages, file transfer, etc. and those that know about
communication, error control, etc.  Communication between the high level
and low level programs it done by protocols that don't assume packet
switching and are compatible with either a packet switching or a dial-up
connection.

	2. The ARPAnet programs which are capable of handling multiple
dialogs are also capable of communicating with Dialnet programs that can
handle only a single dialog.

	There are many issues concerning error control, flexibility in
communication rate, and core protocols that remain to be identified and
resolved.  We think the basic facility should be that a program in one
computer can ask to have a program in another computer waked up and then
exchange messages with it.

	The Stanford AI Lab is currently soliciting NSF support for
implementing Dialnet, but it will probably be done anyway at a slower pace
even if support isn't forthcoming, because there is much aspiration for
ARPAnet like facilities among people who use computers that will never be
part of ARPAnet.

	The Defense Department should co-operate because

	1. It has developed much of the basic technology that will be
used.

	2. It will need Dialnet-like facilities in some form for smaller
DoD computers.

	3. ARPAnet users will need to communicate with non-ARPAnet users.

	4. It can relieve some of the pressure to join ARPAnet on the part
of installations which are either too small or whose DoD connection is too
small to justify it.

	Replies to this communication should be sent to

MCCARTHY@SU-AI

Professor John McCarthy
Artificial Intelligence Laboratory
Stanford University
Stanford, California 94305

(415) 497-4430

∂11-Dec-76  1335	JMC  
To:   PAT    
I am getting unhappy about the number of times I have to ask
for things.  Please bind my MACLISP manual today.

∂10-Dec-76  2030	JMC  
To:   LES    
How about bureau.1[let,jmc] as a computer use request form?

∂10-Dec-76  1458	JMC  	samefringe    
To:   MFB    
I don't see how your samefringe could possibly work with arguments
(a.(b.c)) and ((a.b).c), and, in fact, it doesn't.  Even if repaired,
it would still suffer from the same problem as mine.  Namely, if its
arguments were referred to elsewhere in the program, it would wrongly
modify them.

∂09-Dec-76  1507	JMC  
To:   RDR    
I want to check with Ralph before I post it, because it may be
that a bunch of eager users would interfere with getting the machine
ready for Winter quarter.  Assuming he assents, we can post it - or
a revised version - this weekend.
Incidentally, I check mail in LOTS frequently enough for that to
be reliable.

∂09-Dec-76  0115	JMC  
To:   CG
Better check the change, because I tried it right away.  I don't know
where the extras are, so in case I can't find it tomorrow morning,
please xs theory.xgp[f76,jmc] and xerox a couple.

∂08-Dec-76  0104	JMC  	dialer   
To:   REM    
Please lay off the dialer unless you can convince me that there is advantage
to the AI Lab in your using it.

∂08-Dec-76  0045	JMC  
To:   REM    
Please minimize use till end of next week because of end quarter rush.

∂07-Dec-76  2326	JMC  
To:   sigart at CMU-10B
	The following paragraphs are a reply to Grosch in honor of
his office.  However, if you are tired of the whole matter, I won't
be offended.

	I doubt Herb Grosch can find a recent literate use of "moralistic"
that isn't pejorative.  Three of the four examples in the OED are
pejorative, e.g.  "Such an action is moralistic rather than moral, for it
has not been prompted by the sentiment of goodness".

	Like Weizenbaum, Grosch deigns neither to justify nor delineate
his judgments.  He substitutes bluster for moral argument, and he won't
tell us whether houseflies are to be protected just from connection to
computers or also from general biological experimentation, medical
experimentation, feeding to pet lizards, and from fly swatters.

	I hope someone will run for President of ACM on a platform of
literacy, morality (not moralism), and intellectual coherency.

∂07-Dec-76  2143	JMC  
To:   carl at MIT-AI   
	I haven't been following the SAMEFRINGE controversy fully, but
the following seems to be a reasonable pure LISP solution.  No extra
storage is used in spite of the CONSes if SAME is compiled iteratively,
because whenever SAME conses, it forgets a variable, remembering only
its CAR and CDR so that the storage required remains constant.

	If no such simple example has been published so far, I suppose
I'll send this to SIGART.  I await your advice.



(DEFPROP SAMEF
 (NIL SAME SAMEFRINGE)
VALUE)

(DEFPROP SAME
 (LAMBDA(X Y U V)
  (COND	((ATOM X)
	 (COND ((ATOM Y)
		(AND (EQ X Y)
		     (COND ((OR (ATOM U) (ATOM V)) (EQ U V)) (T (SAME (CAR U) (CAR V) (CDR U) (CDR V))))))
	       (T (SAME X (CAR Y) U (CONS (CDR Y) V)))))
	((ATOM Y) (SAME (CAR X) Y (CONS (CDR X) U) V))
	(T (SAME (CAR X) (CAR Y) (CONS (CDR X) U) (CONS (CDR Y) V)))))
EXPR)

(DEFPROP SAMEFRINGE
 (LAMBDA (X Y) (SAME X Y NIL NIL))
EXPR)

∂03-Dec-76  1234	JMC  
To:   CCG    
I'm not in touch.

∂02-Dec-76  2302	JMC  
To:   MRB    
The technical area is where your help will be useful.  I
think perhaps I'll get Feigenbaum to sound them out as to whether the
proposal will fly.

∂02-Dec-76  2258	JMC  
To:   mrb at PARC-MAXC 
The technical area is where your help will be useful.  I
think perhaps I'll get Feigenbaum to sound them out as to whether the
proposal will fly.

∂02-Dec-76  0930	JMC  
To:   gjs at MIT-AI    
No, could you U.S. Mail me a copy of AI 380?

∂01-Dec-76  2147	JMC  	Commodore calculator    
To:   PMF    
The gamma function on the Commodore doesn't give exact results for integer
arguments and the factorial does.  If the gamma argument were gimmicked to
give exact results for integer arguments, it would have jumps at the integer
values which might have worse consequences.  Hever, the algorithms I can
think of for the gamma function would give exact results for integers, so
I don't see what they're doing.

∂01-Dec-76  2057	JMC  
To:   RPG    
JBR says stack I-O has been in this system for three years.

∂01-Dec-76  1652	JMC  
To:   gls at MIT-AI    
Thanks, I just didn't have it at hand.

∂01-Dec-76  1644	JMC  
CC:   gls at MIT-AI, gjs at MIT-AI    
I understood but misprinted and further didn't have the documentation.
However, I didn't think of making the function the value of the
labels, and this meets my objections - at least the practical
objection.  The possible further objections would probably apply
to the original form of label also.

∂01-Dec-76  1542	JMC  	SCHEME   
To:   GJS at MIT-AI, GLS at MIT-AI    
	The most important of the comments on SCHEME that got
lost in the mail was a grumble about LABELS.  The result of
defining a function by LABELS is not syntactically a function
symbol since it has to be provided with its arguments.  Therefore
it cannot serve as the f in maplist[u,f] (of which M.I.T. 
reversed the arguments many years ago).  Instead of writing
(MAPLIST U (FUNCTION (LABEL FF (LAMBDA (X) (COND ((ATOM X) X)
(T (FF (CAR X)))))))), one must write (as I understand it)

(MAPLIST U (FUNCTION (LAMBDA (X) ((LABELS (FF (LAMBDA (Y)
(COND ((ATOM Y) Y) (T (FF (CAR Y))))))) X)))

which has an unnecessary level of abstraction.  This will make
the semantics of SCHEME more difficult to axiomatize also.

∂30-Nov-76  2143	JMC  	MONTAGUE BOOK FOR JOHN  
To:   TW, WINOGRAD at PARC-MAXC  

∂30-Nov-76  1413	JMC  
To:   CCG    
Yes, why should KRD do his PUBbing here?

∂30-Nov-76  0314	JMC  
To:   LES    
ok, do it for a while

∂30-Nov-76  0312	JMC  	Carl Farrell  
To:   CCG    
Why did you ask us to leave Carl Farrell on the list of users.
Use of the XGP is not ordinarily a legitimate reason unless
there is a benefit to the AI Lab.  Sometimes I think you don't
mind overcrowding this machine.

∂30-Nov-76  0311	JMC  
To:   LES    
We shouldn't give our machine away for documentation purposes.
Please ask me in the future.

∂29-Nov-76  1943	JMC  
To:   RWW    
I thought we agreed to get him to take over simplification.

∂27-Nov-76  2154	JMC  	bug in spindl 
To:   REM    
Please explain whether what should be done about files previously spindled
and whether the new version will properly uncrunch and unspindle old files.

∂26-Nov-76  0141	JMC  
To:   CDR    
I sent you a message, but it evidently got lost.

∂26-Nov-76  0013	JMC  
To:   RWW    
It sure would be nice to have those subsitution rules for conditional forms.

∂25-Nov-76  1109	JMC  	Abuse of service level system.    
To:   BPM    
Jeff considers that you have been abusing the service level
system to get an advantage over others.  What exactly are
you doing now and what have you been doing?

∂25-Nov-76  0204	JMC  	PROOF    
To:   RWW    
APPNIL.PRF contains a proof that U*NIL=U from the axioms LIST.AX (not LISP.AX).
Note that the DISTRIB command of step 3 wasn't listed with the step.  There
are some mysteries about why I couldn't use TAUTEQ sooner to get the conclusion
of step 12.  It is pretty ugly to have to call NIL NILL.
Associativity of * is the next step.

∂24-Nov-76  2208	JMC  
To:   PMF    
Thanks, I'll try it.

∂24-Nov-76  2149	JMC  	Dialer   
To:   TED    
I have been trying to use the dialer to call into LOTS without success.
Can you show me how to get around its flakiness, or better, fix it.

∂24-Nov-76  2058	JMC  
To:   RWW    
Not unless you've debugged it.

∂24-Nov-76  1555	JMC  	Fairy story   
To:   newell at CMU-10A
Congratulations on FAIRYTALES.  How wonderful that it
is so easy to indulge my impulse to congratulate you.

∂24-Nov-76  1325	JMC  
To:   RWW    
More bugs:  Why does the last step lose?

.RU PROOF

Saving input on: BACKUP.TMP

*****SHOW PROOF;

*****∧I INDUCTION2;

1 ∀X.((NULL X∨(CDR X*NILL)=CDR X)⊃(X*NILL)=X)⊃∀X.(X*NILL)=X  

*****∀E APPEND X,NILL;

2 (X*NILL)=IF NULL X THEN NILL ELSE CONS(CAR X,CDR X*NILL)  

*****;

3 IF NULL X THEN NILL ELSE CONS(CAR X,CDR X*NILL)=(X*NILL)≡IF NULL X %
THEN NILL=(X*NILL) ELSE CONS(CAR X,CDR X*NILL)=(X*NILL)  

*****TAUT 3:#2≡(NULL X⊃=(NILL,X*NILL))∧(¬(NULL X)⊃=(CONS(CAR X,(CDR X)*NILL),X*NILL)
);

Not a tautology

*****↑C

∂24-Nov-76  0013	JMC  	bug 
To:   RWW    
The current problem is that distrib balks at
DISTRIB λX.(X=X) IF TRUE THEN Y ELSE Z;

∂23-Nov-76  2145	JMC  	BUGS
To:   RWW    
How come you don't debug things.  The conditional expression examples
in your writeup don't work.  The first of them gave
arg
UNDEFINED FUNCTION

∂23-Nov-76  1009	JMC  
To:   RWW    
EMSNOPROPFUNC is undefined.

∂22-Nov-76  1217	JMC  
To:   gjs at MIT-AI    
It was a batch of comments on your second SCHEME paper, and I'll try
to find them and resend them.  Please acknowledge this message, however,
because, as far as I can see, the messages were properly sent, and
I don't want to be shouting down a well again.

∂19-Nov-76  0252	JMC  	bug 
To:   RWW    
TAUTEQ CAR X = CAR(X);
produced 
PROPSYM
UNDEFINED FUNCTION

∂16-Nov-76  0235	JMC  
To:   PAW    
What happened to the copies of MACLISP I asked for.

∂16-Nov-76  0056	JMC  	FACT
To:   gjs at MIT-AI    
Note that your FACT is not the same algorithm as the recursive definition in
the sense that if multiplication were not commutative and associative, it
mightn't give the same answer.

∂16-Nov-76  0050	JMC  	IF  
To:   gjs at MIT-AI    
Your (IF p a b), indeed preferable to COND, was originated by Harold
McIntosh.  Yates or Hawkinson could give a precise reference.  Had I
but known I was inventing a programming language that would be used
when I wrote eval as a rival to universal Turing machines, I would
have used a more convenient notation.

∂16-Nov-76  0036	JMC  	LABEL    
To:   gjs at MIT-AI    
Your LABEL construction, unlike mine, unless I misunderstand, doesn't
define a function object (unapplied as yet) that can replace a function
letter in an arbitrary formula or can serve as a functional argument.
Mine allows simultaneous recursion, because if a LABEL named function
needs another, another LABEL definition can be used inside the first
with the outer function name appearing as a free function variable.
Each occurrence of an inner function may require a separate LABEL
expression which is unaesthetic, but it works.  Thus

(LABEL FOO (LAMBDA (X) (COND (((LABEL BAZ (LAMBDA (Y) (COND ((EQ (FOO Y) ... etc.

I seem to be reading your paper both ends toward the middle.

∂16-Nov-76  0027	JMC  	Alonzowins    
To:   gjs at MIT-AI    
I think you are mistaken in saying that λ-calculus was developed as an
axiomatic system of logic.  In Church, (Calculii ...), it is used solely
to describe functions, and a formal language for making assertions
about functions is not given.  It is my impression that Church wanted
a logical language, but the natural one proved inconsistent.

I agree completely with the final remark of your paper that eliminating
bad constructs is unimportant compared to inventing or discovering good
ones.

∂16-Nov-76  0015	JMC  	Jensen   
To:   gjs at MIT-AI    
I remember meeting the afore-mentioned Jensen at the Regnecentralen
in Copenhagen in early 60's.  Any Dane, especially Peter Naur
will know him.  I don't know if the agreement between ACM and their
computer society provides for extradition.

∂15-Nov-76  1525	JMC  
To:   JAB    
yes

∂14-Nov-76  2120	JMC  
To:   RDR    
No, something quite different.

∂12-Nov-76  2330	JMC  
To:   JFR    
p.31, line -9 its → it's

∂12-Nov-76  2131	JMC  	SAIL manual   
To:   JFR    
The term "disjunctive expression", p. 26, for an expression whose major
connective ∧ is a misnomer.  Disjunctive means "or"; an expression with
∧ would have to be a conjunctive expression.

∂12-Nov-76  1453	JMC  
To:   reddy at CMU-10B 
Ohlander's thesis.

∂12-Nov-76  1423	JMC  
To:   * 
If you have trouble fitting into your disk quota, read SPINDL.REM[UP,DOC]
for ways of compressing and decompressing files.

∂12-Nov-76  1402	JMC  
To:   RDR    
Why not adopt a less paranoid literary style.  Perhaps you suspect they
haven't yet solved the heating problem.

∂11-Nov-76  2210	JMC  	Telco data sets    
To:   REG    
Have ←ou considered the option of getting started in one of the locations
by renting data sets from Telco?  We could stop when our own cables were
ready.

∂10-Nov-76  2329	JMC  
To:   REG    
I'm back.  Didn't get your note till now and didn't have time to scream anyway.

∂09-Nov-76  2139	JMC  
To:   PAW    
OK to change Todorovich appointment.

∂08-Nov-76  1004	JMC  
To:   JFR    
Naur is misspelled Nauer on p.1 of SAIL.

∂05-Nov-76  2343	JMC  	Comments on your "Future Studies" paper.    
To:   licklider at MIT-DMS  
The file LICKL2[F76,JMC] is close to the version that will be mailed to
Dertouzos after I sit on it for a few days.  It is quite different from
the earlier version but contains many of the same grumbles.

∂05-Nov-76  2246	JMC  
To:   RDR    
They are a cost center in the sense that the University adds up all
the costs of operating it and charges accordingly.  I don't know
that Reprographics has a different status.
Some days ago, I asked you what you were doing for LOTS and haven't
received a reply yet.

∂05-Nov-76  2159	JMC  
CC:   RDR, REG    
LOTS isn't a business - by charter, while SCIP is.  The University
won't let us go into the terminal business merely to compete with
SCIP.  I might prefer to run a competitive computation center,
because I also think I could do better than SCIP, but we can't.
The justification would have to be serving our particular clientele,
but the meeting Wednesday showed no-one but JAB who wanted to rent
a terminal, and he can be treated as a special case.  If the desire
to rent terminals appears later, we can consider it concretely at
that time.  I have no objection to your trying to drum up interest
among students - but not on our time.

	The worry that SCIP would mismanage and overcharge for an
XGP has been mentioned by everyone to whom I have spoken.  Once the
ducks are lined up - in both supply and demand, we can dicker with
SCIP about what the charges would be.  If we don't like what we hear,
or if we can't get a reasonably definite answer, then we can look for
another solution - but it won't be LOTS alone, because the University
won't buy it for us, and it won't be free.  By the way, SCIP running
the XGP wouldn't mean attaching solely to a present SCIP computer and
wouldn't necessarily involve a PUB on a SCIP computer except to serve
SCIP's customers.  The minicomputer running the XGP would get files
ready for printing by telephone directly from all its client computers,
and the 168 would just be one of them.  The reason for preferring SCIP
is that they already employ people to remove output from printers and
ought to be able to operate one more printer more cheaply than other
installations, especially since it ought to be 24 hours a day.  If this
idea proves false, alternatives exist.

	Incidentally, the problem of an inexpensive PUB for LOTS exists independently
of whether the printer is a Printronix or an XGP and independently of
whether it is operated by LOTS, SCIP or some co-op.  If you could
co-ordinate some students into volunteering, it would be worthwhile.
TJ6 may be similar to RUNOFF; the M.I.T. people seem to have mainly
switched to PUB for their fancier documents.

∂05-Nov-76  2134	JMC  	Communication 
To:   REG    
I mentioned my worry about getting a good solution to the remote terminal problem
to Franklin.  His only suggestion was Bill Sanders at IMSSS whose expertise
and experience in such matters he praised.  I suggest you talk to him.

∂04-Nov-76  1902	JMC  
CC:   DSB, REG    
I am pursuing the idea of getting such a printer - somewhat better -
for Stanford, but not for LOTS's exclusive use.  There are several
problems: Stanford won't pay anything for it until it is determined
that LOTS can meet its main goals, i.e computing for courses and
independent student research.  Even then, the demand for high
quality printing goes beyond LOTS to paying users, so that would
have higher priority with Stanford.  If someone were to give LOTS
such a printer, we would accept it, of course, but we might have
to severely limit the use of PUB and similar programs in order to
serve the computing ccurses and the independent research.  We are
authorized to experiment with letting students unused to computers,
e.g. in an English class, use it for reports to see if it proves
popular.  Naturally, one expects that students who are already
computer users will prepare reports with LOTS using the Printronix
which is better than the average student's typewriter.  LOTS will
get its own XGP, but I am afraid it can't be soon.

	Now, as to what you can do to hurry the day.  We need an
accurate estimate of the costs of editing and using RUNOFF or PUB.
We know PUB is very expensive and also that it can be made much
faster by recoding it in machine language with better string handling
programs.  A better editor, more friendly to the unwashed than SOS,
would also help.  Again it must be economical.  My main idea about
how to achieve economy is that it should have relatively big
commands and make heavy use of what passes for a line editor in TENEX
if there is such a thing, but this idea may not be correct.

∂04-Nov-76  0028	JMC  
To:   moon at MIT-MC   
Done.

∂03-Nov-76  0040	JMC  
To:   JAB    
It's a question of whether the University will buy terminals for re-rental;
not how to administer it.

∂02-Nov-76  2154	JMC  
To:   LOTS.DIS[P,DOC]:;
RDR doesn't have that exactly right.  Prof. Massy expressed doubts
about terminal rental, but because of time shortage, the question
was continued till next time which will be a meeting with the LOTS
Advisory Board - Nov. 16.  In the meantime, it would be worthwhil
to have an indication of the demand at rentals between $55 and $65
per month.  The main reason why we are worried about buying more
terminals for rental now is that INTEL and other manufacturers are
coming out with integrated circuits that will do on one chip what
most of that ADM-3 circuit board does.  Therefore, Stanford might
be left holding the bag on the rental terminals as soon as a year
from now.  The LOTS terminals can be held until amortized even if
better and cheaper are available, but rental terminals may be
unattractive.  This is only a worry and not necessarily a decisive
consideration, but a rental price would reflect the worry to some
extent.

	The security problem was discussed, but it seems that
Cedar and Eng. 303 are the best we can do.  Those most worried
can concentrate their late night activities in SCRDT, but frankly
I think it will only be marginally less worrisome.  When the
terminals are heavily used, there will be plenty of people, and
otherwise SCRDT can be used by the worried.

	The computer was shipped Friday Oct. 30 by truck; arrival
date depends somewhat on what else is on that truck.

	Prof. Massy agreed to the purchase of a Printronix printer
for each terminal room which will make listings easily available.

∂02-Nov-76  0101	JMC  
To:   CG
Can you write out and distribute sol'ns to midterm problems tomorrow?

∂01-Nov-76  1916	JMC  
To:   ef at MIT-AI
I have put you on the LOTS distribution list.  Items are somewhat random.

∂01-Nov-76  1603	JMC  
To:   licklider at MIT-DMS  
The file LICKLI[F76,JMC] contains an evolving draft of comments on
your paper for the future study.  The first page is just notes and
won't appear per se, but  you might be interested in comparing them
with the corresponding pages.  Next comes an incomplete draft of
the comments that will be printed unless you obviate them.  I have
some technical disagreements, but mainly I think you overestimate
the desirable role of government.  The language of the printed
version will be milder.

∂01-Nov-76  1038	JMC  
To:   REG    
ok, do it.

∂31-Oct-76  1704	JMC  	terminal rental    
To:   RDR    
We can think about it now.  JAB is no? a student, so a more typical case
wodbe better for the Advisory board to think about.

∂31-Oct-76  1703	JMC  
To:   REG    
What do you think about renting JAB a terminal or even lending him one
temporarily.

∂30-Oct-76  1028	JMC  
To:   LES, TOB    
 ∂30-Oct-76  0255	LES  
Here is a note to think about.

 ∂29-Oct-76  1010	FTP:CERF at USC-ISI	soviet work on robotics  
Date: 29 OCT 1976 0952-PDT
From: CERF at USC-ISI
Subject: soviet work on robotics
To:   les at SU-AI
cc:   cerf

Les,
In a review of soviet literature, i recently came across an
article by S. Adrieyevskiy on 2Research on Robots at Kiev Institute of
Automation" printed in Pravda Ukrainy in Russian 4 July 1976,
p.3.
Rather incredible claims were made for a robot assembling a toy
automobile from parts. The robot has 2 TV eyes, range finder, movable "hands" 
so i assume more than one. The claim was the the robot got
a word description of a car and its components but nothing as elaborate
as the instructions Stanford has been using for AL. 
"the robot examines them [the parts] for 40 seconds and lays them out in
the order in which he needs them, then with fantastic speed, he
assembles the play automobile on a movable platform. and note -- this
is without specific mathematical programs, but only from a general word
description of the automobile." This sounds like pure BS tome, but do you
or John McCarthy have nay feeling for the state of the Kiev work? Articles like
this can make us look bad.
Vint
-------

Ask Tom Binford to look into it and inquire of Vint where the translation
can be found.  I doubt that Pravda Ukrainy can be found here.  The
task, complete with movable platform, sounds like what Donald Michie's
group did at Edinburgh with a toy car, etc.  Have they copied Freddy
or has the reporter or translator slipped and ascribed to them
work done at Edinburgh.  Harry Barrow at SRI might know.

∂30-Oct-76  0150	JMC  	493-5505 
To:   LES    
mIt is proposed to make this permanently 30 chars/sec.  The few remaining
model 33s can come in through the Ames Tip.  What say you?

∂30-Oct-76  0150	JMC  
To:   LES    
It is proposed to make 5505 30 chars/sec.  The few remaining model

∂29-Oct-76  2224	JMC  
To:   JAB, REG    
 ∂29-Oct-76  2138	JAB   via TYMT	terminal  

i was wondering about my request to rent a terminal. i told reg
that i would want a coupler also and would pay about 50-65/month
for a term/coupler deal. i also would honor your request to have 
it back should the need arise. let me know sooon. thanx

When do you want to rent it?  If you don't want to rent it soon
I will take my time, because there is a policy issue involved
in Stanford buying terminals to rent to people.  If I had to
give an instant answer, it would have to be no.
What do you intend to use it for.  I wouldn't ask, but if I
discuss it with Vice-Provost Massy, it will have to be on the
basis of being the first swallow of a summer, and the question
is what summer?

∂28-Oct-76  2227	JMC  
To:   RPG    
Maclisp distribution list

∂27-Oct-76  1606	JMC  
To:   CG
I think I've fixed the equation numbering stuff myself.

∂25-OCT-76  1600	JMC  
To:   JB
Fine.

∂25-OCT-76  0045	JMC  
To:   PAW    
Please tell Carolyn Tainai that I won't be at dinner today (Monday).

∂25-OCT-76  0020	JMC  
CC:   RDR, REG    
I can't help getting the impression that you spend most of your
time thinking up things for others - especially Ralph and me -
to do.  You could dispell this impression by mailing me a list
of substantive things you have been doing recently and expect
to do in the next month.

∂24-OCT-76  1735	JMC  
To:   LOTS.DIS[P,DOC]:;
I agree with SAU and will make the point to the authorities.

∂23-OCT-76  2348	JMC  
To:   REG    
Do you have an objection to saying that there will be terminals in 105 SCRDT
and 103 Engineering corner?  I have made other minor changes.

∂23-OCT-76  2129	JMC  
To:   REG    
Thanks.

∂22-OCT-76  2058	JMC  
CC:   RDR, REG    
I am also worried about whether 45 terminals are enough.  It is unrealistic
to expect human nature to change and people to do their work well
in advance of the end of the quarter.  There will be some very late
hours towards the end of the quarter.  This is another reason for
caution about terminals in dorms until we see how hard it is to
handle the courses.  Let me remind you that the university has taken
money mainly spent on course computation to set up LOTS.  I see the
largest new application to be on-line editing of reports, etc.  However,
when this application begins to compete for terminals, disk space,
and compute cycles with the courses, the University will have to
decide whether to put in some more money.  I think they'll do it, but
not until there are some users to say they like it.

∂22-OCT-76  1810	JMC  
To:   ZM
Have a good trip.

∂21-OCT-76  0906	JMC  	Phone connection to Cedar.   
To:   REG    
It seems to me that the phone company usually has enough excess capacity
for the number of lines we would need.  Has that been checked together
with the prices for the number of lines required - say 15 to 20?

∂20-OCT-76  2315	JMC  
To:   EHF    
You're welcome to look into it, but when we considered a co-ax or
microwave between the AI Lab and Campus, it seemed like the modulation
and demodulation were very expensive.

∂19-OCT-76  1508	JMC  
To:   RDR    
We have been pursuing it.

∂18-OCT-76  1110	JMC  
To:   PAW    
Please integrate Feigenbaum biography into csddis.pro[f76,jmc].

∂17-OCT-76  1817	JMC  
CC:   REG, RDR    
After thinking it over, I think we should ask to keep two rooms
in Cedar, a terminal room and a shop (unless we can get shop space
in SCRDT).

∂16-OCT-76  1713	JMC  	Stanford network addresses   
To:   feinler @ OFFICE-1    
Last names work at Stanford except when there are two with same
name, and I think they are preferable to the login initials.

∂14-OCT-76  2307	JMC  
To:   RDR    
Indeed!  If I go it will be to get ideas.

∂14-OCT-76  2004	JMC  
To:   RDR    
Also please put some in Computer Science Library.

∂14-OCT-76  2003	JMC  
To:   RDR    
I forget.  Irwin Remson was to be, but I don't think that was he.
Please ask Adams without letting him know I forgot and let me know.

∂14-OCT-76  1938	JMC  
To:   RDR    
hmm, we'll have a problem with student use then.

∂14-OCT-76  1937	JMC  
To:   QIB    
Please make sure Ralph and David get in at 7-3214 and I have Director of
LOTS added to my phone book identification but with 7-4430.  Remove
Ralph from 7-1360.

∂14-OCT-76  1532	JMC  
To:   MS
How about 4:30 in conference room, and why don't you tell Richard, etc.

∂14-OCT-76  1404	JMC  
CC:   RDR, REG    
We'll have to establish priorities.  You need to fix the terminal a bit more.

∂14-OCT-76  1401	JMC  
To:   JB
Thanks, we'll be there.

∂14-OCT-76  1028	JMC  
To:   JB
Monday or Wednesday would be fine.

∂13-OCT-76  2334	JMC  
To:   REM    
No, I don't actually object.

∂13-OCT-76  1607	JMC  
To:   REM, LES    
I don't see any point in putting the propaganda in the program.  As for
how many people use SPINDL, unless there are bugs, don't worry about it.
The availability of SPINDL co-incided with the availability of more
disk so the motivation to use SPINDL was low.  However, when the next
disk crunch comes, and it will be soon, the management will resist
increasing the disk quotas of people who don't have good arguments
for not using SPINDL.  It wwuld be worthwhile to know how much computer
time in this computer DFTP uses, but I'm not sure you can find out.

∂13-OCT-76  1533	JMC  
To:   ef @ MIT-AI 
I just received Davenport's letter about Winston and will respond.

∂13-OCT-76  1332	JMC  
To:   LES    
 ∂13-OCT-76  1330	REG  
You'd better tell Les to do something to keep NXL from being purged.
Do Something - JMC

∂13-OCT-76  0102	JMC  
To:   MJC, DEK    
Ignore previous message.

∂12-OCT-76  2322	JMC  	204 
To:   MJC, DEK    
I notes six 204 students in RUNQ.  Some soaking up to 13% of computer time.
I hope the problem isn't one in which results are proportional to computer
time taken, e.g. a depth first search, because this can swamp any computer.

∂12-OCT-76  1953	JMC  
To:   LOTS.DIS[P,DOC]:;
The fact that D.E.C.'s Vice-President for Engineering is a professor
at CMU gives them an in.

∂12-OCT-76  1952	JMC  
To:   LOTS.DIS[P,DOC]:;
The fact that D.E.C.'s Vice-President for Engineering is a professor
at CMU gives them an in.

∂11-OCT-76  1307	JMC  	Manna    
To:   feigenbaum @ SUMEX-AIM
He recently gave a talk at U.S.C., and they offered him a full
professorship at a salary of $38K for 11 months.
Why can't we move.

∂10-OCT-76  1640	JMC  	Sussman thesis
To:   PAW    
Do I have a copy of Sussman's thesis "A Computer Model of Skill Acquisition"
If not please order a copy as published by Elsevier.  I prefer ordering
from the publisher.

∂10-OCT-76  1617	JMC  
To:   PAW    
Why are the 4 Sussman papers in lib.bib absent and another sussman paper there?

∂10-OCT-76  1303	JMC  	Syntax for simplification.   
To:   BG, RWW
The syntax proposed in BG's latest memo is satisfactory, although
there is one point not clear to me.  However, if it is not difficult
to implement, I would suggest replacing

SIMPSET <simpset> ← <simpset-expr>

by 

ASSIGN <wffset> ← <wffset-expr>

or yet better

<wffset> ← <wffset-expr>

without excluding the possibility
of other assignments

<var> ← <expresion>.

My idea is that FOL should allow assignments to variables for a variety
of purposes in as general a way as is currently feasible.  In particular,
one can imagine either the user or a program using sets of wffs for
a variety of purposes.  This is a part of my general doctrine that
many kinds of command in many on-line languages including operating
systems can be regarded as assignments with complex expressions on
the right hand side.

∂09-OCT-76  2049	JMC  
To:   REG    
No!  It refers to RDR's message about writing our own versions
of the D.E.C. manuals.

∂09-OCT-76  0023	JMC  
To:   LES    
Before getting enthusiastic about dftp, compare price with local storage.

∂09-OCT-76  0004	JMC  
To:   RDR, REG    
In general, this is impractical given the manpower we have
available though it might be done for selected topics.  However,
it wouldn't unsettle them; they would simply be glad to have
some of their work done for them - unless the documentation
contained some harsh words about the reason.

∂08-OCT-76  2014	JMC  
To:   AJT    
I'll be here till about 10 and will read more now.

∂08-OCT-76  2000	JMC  
To:   MRC    
Thanks for the information;I would like to be able to try it.

∂08-OCT-76  1906	JMC  
To:   MRC    
What is DFTP?

∂08-OCT-76  1120	JMC  
To:   REG    
Keep in mind just doing it and letting them sue.

∂08-OCT-76  0129	JMC  
To:   RPG    
Well, it produces output, but PUB finds many faults with it.

∂08-OCT-76  0047	JMC  
To:   REG    
I can't help feeling we aren't getting actual work out of Roode.

∂08-OCT-76  0046	JMC  
To:   RDR    
No, we have enough people pursuing him for now.

∂06-OCT-76  2319	JMC  
To:   REG    
I will pursue all possible officials tomorrow.

∂06-OCT-76  1214	JMC  	Pretty print  
To:   CG
I tried out Littlestone's Blackboard Pretty Printer (BBPP) as
described in BB.DOC[206,NXL].  The teletype and line printer
versions seem to work (one needs 40 pages of core and 5000
words of binary program space to load the program), but
BBXGP and BBPUB, the versions that produce output for the XGP
and PUB blow up with an ILL MEM REF.  Could you try to figure out
what's wrong?

∂05-OCT-76  1342	JMC  
To:   TOB    
Please see kanade.le1[let,jmc]

∂05-OCT-76  1341	JMC  
To:   PAW    
please pub kanade.le1[let,jmc] and enclose map.

∂04-OCT-76  2259	JMC  
To:   AJT    
Wednesday afternoon

∂22-Jan-77  2039	JMC  	about messages
To:   ROZ    
If you SEND a message without also MAILing it, you run the risk
that the recipient who may not be looking will be zapped by a
system crash before he sees it.