perm filename JUN76.IN[LET,JMC] blob
sn#225042 filedate 1976-07-04 generic text, type C, neo UTF8
COMMENT ⊗ VALID 00013 PAGES
C REC PAGE DESCRIPTION
C00001 00001
C00010 00002 VALID 00107 PAGES
C00020 00003 ∂02-JUN-76 2250 TOB via AI
C00028 00004 ∂10-JUN-76 0619 FTP:J. Baratz (BARATZ @ MIT-AI)
C00050 00005 ∂13-JUN-76 0816 RDR via AMET
C00064 00006 ∂15-JUN-76 2233 RDR LOTS terminals
C00084 00007 ∂17-JUN-76 2029 FTP:LEVY at PARC-MAXC ARPA report
C00102 00008 ∂18-JUN-76 1038 TAG
C00127 00009 ∂19-JUN-76 1546 RDR Pine Hall
C00140 00010 ∂23-JUN-76 0821 FTP:LEVY at RUTGERS-10 lot.jmc directory
C00161 00011 ∂25-JUN-76 1340 RSC Cartwright's Ph.D. Orals
C00183 00012 ∂28-JUN-76 1900 LES Remember to talk to Lieberman
C00200 00013 ∂03-JUL-76 1810 REM via AMET KL10 timing for R SPINDL
C00210 ENDMK
C⊗;
COMMENT ⊗ VALID 00107 PAGES
C REC PAGE DESCRIPTION
C00001 00001
C00011 00002 ∂02-JUN-76 2250 TOB via AI
C00012 00003 ∂04-JUN-76 1444 FTP:PRATT at MIT-AI
C00014 00004 ∂04-JUN-76 2228 PLW at TTY75 2228
C00015 00005 ∂05-JUN-76 0022 PLW project description meeting
C00016 00006 ∂06-JUN-76 1248 REG
C00017 00007 ∂08-JUN-76 1244 100 : REM via AMET Chope et al
C00018 00008 ∂08-JUN-76 1312 FTP:Robert Elton Maas (REM @ MIT-MC)
C00019 00009 ∂08-JUN-76 1315 FTP:Robert Elton Maas (REM @ MIT-MC)
C00020 00010 ∂08-JUN-76 2330 REM via AMET System busyness, etc.
C00022 00011 ∂08-JUN-76 2337 REM via AMET
C00023 00012 ∂10-JUN-76 0619 FTP:J. Baratz (BARATZ @ MIT-AI)
C00027 00013 ∂11-JUN-76 0426 TED TWO ANSWERS
C00028 00014 ∂11-JUN-76 0800 FTP:J. Baratz (BARATZ @ MIT-AI)
C00030 00015 ∂11-JUN-76 2144 RDR via AMET pine hall space
C00032 00016 ∂12-JUN-76 1358 REM via AMET TROUBLE IN MAY76.IN[LET,JMC]
C00033 00017 ∂12-JUN-76 1430 REM via AMET No trouble with MAY76.IN[LET,JMC] beyond reformatting bare linefeeds.
C00035 00018 ∂12-JUN-76 2203 FTP:LEDERBERG at SUMEX-AIM Felicitations
C00036 00019 ∂12-JUN-76 2333 100 : REM via AMET IF YOU "TY FOO.SPI"
C00038 00020 ∂12-JUN-76 2341 100 : REM via AMET EMPLOYMENT
C00041 00021 ∂13-JUN-76 0314 REG
C00045 00022 ∂13-JUN-76 0816 RDR via AMET
C00047 00023 ∂13-JUN-76 1215 EHF Plans
C00048 00024 ∂13-JUN-76 1312 REG
C00049 00025 ∂13-JUN-76 1435 REG
C00050 00026 ∂14-JUN-76 0227 RDR via AMET LOTS configuration
C00051 00027 ∂14-JUN-76 0315 JMC BUDGET
C00052 00028 ∂14-JUN-76 1124 REG
C00053 00029 ∂14-JUN-76 2209 JMC Present state of LOTS.
C00056 00030 ∂15-JUN-76 1123 REG
C00057 00031 ∂15-JUN-76 1244 REM via AMET Employment
C00059 00032 ∂15-JUN-76 2233 RDR LOTS terminals
C00062 00033 ∂15-JUN-76 2307 RDR SCIP's TRAN (communications switch for terminals)
C00064 00034 ∂16-JUN-76 1012 FTP:RINDFLEISCH at SUMEX-AIM LOTS AND SPACE
C00067 00035 ∂16-JUN-76 1014 FTP:FEIGENBAUM at SUMEX-AIM ai qual grading
C00069 00036 ∂16-JUN-76 1051 JMC file directory
C00070 00037 ∂17-JUN-76 1136 RDR more LOTS accounts
C00072 00038 ∂17-JUN-76 1143 RDR continued
C00073 00039 ∂17-JUN-76 1544 EHF LOTS in general
C00077 00040 ∂17-JUN-76 1611 EHF correction to previous message
C00078 00041 ∂17-JUN-76 1658 MFB paragraph
C00079 00042 ∂17-JUN-76 2029 FTP:LEVY at PARC-MAXC ARPA report
C00082 00043 ∂17-JUN-76 1643 CCG Accomplishments for ARPA
C00083 00044 ∂17-JUN-76 2032 FTP:LEVY at PARC-MAXC ARPA report
C00086 00045 ∂17-JUN-76 2146 GF via AMET TO RDR
C00087 00046 ∂17-JUN-76 2334 DBL AM paragraph
C00088 00047 ∂17-JUN-76 2339 DCL carlstrom accomplishments
C00089 00048 ∂17-JUN-76 2327 RDR Space and RP06's
C00092 00049 ∂17-JUN-76 2351 RDR
C00093 00050 ∂18-JUN-76 0247 RWW ARPA PARAGRAPHS
C00094 00051 ∂18-JUN-76 0823 100 : REM via AMET S.I.S. JOB
C00097 00052 ∂18-JUN-76 1038 TAG
C00099 00053 ∂18-JUN-76 1230 100 : rem via AMET
C00108 00054 ∂18-JUN-76 1233 100 : REM via AMET System still losing
C00109 00055 ∂18-JUN-76 1354 REG Letter to John Leng
C00110 00056 ∂18-JUN-76 1403 EHF
C00112 00057 ∂18-JUN-76 1620 REG
C00113 00058 ∂18-JUN-76 2348 JMC Reasons for accepting delay
C00115 00059 ∂19-JUN-76 1045 FTP:Marvin Minsky (MINSKY @ MIT-AI)
C00118 00060 ∂19-JUN-76 1327 SGK
C00121 00061 ∂19-JUN-76 1544 RDR
C00122 00062 ∂19-JUN-76 1546 RDR Pine Hall
C00123 00063 ∂19-JUN-76 1550 RDR machine room /LOTS
C00124 00064 ∂18-JUN-76 1111 JMC
C00125 00065 ∂21-JUN-76 1140 REG
C00126 00066 ∂21-JUN-76 1832 FTP:FEIGENBAUM at SUMEX-AIM fellowship award for randy davis
C00128 00067 ∂21-JUN-76 2216 REM via AMET
C00129 00068 ∂22-JUN-76 1301 JMC*
C00130 00069 ∂22-JUN-76 2047 ELM LIBRASCOPE
C00131 00070 ∂23-JUN-76 0051 RDR last message
C00133 00071 ∂23-JUN-76 0157 RDR Gloria Guth
C00135 00072 ∂23-JUN-76 0821 FTP:LEVY at RUTGERS-10 lot.jmc directory
C00137 00073 ∂23-JUN-76 1615 RDR
C00142 00074 ∂24-JUN-76 1136 FTP:UNCAPHER@ISIB 2040
C00144 00075 ∂24-JUN-76 2133 RDR via AMET Meeting
C00145 00076 ∂24-JUN-76 2201 RDR via AMET MAINSAIL
C00149 00077 ∂24-JUN-76 2352 REM via AMET
C00150 00078 ∂24-JUN-76 2357 JMC
C00152 00079 ∂25-JUN-76 0640 FTP:Stavros M. Macrakis (MACRAK @ MIT-AI)
C00153 00080 ∂25-JUN-76 0700 FTP:Marvin Minsky (MINSKY @ MIT-AI)
C00154 00081 ∂25-JUN-76 1152 REG
C00156 00082 ∂25-JUN-76 1340 RSC Cartwright's Ph.D. Orals
C00157 00083 ∂25-JUN-76 2112 REM via AMET Librascope
C00159 00084 ∂25-JUN-76 2313 REM via AMET JOB ETC.
C00161 00085 ∂26-JUN-76 1615 RDR CHEAP terminals
C00163 00086 ∂26-JUN-76 1624 FTP:Mark R. Crispin (MRC @ MIT-AI)
C00168 00087 ∂26-JUN-76 1641 RDR MIT
C00169 00088 ∂26-JUN-76 1707 100 : RDR
C00171 00089 ∂27-JUN-76 1746 REG
C00173 00090 ∂28-JUN-76 0610 FTP:Benjamin J. Kuipers (BEN @ MIT-AI) Home terminal suggestion
C00177 00091 ∂28-JUN-76 1849 CCG ARPA VISIT
C00178 00092 ∂28-JUN-76 1900 LES Remember to talk to Lieberman
C00179 00093 ∂29-JUN-76 0055 RDR terminals
C00181 00094 ∂30-JUN-76 0213 RDR another LOTS person
C00184 00095 ∂30-JUN-76 0214 RDR campus dial-in terminal
C00188 00096 ∂01-JUL-76 2126 MM via AI sarah
C00189 00097 ∂01-JUL-76 2249 KRD
C00190 00098 ∂02-JUL-76 0909 REM via AMET Preliminary timing figures for job running on crippled KL/10
C00192 00099 ∂02-JUL-76 1117 FTP:Robert Elton Maas (REM @ MIT-MC)
C00193 00100 ∂02-JUL-76 1109 REG
C00194 00101 ∂02-JUL-76 1221 REG TOPS-20 sources
C00195 00102 ∂02-JUL-76 1600 JMC*
C00199 00103 ∂03-JUL-76 1810 REM via AMET KL10 timing for R SPINDL
C00201 00104 ∂03-JUL-76 1825 JMC
C00202 00105 ∂04-JUL-76 1305 SYS Your message file exceeds 10K words. Unless you delete some messages,
C00203 00106 ∂04-JUL-76 1339 REG
C00204 00107 ∂04-JUL-76 1506 REG
C00205 ENDMK
C⊗;
∂02-JUN-76 2250 TOB via AI
THANKS FOR THE INFORMATION. I WON'T HIRE HIM.
∂04-JUN-76 1444 FTP:PRATT at MIT-AI
Date: 4 JUN 1976 1740-EST
From: PRATT at MIT-AI
To: JMC at SU-AI
Well, I was wrong, it WAS Montague that Kripke referred to. Let me
quote from Kripke's paper, which appeared in Acta Philosophica
Fennica, 1963.
"It has been argued that such "provability" interpretations of a
modal operator are dispensable in favor of a provability
predicate, attaching to the Goedel number of A; but Professor
Montague's contribution to the present volume casts at least
some doubt on this viewpoint."
Iassume that "present volume" refers to APF, though it could conceivably
refer to JSL 1959 where an earlier version of Kripke semantics appeared;
the reference may have been inadvertently lifted.
-------
Thanks. I know the Montague papers referred to and have ideas on
how to restrict the knowledge axioms to avoid the paradoxes. The
papers - with and without David Kaplan - are reprinted in Montague's
collected papers. If you are interested CONCEP[S76,JMC] contains
a version of my ideas, but the section on applications to AI should
be ignored in its present form; it's a false start.
∂04-JUN-76 2228 PLW at TTY75 2228
so, what do you want me to work on?
I have a project that involves developing a way of describing
(telephone directory e.g.) files at different sites on the ARPAnet
so that a program armed with a description of the file can
get information from it. It will require some explanation.
When can you come in?
∂05-JUN-76 0022 PLW project description meeting
i am done with finals as of next Wednesday. how about meeting to describe the
project then?
ok
∂06-JUN-76 1248 REG
David Roode asks that several accounts be setup to facilitate communication
among the LOTS movers and shakers. He wants to use MAIL to keep everyone
informed of what's happening. His list of people and proposed programmer
names follows. Should I ask LES to set this up?
RDR David Roode
SAU Steve Uhlir
EHF Ed Frank
DSL Drew Lanza
GFF Gene Franklin
DL Dave Lennert
RDN Richard Nielson
MXB Maurice Bizzarri
Sure, let's do it, but how does he think Franklin will receive his mail?
∂08-JUN-76 1244 100 : REM via AMET Chope et al
To: RWG, JMC
Yesterday they gave me a pass and bus ticket to get my car and glasses.
I went from San Mateo by bus to S.F.,
by streetbus and ferry to Sausalito,
by taxi to D+J towing to get my car. (all semi-blind)
My glasses, which the police had knocked off me a week ago weren't there like they had told me.
Now I'm out, as you may have guessed by now.
∂08-JUN-76 1312 FTP:Robert Elton Maas (REM @ MIT-MC)
Date: 8 JUN 1976 1610-EST
From: Robert Elton Maas (REM @ MIT-MC)
To: jmc at SU-AI
<FF> in text file being spindled or crunch-and-spindled is removed, then
size is rounded up to next word boundary, and two words are added
to the directory, thus the overhead is between 9 and 13 characters.
Deleting page marks will remove this tiny overhead at such places.
-------
∂08-JUN-76 1315 FTP:Robert Elton Maas (REM @ MIT-MC)
Date: 8 JUN 1976 1612-EST
From: Robert Elton Maas (REM @ MIT-MC)
To: JMC at SU-AI
Directory page is deleted upon spindling. There is a planned feature
to allow a portion of the directory to be re-created by SPINDL if you
specify page 0 when using the Type command (typeing a portion of
a file in the spindle so you can see what it looks like, for example
if the PRUNE feature isn't implemented, or your PRUNE line is too vague.).
-------
∂08-JUN-76 2330 REM via AMET System busyness, etc.
I just now measured system busyness, by DO TIM1[1,REM], which does a
DI [1,3]/Q/G, and it took 3 minutes. On the KA/10 it only took 1 minute
when nobody else was trying to run, and I generally accept 1-2 minutes
as ok but 3 minutes as alittle too busy for getting much work done,
so I conclude I was right about the disk being a major bottleneck.
(The speed of the KL/10 is not enough to compensate for system overload
and broken swapping disk. Disk-bound jobs like TIM1.DO are especially
slowed by the lack of a swapping channel or disk, etc....) Paging
might improve the situation by reducing the thruput on the disk channel,
or swapping disk or lsi-device would releave the disk bottleneck, I predict.
∧
∂08-JUN-76 2337 REM via AMET
P.S. 3 minutes was about the median for the KA/10 during the 2300-2400 time-of-day, so the KL/10 isn't one iota faster for DI[1,3]/Q/G
∂10-JUN-76 0619 FTP:J. Baratz (BARATZ @ MIT-AI)
Date: 10 JUN 1976 0917-EST
From: J. Baratz (BARATZ @ MIT-AI)
To: MCCARTHY at SU-AI
Dear Prof. McCarthy,
I am currently a graduate student at MIT and
am finishing my masters. I am interested in transferring to
Stanford to pursue a Ph.D. As I am going to be on the
West coast in a week or so, I would like
to use that opportunity to speak to some people there.
Specifically, I will be able to visit Stanford on the 17 (Thurs)
or 18 of June and would like to speak to you and/or some
other member(s) of the department whose research interests are
inAI. My current research interests
are learning and the representation of knowledge.
I can be reached at "Baratz@MIT-AI" and am grateful
for anything you could arrange.
Sincerely yours,
James M. Baratz
-------
You will be welcome, but probably I will be able to see you on the 17th only.
I presume you are proposing to come to Stanford in September 1977. You
should apply in Fall 1976. Anything else would be difficult to arrange.
∂11-JUN-76 0426 TED TWO ANSWERS
I KNOW.
PROBABLY, IF YOU CAN FIND ONE.
∂11-JUN-76 0800 FTP:J. Baratz (BARATZ @ MIT-AI)
Date: 11 JUN 1976 1057-EST
From: J. Baratz (BARATZ @ MIT-AI)
To: JMC at SU-AI
The 17th will be ok. How about 1 P.M.? Also,
can you give me a rough idea of where your office is
and how to get there.
-------
1pm is ok tentatively, and I'll let you know otherwise.
The AI Lab is 5 miles from campus, at 1600 Arastradero Rd.,
Palo Alto. It is on Arastradero between Page Mill Road
and Alpine Road and is .8 mile from Page Mill Road and
is marked Stanford University Donald C. Power Laboratory.
If you have a car, you go West on Page Mill till you pass
Interstate 280 and take the next right onto Arastradero.
If not you scrounge a ride or take a taxi. Strong men
bicycle.
∂11-JUN-76 2144 RDR via AMET pine hall space
i talked to Dickens today and we need to get moving on getting our
space in Pine Hall because there will be remodeling to be done...
Franklin and I must present a budget to Massy on Monday, and then
we must present an order to D.E.C. The current planned configuration
is config[s76,jmc]. We removed the second RPO6 in order to bring
the price below $400,000. I suppose that space will be the next
major issue, and Ed Shaw and I have agreed to meet, and I would like
you to keep planning until then. Shaw is the space tsar.
Please telephone me or come by at your convenience.
∂12-JUN-76 1358 REM via AMET TROUBLE IN MAY76.IN[LET,JMC]
I'm puzzled. On page 164 there is a note from JOSEPH at MIT-MC.
This message contains bare linefeeds even though it is included in
the directory of the E-format file. How did it get included in the
directory without E reformatting it to SU-AI standards???????????
I suppose you must ask Martin Frost about this. Evidently, the mail
receiving program creates directory entries for the pages of the
files without modifying the files themselves.
∂12-JUN-76 1430 REM via AMET No trouble with MAY76.IN[LET,JMC] beyond reformatting bare linefeeds.
I spindled MAY76.IN[LET,JMC] to FOO.SPI[1,REM], telling it R when
it asked about reformatting or omitting, then unspindled to MAY76.IN[1,REM].
SRCCOM shows no changes other than the directory deleted (a feature), i.e.
SRCCOM can't tell the difference between a bare linefeed and a crlf pair.
I therefore don't understand or believe your bug report.
Well, perhaps I thought SPINDL behaved differently. Namely I tried to
TY the spindled file to check whether the messages were readable. The
Each line seemed to take two on the screen and their seemed to be some
garbage at the beginning. Should a spindled file be readable?
∂12-JUN-76 2203 FTP:LEDERBERG at SUMEX-AIM Felicitations
Date: 12 JUN 1976 1825-PDT
From: LEDERBERG at SUMEX-AIM
Subject: Felicitations
To: jmc at SU-AI
and FELICITATIONS again.
But I will be abroad that week; hope to see you
soon after July. All the best
Josh
-------
∂12-JUN-76 2333 100 : REM via AMET IF YOU "TY FOO.SPI"
The spindle directory is in binary, see the format in SPINDL.WRU or
whatever it is. Of course if you type it in ASCII you get random shit.
The text of the sub-files are ASCII except that a CRLF is denoted by
a LF so you get wierdness if you type it out, namely each CRLF token
prints as a LF but then automatic-crlf-upon-line-overflow generates
another CR and LF at j random point depending on your current linelength.
At the end of each page, there is no pagemark, and there may be up to
four random characters that type out but are not logically part of the
page (the "short byte count" field in the directory deletes zero to four
such characters when unspindling). Also the pages may be in any random
order physically in the spindle depending on bubbling that occurs.
Normally they are in the same order as they are logically except that
the first few pages have been bubbled to the end to make room for expanding
the spindle directory. When unspindling, random access is used
to fetch each page in correct sequence according to pointers from the directory.
∂12-JUN-76 2341 100 : REM via AMET EMPLOYMENT
I might be hired for a short-term project (soon to begin of Aug)
converting a large accounting program from COBOL to INFORM (a data-
management language of Standard Information Systems, Inc. which
just recently opened a branch office 1 mile from where I live).
If I accept it, I will be only able to do minimal work maintaining
SPINDL, POX, et al during that time period. Their INFORM language
is pretty nice, they even have string variables that can be used
as parameterless macros to patch things into command strings, including
asking the user for an input string at run-time and patching that
string into a statement to be executed. Mostly the language works with
a database that is an array of fields with columns having fieldnames
and rows having recordnumbers, but ordinary processing a la BASIC
can be done, and upto 10 levels of required source files with RETURN
to pop back one level. Of course, everything is interpreted, except
the interpreter itself and the built-in routines such as SORT and
MERGE and PROFILE etc. (PROFILE counts the occurrances of each
value of some field, sorts by frequency-of-occurrance, and prints a table
of most-common-value first down to those occurring once each last).
Good luck with it.
∂13-JUN-76 0314 REG
Comments concerning BUDGET[S76,JMC]. REG 6/13/76
The staff benefit rate is, I believe, 19%. You'd better check that.
Someone has to provide office furniture and equipment (e.g.,
typewriter, desks, chairs, filing cabinets, phones, copying, etc.)
I don't know how Mr. Augsburger does arithmetic, but the way that I
do it says that a $50,000 payment on $400,000 at 6% reduces the
principal to $181,000 (not 127,000) at the end of 8 years.
I think it rather silly to amortize over 8 years. Recall that in
1970 we concluded that we had run out of compute power on the (then 2
year old) KA10 and so Foonly was started. If we hadn't started
Foonly we probably would have upgraded to a KI in 1971 or 1972.
After the FOONLY debacle, if DEC hadn't promised us a KL10, probably
we'd have gotten a KI in 73.
LOTS is going to run out of computes (and disk space, and terminal
ports) too; if you amortize too slowly you can't afford to replace or
upgrade the equipment. It's fine to say that the 67 had a useful
life of 8 years and had a significant residual value at the end of
that time. This ignores the fact that during the lifetime of the 67
Stanford added a large variety of other computers because the 67 was
fully loaded. There's no budget provision for additional equipment
and upgrades for LOTS; the analogy with the 67 breaks down since
additional computation cannot be provided within the LOTS budget.
I foresee LOTS having to fight tooth and nail through several
committees and the provost to get additional funds. The small staff
of LOTS can be easily ground up by the bureaucratic millstone
(especially if they concentrate on running LOTS instead of on
university politics). There had better be a budget provision for
capital improvements. And it ought to be substantial (i.e., $50,000
doesn't buy very much any more).
I suppose there's something to be said for getting our foot in the
door. This budget looks plausible, but I don't relish the prospect
of telling the provost that LOTS has run out of money and "Please,
sir, more porridge."
[400,000 at 6%, 100% payout in 5 years is $92,000/year,
100% payout in 6 years is $80,000/year]
∂13-JUN-76 0816 RDR via AMET
To: LOTS.DIS[P,DOC]:;
To facilitate communication among those interested in Stanford's new
DECsystem-20 to be run to provide a Low Overhead Timesharing System
(LOTS) for general academic use, for class work as well as
some unsponsored research; a distribution list has been
set up at SAIL. This message is being directed to those on that list.
Please advise me of any inaccuracies in the list, or of any additions
to it.
It is important that everyone be informed, so please use the full
distribution if you know anything others do not. This is done
via the local command:
.mail @lots
One of the destinations of this mailing list is the file
LOTS.LOG[LOT,RDR]. Thus a permanent record of all messages will
be kept.
∂13-JUN-76 1215 EHF Plans
To: LOTS.DIS[P,DOC]:;
For those of you who haven't (or don't know about it (except sau who does))
left plans using the PLAN command. you should. just type PLAN.
GOOD things to leave in PLAN
are your phone number an maybe an address where you will be.
ehf
∂13-JUN-76 1312 REG
My comment about $50,000/year being insufficient to amoritize the LOTS
system is wrong. $50,000 is very nearly the right amount.
∂13-JUN-76 1435 REG
You can't pay secretaries as little as 7000. Try 8300 to 11500.
∂14-JUN-76 0227 RDR via AMET LOTS configuration
To: LOTS.DIS[P,DOC]:;
A proposed system is in CONFIG[S76,JMC].
∂14-JUN-76 0315 JMC BUDGET
To: LOTS.DIS[P,DOC]:;
A tentative budget for LOTS is in BUDGET[S76,JMC]. It will probably be
modified today.
∂14-JUN-76 1124 REG
2040 prices are FOB Marlborough. You get to pay an additional
$1800 to $2600 (my estimate) for shipping.
∂14-JUN-76 2209 JMC Present state of LOTS.
To: LOTS.DIS[P,DOC]:;
1. The Trustees have approved LOTS.
2. J. McCarthy has been appointed Director. There will be an Advisory
Board.
3. The Director reports to the Provost; SCIP is not involved for now.
4. A formal search will be made for a manager.
5. The machine with RPO6 can be had in October or November, but
D.E.C. will try to get it for us in August.
6. Decwriter II is the current most probable terminal.
7. Further information in CONFIG[S76,JMC] and BUDGET[S76,JMC].
∂15-JUN-76 1123 REG
By the way, it's "RP06" not "RPO6"
↓
∂15-JUN-76 1244 REM via AMET Employment
To: JMC, LES
I have been hired at Standard Information Systems, officially
beginning 1100 PDT today. I will continue to check my mail and reply
to bug reports as time permits, but will minimize the time I spend here
until the project at S.I.S. is over appx. Aug 01. They don't have a
good mail system, and don't have a TTYMES capability at all, so to contact
me just leave a note at SU-AI and MIT-MC and wait until I read it, unless
you can reach me by telephone if my line isn't busy.
If there are any finishing touches on the SPINDL etc. that ought
to be thought about, other than the article for publishing, MAIL me a
note and I might look at or think about them as time permits.
At S.I.S. I will be working mostly at home into their dialup
line, with daily short meetings in their office at Bayshore and Embarcadero.
Flexible hours, I keep track of my hours, and bonuses for getting things
done faster than deadlines.
∂15-JUN-76 2233 RDR LOTS terminals
To: LOTS.DIS[P,DOC]:;
It should be mentioned that what is planned as the eventual LOTS terminal
[the mainstay variety anyway] is, as a "high quality display system", very
nice. It is also very cost effective. One will be able to switch character
sets under computer control (for example to APL characters). High quality
graphics will also be possible. Perhaps the central controller of the
displays will be the PDP-11 front end that comes with the system. (Hence it
is "free.") Central control of the units means that each station will be
merely a set of two modules--a high resolution video monitor and a keyboard.
Also, memory is not present at each station and so is not wasted when a
station is idle.
The idea of the interim terminals is stopgapness. The first proposal was
model 33 TTY's. The current proposal of DECwriters came about when SCIP
offered them @$52.50/mo. and we found that Suppes would charge $30/mo. for
some beat up old model 33's he had lying around gathering dust and growing old
some beat up old model 33's he had lying around gathering dust and growing old.
SCIP may well be able to offer us CRT terminals for even less if they do buy
the Lear Siegler and/or DataMedia models as planned.
ε
∂15-JUN-76 2307 RDR SCIP's TRAN (communications switch for terminals)
To: LOTS.DIS[P,DOC]:;
According to Gene Franklin, SCIP will give us 5 or 6 TRAN ports for free.
I see several benefits to this:
1) free dial in lines
2) 5 or 6 terminals we don't pay for but can use (in Pine Hall or
elsewhere).
3) If we can get SCIP to put SCIP terminals in dorms, for SCIP users,
these terminals can access LOTS as well as SCIP. We don't pay for
installation or added communications cost of distance.
4) The TRAN acts as a partial concentrator for LOTS ports. That is,
we will have n lines accessible by >>n (different) terminals
(not all at once of course).
Yes, we'll do it.
∂16-JUN-76 1012 FTP:RINDFLEISCH at SUMEX-AIM LOTS AND SPACE
Date: 16 JUN 1976 1011-PDT
From: RINDFLEISCH at SUMEX-AIM
Subject: LOTS AND SPACE
To: JMC at SU-AI
Congratulations on the approval of LOTS. I have little to contribute to
the space issue in Pine hall except to relate an experience I had with Shaw
and SCIP about trying to relocate SUMEX to Pine hall to integrate better
with IMSSS. The space SCIP was willing to allocate (Mike Roberts and Gene
Franklin were spokesmen) was about 1200 sq ft. For the early LOTS
configuration that is probably adequate - I wasn't trying to get any office
space over there though either. Some of the arguements focussed on the
relative budgets of SCIP and SUMEX (3-5:1 difference) to weight decisions
in favor of SCIP. SCIP has some extensive plans for reoccupying Pine hall
in the future and since LOTS is not part of SCIP (now) that may make them
less than generous or cooperative regarding space. I think you should
negotiate a very careful agreement about space with the right recourse route
(I guess the Provost) for (anticipated) boundary disputes.
Tom R.
-------
I will try to follow your advice.
∂16-JUN-76 1014 FTP:FEIGENBAUM at SUMEX-AIM ai qual grading
Date: 16 JUN 1976 1013-PDT
From: FEIGENBAUM at SUMEX-AIM
Subject: ai qual grading
To: jmc at SU-AI, ccg at SU-AI, winograd at PARC
cc: feigenbaum
apparently, all of the people who wanted to take the ai qual this spring
(eight people) opted for the option of writing eight articles for the
emerging handbook of artificial intelligence--an option that was
agreed uponby all of us during the spring. None of these opted to
go back to the traditional orals format.
Now the articles are cascading in. Sine this was my idea and my "baby",
I will take the primary responsbility for grading them. Most of thhe
articles I have already seen look very good, so we will have little
trouble with this exam. I am hereby inviting the rest of you to
participate in the evaluative procedure to the fullest extent that
your time and motivation allows. Please let me know the extent
of your desired involvement and your time schedule for getting involved.
CCG, you told me earlier that you wanted to be involved in evaluating.
Is that still true?
Best regards,
Ed
-------
∂16-JUN-76 1051 JMC file directory
To: LOTS.DIS[P,DOC]:;
The BUDGET and CONFIG and SPACE files are now in a directory called LOT,JMC.
LOTS group members are welcome to look at it, but comments on unannounced
drafts should be made to me, because they are not ready for comment until
they are finished.
∂17-JUN-76 1136 RDR more LOTS accounts
Talking to Franklin, I came to the realization that it would be a fXXX good idea
(excuse XXX this terminal has no rubout) to give SCIP's Chuck R. Dickens
and Ed S. Williams accounts to read LOYXXX LOTS mail. It would be the
friendly thing to do since they would have ben running it until recently. They
won't use much and besides they might learn something--about mail anyway which
their systems could use!!!
Also Franklin suggests Mech. E. ProgXXX Prof. Joel H. Ferziger have an coXXXaccount
There is no problem provided they can log in conveniently enough so that
communicating with them in this way is not an inconvenience to them. If
so please arrange the accounts with LES. By the way we will have to see
about message facilities in LOTS itself.
We now have a delivery date from D.E.C. It is 19 November and is for
the whole configuration described in CONFIG[LOT,JMC] - not changed recently.
We still don't have good answers on getting manuals at reasonable prices.
∂17-JUN-76 1143 RDR continued
(These terminals I have to use down here make real problems doing thinmgs all aty once
instead of in peics) Ferziger will maybe be advisory board chairman.
Anyway, we realize giving out all these accounts may be a problem, but aren't sure
you don't agree λ on the usefulness.
There are also 2 students I foolishly left off of the list and this matters because
they are social scientist types!!! They are Linda C. Fields and Gloria J. Guth...
∂17-JUN-76 1544 EHF LOTS in general
To: LOTS.DIS[P,DOC]:;
Here are som of my ideas about LOTS:
TERMINALS:
It seems reasonable and pracitical to build SLAC Video
Graphics Terminals. In the long run I think these will prove to be the
best and most reliable terminal we could have. I think that these
will provide much better facilities than any sort of centralized
terminal system (using a pdp11, for example as a terminal processor which
has all the memory for the displays).
CABLING FOR TERMINALS:
It is trivial to run twisted pairs over long distances (i.e several
of line drivers and receivers.
By doing so we would be able to run terminals at any speed
we desired (e.g. 9600 baud) without buying
special modems. Moreover, we could
run multi conductor cables (i.e. 100 twisted pair)
around campus and save large amounts of money.
Yale university has done just this, and has meet with
great success.
SYSTEM CONFIGURATION AND DELIVERY DATE.
I think it is a great mistake to get a computer with only one disk drive.
Having worked on a a10 which had only one disk, I know that
response time will be very poor. Therefore it seems reasonable to
me to get an RP04 disk now, get delivery of the machine
in AUGUST, and get and RP06 disk when thy
become available in NOVEMBER. This will allow us to get the system up
and running much sooner, and will give us
a much better system. It seems to me that the
additional cost of getting the RP04 will be well worth it.
MORE ON TERMINALS:
It seems a real waste to spen $20k on renting terminals.
This money would be much better spent on buying cheap glass
teletypes, or on building our own terminals (it would probably take
about two months to get the SLAC VGT ready for building a production
prototype, if we were to start working on it now).
LASTLY:
LOTS was originally set up as a system in which students would be
an important part of the decission making process. At present it appears
that many important decissions are being made without us. I suggest
that a LOTS meeting be held within the next week or so, in order
that the rational behind many of the current system plans can be
discussed.
Ed Frank
Student Rep., Committee on Computatinational Facilities
SLAC x2467
p.s. keep those cards and letters comming
∂17-JUN-76 1611 EHF correction to previous message
To: LOTS.DIS[P,DOC]:;
part about line drivers an recievers should have read:
It is trivial to run twisted pairs ove long distances (i.e. several
miles) by using about $2 worth of line drivers an receivers
with each terminal.
ehf
∂17-JUN-76 1658 MFB paragraph
see ARPA[NEW,MFB] for a short, general description of my work. If you want more,
leave a msg or call me (996-2569). martin.
∂17-JUN-76 2029 FTP:LEVY at PARC-MAXC ARPA report
Date: 17 JUN 1976 1918-PDT
From: LEVY at PARC-MAXC
Subject: ARPA report
To: jmc at SU-AI
cc: winograd, winograd at SU-AI, levy
John: Following is the paragraph you requested. Don't hestitate to call if you
have any questions. I may be at Xerox (494-4423) or at home (327-2981).
- David Levy
The research of the language understander group, directed by
Terry Winograd and supported by several sources, is guided
by the belief that more powerful and flexible representational
tools are needed to construct programs which reason and use
natural language. Over the last year, in conjunction with
several other researchers, we
have been developing a Knowledge Representation Language.
This language is intended to facilitate the integration of
procedural and declarative knowledge, the representation and use
of partial knowledge, and the ability to reason by analogy.
In the initial implementation of this language, we have thus far
written and debugged two extensive test programs. The first of
these is a language understanding system which accepts several
simple stories in English, then summarizes and answers questions
about them. It demonstrates the capability in the language for
representing large, related segments of real-world knowledge,
and for integrating syntactic and semantic knowledge. The
second program, a medical diagnosis system which accepts a
set of patient symtoms and responds with a set of explanatory
diseases, exercises the facilities provided in the language for
factoring knowledge along many dimensions, and reasoning from
partial information.
-------
∂17-JUN-76 1643 CCG Accomplishments for ARPA
Feigenbaum may also report Lenat's work on his learning program,
called AM. I presume you'll coordinate that.
Cordell
↓
∂17-JUN-76 2032 FTP:LEVY at PARC-MAXC ARPA report
Date: 17 JUN 1976 2033-PDT
From: LEVY at PARC-MAXC
Subject: ARPA report
To: jmc at SU-AI
cc: winograd at SU-AI, levy at SU-AI
John: I'm not sure that my earlier message made it across the net. If it did,
this note is redundant.
Below is the research statement for Terry's group. Feel free to call me if you
want to discuss it. I'll either be at Xerox (494-4423) or home (327-2981).
-- David Levy
The research of the language understander group, directed by
Terry Winograd and supported by several sources, is guided
by the belief that more powerful and flexible representational
tools are needed to construct programs which reason and use
natural language. Over the last year, in conjunction with
several other researchers, we
have been developing a Knowledge Representation Language.
This language is intended to facilitate the integration of
procedural and declarative knowledge, the representation and use
of partial knowledge, and the ability to reason by analogy.
In the initial implementation of this language, we have thus far
written and debugged two extensive test programs. The first of
these is a language understanding system which accepts several
simple stories in English, then summarizes and answers questions
about them. It demonstrates the capability in the language for
representing large, related segments of real-world knowledge,
and for integrating syntactic and semantic knowledge. The
second program, a medical diagnosis system which accepts a
set of patient symtoms and responds with a set of explanatory
diseases, exercises the facilities provided in the language for
factoring knowledge along many dimensions, and reasoning from
partial information.
-------
∂17-JUN-76 2146 GF via AMET TO RDR
To: LOTS.DIS[P,DOC]:;
DAVE, I WILL ASK FOR A BUILDING AND ROOM KEY SO THAT YOU CON USE THE
ISL TERMINAL AFTER HOURS. I GOT THE COPIES OF TTHE OTHER MEMOS. THANKS.
I AGREE WITH ED FRANK THAT WE SHOULD MEET SOON TO TALK ABOUT PLANS.
ESPECIALLY IMPORTANT IS THE MATTER OF TERMINAL SELECTION AND
ASSIGNMENT OF RESPONSIBILITY FOR HELPING SPECIFIC CLASSES
TO CONVERT THEIR COMPUTING TO THE 2040. I'LL BE OUT MOST OF NEXT WEEK
BUT WOULD SUGGEST TUESDAY JUNE22 AT 4 PM. ISL CONFERENCE ROOM.
SHALL I MAKE IT DEFINITE?
π
∂17-JUN-76 2334 DBL AM paragraph
The one-paragraph description of AM, which Cordell askd me to write
for you to give to ARPA, is found on file AMP[NOT,DBL].
Please forgive its boasting tone; I was instructed to make it qite
positive-sounding.
Regards,
Doug Lenat
∂17-JUN-76 2339 DCL carlstrom accomplishments
ACCOMP[VCG,DCL] contains 4 paragraphs with references. Let me know if
they are suitable-i think they may be toolong.-David
¬
∂17-JUN-76 2327 RDR Space and RP06's
To: LOTS.DIS[LOT,RDR]:;
I talked today with Amy Blue about LOTS space needs in Pine Hall or elsewhere.
Pine seems to be our only alternative since it exists as a computer ready location
as far as floor and air goes. She seemed to understand LOTS needs and, except that
I may be overoptimistic, seemed also to think that thew LOTS plan was the most
reasonable of the three presented so far [SCIUP's and IMSSS's too].
It asks for:
a) 15' x 24' machine room carved out of the parkinglot/med center side of
the old /67 machine rroom,
b) 15' x 24' hardcopy terminal room taking up the rest of the space along that
wall of old /67 room
c( Pine Hall 186 as a headquarters for LOTS student staff
d) Current SCIP engineering managers and 2 assistants offices for LOTS
secretarty, director, amanager, and systems programmer.
e) current SCIP engineering lab as a CRT terminal room
It achieves these goals by slight reductions in the amount of space offred
SCIP and IMSSS, better arrangement of SCIP's area to avoid wasted lobby areas,
and relocation of engineering to another building.
It also is important that SCIP maintain desirable space for their users rather
than converting it into offices because their users will "invade" LOTS space
otherwise.
There will be a meeting next Tuesday from 2pm until 3:300 about this matter in
building 10 of the Quad. I think a large turnout for LOTS by students would help,
as SCIP and SUPPES will bring their armies for sure...
The delay in DEC-20 delivery stems largely from the request for 256K of core as
opposed to only 128K, rather than from the RP04 vs. RP06 question!!! Without
cache, we need that core more than we need disk space, but that isn't even a
choice that has to be made.
We're working on getting time on a 20 elsewhere, over the NET.
∂17-JUN-76 2351 RDR
should we put any other C.S. profs. on the LOTS distribution list?
I think I would like to confine the distribution list to people who
are likely to do some work. Perhaps we should have a larger distribution
list for potential LOTS users who are not involved in making LOTS work.
I don't feel strongly about the matter, and feigenbaum%sumex is the
obvious candidate. I have put my secretary Patte Wood on the list,
and she should be informed of any members of the list who should get
their notices by campus mail.
∂18-JUN-76 0247 RWW ARPA PARAGRAPHS
To: JMC, LES
My paragraphs are on CARLS.S76. I will be at home all morning if they
need repairing.
rww
∂18-JUN-76 0823 100 : REM via AMET S.I.S. JOB
To: JMC, RWG
The job at Standard Information Systems is probably not
for me. The INFORM language, despite its nicenesses, has one absurdity,
complete lack of arrays. Even FAIL and other assemblers can simulate
arrays by means of indexing or patching together an indirect-address.
The hassles involved in attempting to get around this limitation are
probably too much to handle. I'll speak to my therepist today and
might quit the job before a week has elapsed. a
No definite decision so far this morning, but I've been sleeping a lot
rather than get up to face the job, so it's ominous.
I think you are not so easily defeated by difficulties as you fear.
You have probably overcome more difficult problems in POX and SPINDL
than INFORM's lack of arrays.
∂18-JUN-76 1038 TAG
THERE IS A STRONG POSSIBILITY THAT A CHEAPER ALTERNATIVE SOURCE
OF DISK FOR DEC USERS WHOSE MACHINES POSSESS MASSBUSES (I.E., PDP11/70, KL20)
WILL EXIST IN 6 MONTHS. IF YOU DON'T OBJECT TO GOING MULTI-VENDOR FOR EXPANSIONS
TO THE LOTS SYSTEM, YOU MAY WISH TO EXAMINE SUCH AN ALTERNATIVE TO DEC'S RP06.
According to Tom Gafford, who consults for a relevant company, there will
be mass-buss adapaters available in six months or so that will permit
attachment of other vendors disks to 2040s and 11/70s. This will permit
LOTS to expand its disk file at something closer to the $12,000 per
unit that the AI Lab pays rather than the $34,900 that D.E.C. charges.
This doesn't suggest any changes in the present order to D.E.C., but
it makes it look less like a box and confirms the choice of the
200 megabyte disk, because the 100 megabyte disks will probably not
be available from the independent manufacturers (or maybe even from D.E.C.)
much longer.
Feigenbaum does not want to be on the present LOTS distribution list -
but rather one for major announcements to potential users. He already
gets enought net mail. Others may feel the same way.
∂18-JUN-76 1230 100 : rem via AMET
On the contrary, in POX and SPINDL I was using FAIL which does have access to array indexing. Also I've avoided doing "unreasonably-difficult" things in POX.
∂18-JUN-76 1233 100 : REM via AMET System still losing
I just did a W R (who of running jobs) and found among the 10 normal
jobs 8 RUNQ* and one each DIOW and TQ -- thus due to only 256k memory
effectively online, the jobs in core run to completion (speed 5 on kl10)
leaving the cpu waiting for disk time to swap in a runnable job.
Who shows only 300p (150k) usable, so I'll withhold final verdict
until the rest of memory is connected. Perhaps with 512k of memory connected
all runnable jobs will fit in core and we won't need the librascope, we'll see...
↓
∂18-JUN-76 1354 REG Letter to John Leng
If you haven't sent it yet, Stafford asks that you send your letter to
him for forwarding to Leng. Stafford wants to include a memo of his own
with your letter (and to keep it in channels, etc.)
∂18-JUN-76 1403 EHF
To: LOTS.DIS[P,DOC]:;
If what Dave Roode says is true, naely that what is
holding up delivery is not the rp06 disk but rather the extra 128k
of memory then how about the following:
get the 20 with 128k and an rp06 disk, with delivery in AUGUST.
then when the extra 128k becomes available, get it.
This assumes that the extra 128k can be installed as a field
mod. I have no idea whether or not this is true.
My motive for pushing for delivery in AUGUST is that this will give
us three extra months with the machine. In addition two of these
months will be during the summer, when people have
much more time to spend working on the machine.
ehf
∂18-JUN-76 1620 REG
A minimum of 20' by 25' is needed for the LOTS computer room, assuming
that some expansion of disk and tape units is inevitable. Additional
space would be needed (on the 25' dimension) for additional equipment.
∂18-JUN-76 2348 JMC Reasons for accepting delay
To: LOTS.DIS[P,DOC]:;
The delay in delivery from August to November is indeed painful.
Stafford claims that going from 128K to 256K requires a new back
plane that won't be ready until then, and frankly this claim is
puzzling. RPO6s are also only available in November. We were
offered delivery with 128K and an RPO4 in August with the possibility
of trade-in in November with an $8000 loss. This might be acceptable
if the money existed, but we are constrained by the $400,000 limit on
purchase order. In my opinion, RPO6s are an absolute necessity; the
AI Lab now has 800 megabytes in its file system, and I think LOTS
will eventually need even more. The 100 megabyte 3330 type disk is
now a technological orphan, whereas 200 megabyte disks will soon be
available on the 2040 at much lower cost than D.E.C. now charges.
One can speculate that D.E.C. could do better on delivery if they
were so motivated, but I don't know any way to motivate them right
now.
∂19-JUN-76 1045 FTP:Marvin Minsky (MINSKY @ MIT-AI)
Date: 19 JUN 1976 1343-EST
From: Marvin Minsky (MINSKY @ MIT-AI)
To: jmc at SU-AI
Congratulations & Best Wishes
-------
∂19-JUN-76 1327 SGK
Well, I must admit, I am not doing a hell of a lot that should, but then again
I use virtually no computer time, and my file space has been dropping. About
the only constructive thing I do from the AI lab's and CS's standpoint is
demonstrate to as many people as I can that computers can and are being used
to help people (like myself on this system), and that they are not to be feared,
blown up, or lobbyed against. You are as aware as anyone of all the anti technology
propoganda floating around, hardly ever counterbalanced by any interests the
people don't feel are part of big business. Most of my friends have seen the
AI lab, and seen how I use the machine for my correspondence, storage, occasional
computing, and typesetting. They have a much more realistic view of 'big bad'
computers, and in turn pass this on to people they know.
All of that is pretty important, but even I don't fully regard that as justificaion
for free computer time. I really don't know what to say other than I use
the machine infrequently, and at non peak times. Occasionally I fix
some bugs in programs I've left behind, and I keep up with what is going on.
It is a rare visit that I don't get some sort of plea for assistance from
some user. Oh well, please don't flush me. SGK
¬
For now DATADISC is the short resource. Please don't use it when
your using it makes 31 users.
∂19-JUN-76 1544 RDR
TOPS-20 does have a mail facility that is O>K.
∂19-JUN-76 1546 RDR Pine Hall
To: LOTS.DIS[LOT,RDR]:;
Spring quarter I took pictures of how crowded Pine Hall gets when in full
swing. We can use these to argue against SCIP's moving half their offices
into the users area in Pine. I'll bring them to the spcae meeting or ask
me if you want to see them...
ε
∂19-JUN-76 1550 RDR machine room /LOTS
To: REG, REG, JMC
We might be able to leave one wall off of the room so thatr we
merge into and could slide over into SCIP's RJE area. I'm not sure if
this would solve the problem or hinder it, but there are at least some
advantages toi this approach. SCIP already was planning on some sort of
public access to their machine room. I guess we need to talk about this before
the meeting.
α
∂18-JUN-76 1111 JMC
I think you are not so easily defeated by difficulties as you fear.
You have probably overcome more difficult problems in POX and SPINDL
than INFORM's lack of arrays.
[REM - Friday I griped and suggested I might not be right for the job.
My supervisor says arrays might be simulatable in a database, and the
disk activity would be transparent to the user. We don't yet know
if it can be done, but are working on it. - To me it sounds absurd
for any language to completely not-implement in-core arrays.
Only some versions of BASIC are that bad, and INFORM.]
∂21-JUN-76 1140 REG
I have your blueprint of Pine hall in my office
∂21-JUN-76 1832 FTP:FEIGENBAUM at SUMEX-AIM fellowship award for randy davis
Date: 21 JUN 1976 1435-PDT
From: FEIGENBAUM at SUMEX-AIM
Subject: fellowship award for randy davis
To: lederberg, amarel, baker, rindfleisch, knuth, feldman,
To: russell at ISI, carlstrom at ISI, jmc at SU-AI, phw at MIT-AI,
To: nilsson at SRI-AI, hart at SRI-AI, rha at RAND-ISD
cc: davis, feigenbaum
o
To: friends of Randy Davis
Randy Davis has just been awarded a prestigious new post-doctoral
science fellowship award--the chiam weizman fellowship. The
fellowships are awarded in national competition administered
by cal tech.
I thought you would be interested in the good news.
Ed Feigenbaum
-------
ε
∂21-JUN-76 2216 REM via AMET
To: JMC, RWG
I quit that job at S.I.S. earlier today, it was upsetting me too much, back to being available here or ITS or TENEX etc.
∂22-JUN-76 1301 JMC*
Call Uncapher about KL20 and data compression.
∂22-JUN-76 2047 ELM LIBRASCOPE
IT HAS BEEN DETERMINED THAT 3.6 MEGAWORDS OF IT
RETAIN THE WORST CASE DATA PATTERN FOR 48 HOURS.
TESTING CONTINUES INTERMITTENTLY.
Well, that's enough to be useful.
∂23-JUN-76 0051 RDR last message
I just said roughly--
I have been thinking about the space options(I was not too happy with
the meeting today, although I get the impression Provost wants to be
nice to LOTS, and that is good.) and the more I think, the more I like
the Cedar Hall option, despite some strong initial reservations. We would
only need to be sure of electricity which is probably there, and could ark
fnr carpets and airconditioning. This, and any wires run, would not be wasted
when we moved to the Quad, as any SCIP functions in the building will want
lines to Pine Hall (the TRAN) too...
`I'll see what Shaw, who is Assistant to Provost, says on Friday.
∂23-JUN-76 0157 RDR Gloria Guth
Although the accounts for Gloria and the others were arranged earlier,
they just materialized today. Anyway, I haven't told the people
about them yet but Lyssa says Gloria is going to ask you about that
among other things tomorrow, and she has no idea the accounts exist.
So there is no need for anything to be done--there has just been a
communication lag. (I'd added the respective mail files to the
distribution list before the accounts actually existed, so the mail
has been coming.)
I will inform the other people of the accounts.
On the question of who should receive campus mail copies, I'll offer
that option to the others too (but for example Franklin likes to log
in and read his mail.) What about sending copies to Massy? He may not
want to log in, but I think he would at least like to get the messages.
If you think it's a good idea, will you tell your secretary because I
wouldn't want to do it without checking with you (chain of command,
going over a head and all that...)
∂23-JUN-76 0821 FTP:LEVY at RUTGERS-10 lot.jmc directory
Date: 23 Jun 1976 (Wednesday) 1121-Est
From: LEVY at RUTGERS-10
Subject: lot.jmc directory
To: JMC at SU-AI, LEVY
John,
I tried to print ou the directory LOT.JMC, but got a message
"illegal ppn. dsk:lot.jmc[100,100]. Can something be done to
make the directory (and files) accessible?
Thanks.
Cordially,
Saul Levy (Levy@Rutgers-10)
The command to print the directory should be di[lot,jmc]. You
apparently used a period instead of a comma. The two parts of
a file name are separated by period, but the two parts of a directory
name are separated by comma. If this doesn't work, ask again.
∂23-JUN-76 1615 RDR
CC: JMC
Here is the response of an MIT undergrad on MIT's ITS system use by undergrad--
it sounds like the answer is that authorized users have a hard time using ITS...
∂23-JUN-76 0723 FTP:Michael L. Kazar (MLK @ MIT-ML) ITS usage.
Date: 23 JUN 1976 1025-EST
From: Michael L. Kazar (MLK @ MIT-ML)
Subject: ITS usage.
To: rdr at SU-AI
I believe that there are a few reasons that the ITS systems
are so lightly loaded. One is that despite the relative "fame" they
have around the network for being a "free" system, the vast majority
of the students here do not know that they exist. This situation is
not aided by the fact that the dialup phone numbers are not
publicized, although not hidden either. In addition, it is frowned
upon to make a dir for yourself, and as far as I know, the method for
doing this is documented relatively obscurely in only one place. The
reason for this is that disk space is fairly tight on most of the
systems. This is part of a larger problem, namely that the
documentation on the various features of DDT and ITS in general is
not well organized, especially as far as the new user is concerned.
Thus it is not easy to find out even the minimum needed to use the
system. Most of the people at the ITS sytems are not nasty, but in
generaly, the use of these systems is restricted to people who work
for MAC, and while this rule is almost never enforced, some people
will not answer questions by random people. It should be noted that
the vast majority of thhe people will help you to some extent, but
learning the systems usually involves searching through large help
files. Many oif the com,mands to ddt are very obscure, such as
<alt><alt><control R> . In generaly, then , the reason that these
systems are so rarely used by the mit community is thatt there is no
disk space available for general use and also that the system is very
hard to learn. One might say that their motto is "security through
obscurity". One last point is that the most common programming
languages of scientific users, FORTRAN PL/1 BASic apl, etc are not
supported. They now are available through the decsystem 10(tops 10)
simulator, but few people use it, and it doesn't work perfectly.Most
people use lisp or assembler. There is a users dir (users;) on some
of the its sytstems, for general file storage, but this is generally
a limited resource for the above reasons.
Also, I think that I might be able to get the new york
university elf numbers, as a friend is comming up for a visit this
weekend who uses it. If he knows them I will mail you a copy for your
list. -------
∂24-JUN-76 1136 FTP:UNCAPHER@ISIB 2040
Date: 24 JUN 1976 1134-PDT
Sender: JEANNETTE at USC-ISIB
Subject: 2040
From: UNCAPHER@ISIB
To: JMC at SU-AI
Cc: UNCAPHER
Message-ID: <[USC-ISIB]24-JUN-76 11:34:15-PDT.JEANNETTE>
As you know, 2040 standard DEC monitor is largely Tenex based. Hence,
we are adding JSYS and application features to the system to achieve
virtually complete compatability.
DEC probably will not have an NCP or hardware interface until December
1976. We will build a kludge to allow the 2040 to be on the Net at
the end of summer. This is essential so that we can get help from
SRI, BBN and anyone else in completing the compatibility package.
If we do not get help, we will go it alone. It will take a bit
longer. Our goal is to produce the changes in such a way that they
become DEC supported software. Of course, all our software is
available.
We expect a 1080T in late CY '76 as a replacement for the 2040 which
is scheduled for NELC in March '77.
Keith
-------
↓
∂24-JUN-76 2133 RDR via AMET Meeting
To: LOTS.DIS[P,DOC]:;
There will be a meeting of all interested in LOTS next Tuesday,
June 29, at 4:00 in the afternoon in the ISL Conference Room
of the Durand Building on campus. It is room 126. We should
meet frequently this summer as LOTS gets going...
β
∂24-JUN-76 2201 RDR via AMET MAINSAIL
To: REG, JMC, MSG.MSG[LOT,RDR]
As I hear it, mainsail already has double precision arithmetic
within its capacities that may be specified at the time the
compiler is generated on a machine. It lacks debugging facilities
but they are being worked on. Someone shoulkd talk to
Clark Wilcox at SUMEX and see about it. He has also deleted a
lot of the extra facilities of SAIL and redone the record
structure to be close to that of PASCAL. I talked to
an ISL grad. student and Maurice Bizzari about this.
What about Pascal?
One thing that was brought up about the debugging facilities
was tat maybe in some senses better than a good stand alone text editor
plus a separate debugger for the languages programs, is an editor/
debugger implicit in the language that can interact based on
knowledge of the language as the program is being enetered (like
BASIC) only better like the UCI LISP editor. If mainsail is cleaner
and desirably less complicated, and efficient, plus it has double
precision arithmetic, maybe what we should do is develop this
kind of break package for mainsail rather than attempting to hack
awy at regular SAIL. If the editor thus developed makes use of
display capabilities, then that would be nearlier ultimate. For
example if you can display not a page but rather an algol procedure via
editor commands, than is that not more natural? Also being able to
zoom in and out on and of the block structure, as display the first
block in the block containing the current block, etc. is natural. Well whatever, but nobody seems to have this kind of a debugger/editor
except for lisp and it seems that what LISP demands is also an
enhancement to a language such as algol. It is what I have had in mind,
unconsciously assumed to be tthe obvious answer, ever since
it looked like they might make an intercative algol w. Yet they
did something much easier and even what SAIL has
is apparently different.
If MAINSAIL is available, we should have it. We should also have whatever
else is available. What languages will be used in courses is up to the
instructors, and what individual users use is up to them. I belabor this
obvious point, because I don't want LOTS as an institution to have a party
line on languages except in so far as decisions about staff effort are
concerned or conceivably concerning languages like INTERLISP that require
a lot of computer time to support a few users.
∂24-JUN-76 2352 REM via AMET
By the way, when I was calling you from the pay phone at Chope
and my 3 minutes were up, how come you didn't call me back to continue our
conversation (about Liz)?
Usually the calling party re-establishes the connection
after a cut-off, and, when you didn't I assumed you were done.
I didn't notice the three minutes.
Had you mentioned it, I would have.
∂24-JUN-76 2357 JMC
Usually the calling party re-establishes the connection
after a cut-off, and, when you didn't I assumed you were done.
I didn't notice the three minutes.
Had you mentioned it, I would have.
[REM - When I called Liz on June 3, and my 3 minutes were up, she
immediately analyzed the situation correctly and called me back.
When a person is in the psychiatric ward of a hospital he has only
the money in his pocket at the time he is taken into custody, not
even the money in his car or purse or everything-box which are
impounded far away (Sausalito in my case), so every nickel counts.
It might mean the difference between making an emergency call and not
being able to call for help if the shrinks mistreat the patient.
At one point, when a random nurse began forcing me to take Thorazine
as punishment for being too friendly to other patients (she later referred
to me as "Hyper", probably hyper-active) I made a call to RWG in case
I was too tranqualized to get out of bed later. Luckily
she didn't continue it after I got so dizzy I went to bed at 2100 right
after the phone call.]
∂25-JUN-76 0640 FTP:Stavros M. Macrakis (MACRAK @ MIT-AI)
Date: 25 JUN 1976 0939-EST
From: Stavros M. Macrakis (MACRAK @ MIT-AI)
To: JMC at MIT-AI, jmc at SU-AI
I took the liberty of changing your "Inquir" entry here to
list your name as "John McCarthy" rather than "y".
I hope this is OK.
-------
∂25-JUN-76 0700 FTP:Marvin Minsky (MINSKY @ MIT-AI)
Date: 25 JUN 1976 0959-EST
From: Marvin Minsky (MINSKY @ MIT-AI)
To: jmc at SU-AI
yes, I think hilts is
fine. I didnt talk to him much
because i didnt think anyone could do what he wanted..
but he seems to have put in just the
thoroughness that i doubted. Now I'd be glad to
talk to him more, and you could tell him that.
-------
∂25-JUN-76 1152 REG
To: RDR
CC: REG, JMC, MSG.MSG[LOT,RDR]
Despite the availability of MAINSAIL, I think we need some work put in to
SAIL. I don't know anyone besides Wilcox who's used MAILSAIL. There's
a distinct possibility that it'll turn out to be unsuitable for our purposes
in some way (or, equivalently, requires too much effort to make it suitable
within the necessary time frame).
As for PASCAL, we've got one here which I suppose can be transplanted to LOTS
should anyone want it.
Finally, your idea about combining the compiler, runtimes, debugger and editor
into one package is commendable but unworkable. You could spend a couple of years
working on that and not achieve a suitable result. Looks to me like a huge
sink for man power.
π
∂25-JUN-76 1340 RSC Cartwright's Ph.D. Orals
My orals will be held in the Serra House Conference Room on Monday on
July 12 at 2:15 P.M. as planned. -- Corky Cartwright
∂25-JUN-76 2112 REM via AMET Librascope
To: LES, JMC, REG, JBR
I hear that the librascope was actually used for swapping 37
jobs during maintenance. I have a proposal for an experiment:
During the busy part of the day, not much work can get done because of
disk channel tied up with slow swapping. Try putting the system up
noon to 4pm swapping 37 job slots on librascope, with 15 minute
autologout (all jobs, not just datadisk). I betcha more work can get
done in that mode than at present. (but I may be wrong, no money bet)
Of course, the 2-level swapping (librascope then 3330) will be better,
but in the meantime...
Very soon all jobs will be swapping on Librascope again. Do you want
to work more on crunching - or are we still paying you?
∂25-JUN-76 2313 REM via AMET JOB ETC.
I have not yet gotten my paycheck for the last half of the fourth
month, i.e. for the first half of May. After that I haven't been hired,
which is fine with the system so busy. At present there is again
a gap from 10pm to morning when system load is light, but in the middle
of the day it is at least as bad as it ever was on the KA10 with
broken swapper. It is ok to do most work at night, but it is bad
to be totally excluded from daytime use by incredible system slowness.
Hence I'll wait until midday system load goes down to something on the
order of 5 minutes to get 1 minute of work done (currently it
is between 7 and 11 minutes in early afternoon). I normally consider
anything above 2 minutes to be too much for serious work, and above
5 minutes impossible.
ε
∂26-JUN-76 1615 RDR CHEAP terminals
To: LOTS.DIS[P,DOC]:;
As some of you know, but not by any means all, it looks like we can
get a group deal for LOTS and interested individuals on kits
for the LSI terminal. This is a upper and lower case 24 lines by 80
characters CRT display which lacks features like cursor control
(strictly a glass teletype, Ed) and it is on display in Pine Hall.
The estimate is 16 hours to assemble and it is a nice efficient simple design. The tentative price is $750 per unit. Anyone interested
in a home terminal? Know anyone else or any living groups, etc.
MIT--Doesn't this beat DEC VT%@'s?)
Well spread the word and what do you think?
∂26-JUN-76 1624 FTP:Mark R. Crispin (MRC @ MIT-AI)
Date: 26 JUN 1976 1922-EST
From: Mark R. Crispin (MRC @ MIT-AI)
To: jmc at SU-AI
A while back, some loser was claiming you said something
to the effect that while Stanford's system runs 40 jobs on the
average, with restricted access, the ITS systems run only 12 or
so. Therefore, you concluded, (or so this loser claimed) that
the ITS systems are simply stingy about system access(he used
another term, but I forget it right now).
I doubt you would make such an absurd statement, since you
are in a position to know the facts. But, I've been suprised
before, so here they are...
ITS never runs as few as 12 jobs. The overhead jobs alone
account for more than 12. The number printed out at login time
is exactly what it says; the number of users who have jobs
attached to their terminals. That figure does not include detached
or disowned jobs.
In addition, there is virtually never only one job per user.
Every user has a HACTRN (alias top-level DDT) which is his/her
command decoder and counts as a job. From DDT, the user makes
inferiors of the DDT, each of which count as a job. It is not
uncommon for a user to run 5 jobs or more under his/her HACTRN.
The average number of inferiors a person has is about 3; therefore
every local user takes up 4 jobs.
To add to that, every user coming in from the network has a
LOGGER controlling a STY (:=PTY) where he runs on. Then, there
are FTP servers and JOBDEV device servers.
In summary, an average load on ITS is 60 jobs, not 12. There
is a limit on the AI machine(as well on ML and DM) of 69 jobs, with
the MC KL-10 having 90. It is quite common on the AI machine during
the day to run out of job slots and have to go and gun down various
random detached trees and/or beg people to flush some of their
unneeded jobs. It is even more common on the MC machine for it to
start thrashing when it starts getting 20 or more MACSYMAs; it
hasn't hit its limit on jobs(yet), but that doesn't mean that
it can't get loaded either.
While it can be accepted that SAIL might have a heavier
load than AI, AI cannot be considered a lightly loaded machine.
Part of the reason AI's response is as good as it is now is
because the monitor is so efficient. But that doesn't mean it
can't start thrashing either.
Probably this letter is unnecessary. Then again, you should
know that some people are claiming you are making totally inane
statements about the comparison between ITS and SAIL. I'll try
and find out who this loser was so you can set him straight. A
person with your position and reputation does not need this sort
of lossage attached to you. Please don't take this in any other
way than the spirit that it was intended. I have tried in vain
to convince network randoms that "20 Lusers" means a horribly
loaded system, since that implies over 60 jobs. I figure maybe
you could convince these people, since they seem to have talked
to you about the subject.
-------
∂26-JUN-76 1641 RDR MIT
To: LOTS.DIS[P,DOC]:;
MIT's SIPB(student information processing board), together
with some random departments, is thinking of getting a DEC-20.
¬
∂26-JUN-76 1707 100 : RDR
To: REG, JMC
I'd like to investigate the possibility of DataMedia giving us an offer
similar to LSI's...
Go ahead, but we may have to decide about LSI by 15 July.
∂27-JUN-76 1746 REG
Look at P[LOT,JMC]
∂28-JUN-76 0610 FTP:Benjamin J. Kuipers (BEN @ MIT-AI) Home terminal suggestion
Date: 28 JUN 1976 0908-EST
From: Benjamin J. Kuipers (BEN @ MIT-AI)
Subject: Home terminal suggestion
To: jmc at SU-AI
28 June 1976
John,
I had an idea for your home terminal discussion that is
somewhat obvious, but ought to be stated explicitly anyway.
There are some uses for computer mail which are very important
around the lab, but which are quite different from personal mail.
The most important of them is BUG MAIL. Our bug-reporting
facility allows you to specify the name of the offending program,
and send a message describing the problem, which is then automatically
sent to the relevant people. You doubtless have a similar thing.
What is really happening is that you can address a communication with
a short description of the content, and the system forwards it to
the correct people.
This is exactly what suggestion-boxes are supposed to do.
Many of the small annoyances in our complicated society exist not
because no one is willing to fix them, but because there is no
mechanism for getting knowledge of the problem to the person who
is willing, even eager, to fix it. Suggestion boxes in companies
are notoriously ineffective. In the lab, on the other hand, a
BUG or FEATURE message gets reasonably prompt attention, or at least
a reply.
I thought of this the other day when I was struck by a minor
suggestion I wanted to give to Cabot, Cabot, and Forbes, the owners
of the AI lab's building. There is no mechanism for minor suggestions.
(In this case, that half of the elevators should home on the ninth floor
instead of the first.)
While we are thinking of minor ways to make life better, the
other kind of mail I get quite a bit of on the system is JUNK MAIL.
I don't mind junk mail as much as many people, but the problem in the
US Post, is that it wastes enormous amounts of paper. I was struck at
one point by the amount of junk mail I have read on our system which has
probably never had physical existence, from creation to extinction.
That seems an improvement: it gets the information to the concerned
parties, but doesn't waste paper.
Anyway, there's my suggestions for what they're worth.
Ben Kuipers
-------
↓
∂28-JUN-76 1849 CCG ARPA VISIT
To: PSI.DIS[PSI,DRB]:;
CC: LES, JMC
Bill Carlson will visit on July 12, and he wants to spend all day
reviewing automatic programming progress. All the PSI group should
plan to be here all that day,have presentations ready, and demos
ready where applicable. Also answer questions about relevance.
Thanks
Cordell
β
∂28-JUN-76 1900 LES Remember to talk to Lieberman
∂29-JUN-76 0055 RDR terminals
To: LOTS.DIS[P,DOC]:;
The Lear Siegler ADM-3 terminal which we have an offer on through SCIP at
$750 in kit form is also available in a model which understands some
cursor controls sent by the computer and has keys on the keyboard for
some of them. Although this is not the kind in the kit deal (which will soon
be available at the BYTE shop for $895), it usually adds only $50 to
the assembled price--so maybe something can be worked out.
Dickens will try to get a kit for inspection. Suggest you ask him to
inquire about cursor model. Ask Ralph whether its characteristics will
be good for us.
∂30-JUN-76 0213 RDR another LOTS person
To: JMC
CC: REG, LES
Could we open an account for John C. Freed ?
∂30-JUN-76 0214 RDR campus dial-in terminal
To: JMC, REG
I think we should quickly get an office for LOTS with a terminal
in it that can access the AI Lab. There are not that many dial in
(out) lines around campus and it is a major problem for some
people to get out to 1600 Arastadero Road, especially at night.
For example, this latest person, John Freed, is a natural to help
out with LOTS yet the physical isolation is or has been keeping him from it.
He's taken at least 6 computer science classes and has spent a lot of
time hacking away on personal projects on the BI$ school BASIC system.
He just needs a slight push in our direction, and that would have happened
long ago if we had better access. [Of course, this will be the case when
the 20 is on campus.]
If the 20 will have a leased line out to AI, why not get it soon, and
install it connected to a terminal in the LOTS office? This could be
a datamedia, Imlac or LSI--whatever. In any event, it is something to
consider.
Thinking about the gearing up of the 20, I see an awful lot of
potential in people working for credit. There are undergrads who have
been searching for a CS 293 or EE 288 or whatever project to do. I just
talked to another one(in addition to Freed) and I know of more. Of course,
not many will be around until autumn--but we might want to create a list
of projects beyond the absolutely necessary ones because there are
people who might not do what is neccessary because it is dull or not
up their alley but who would gladly do something that was also desirable even
if not crucial...
ε
∂01-JUL-76 2126 MM via AI sarah
Don't forget fireworks in shed.
∂01-JUL-76 2249 KRD
No sir! I wouldn't touch it at all if I didn't have to.
OK, I'm just getting edgy about the KL-10 not being the millenium after all.
∂02-JUL-76 0909 REM via AMET Preliminary timing figures for job running on crippled KL/10
Survey so-far on KL/10 with no librascope and only 256k core
includes the hours of 0,1,3,5,6,8,9,11,12,13,14,16,17,20,22,23 PDT.
The task is the disk-bound job "DO TIM1[1,REM]", which used to take
just under 1 minute on KA/10 in middle of night when not swapping.
The pattern so-far is: daytime (06 to 17) 3-11 minutes (busy to impossible),
late-evening (end-of-maintainance to midnight) 1-2 minutes (good),
wee-hours (midnight to 06) 1 minute (good). Thus anytime after maintainance
before the music people show up next morning is a good time to use the computer.
Any other time is a drag. As soon as 512k and/or Librascope come online
I expect this to change for the better.
∂02-JUL-76 1117 FTP:Robert Elton Maas (REM @ MIT-MC)
Date: 2 JUL 1976 1419-EST
From: Robert Elton Maas (REM @ MIT-MC)
To: JMC at SU-AI
Host not responding
Connection suspended
well, SU-AI was getting too busy to get anything done (> 1 minute delay
just to get echo, several minutes to POXify a 10-line test file)
anyway... I was about to give up and break connection myself.
-------
∂02-JUL-76 1109 REG
To: LOTS.DIS[P,DOC]:;
I talked with Mac McKinley of ISI today about their 2040. They're planning
to make TOPS-20 more like the current version of TENEX, and they're going
to write an ARPAnet NCP. They have the sources for TOPS-20 under a "total
non-disclosure" agreement with DEC. ISI and DEC both expect that ISI's
result will be essentially the operating system for the 1080T configuration.
McKinley thinks they'll be on the net sometime in August. There won't be
any dialup ports. Once they're on the net, we may be able to get an account
for some modest amount of hacking.
π
∂02-JUL-76 1221 REG TOPS-20 sources
To: JMC, REG
Stafford called today and said the TOPS-20 source are normally $30K but
for us, $10K. However, the sources for RSXF20F (the program that runs in
the 11) are normally $50K, for us $30K.
The reason the 11 source is expensive is that obviously the most likely
place to add a brand x peripheral is on the 11.
My own feeling is that we should, budget permitting, buy the TOPS20 source
and expect to write our own 11 front end software should we need it.
(and we will need it if we're going to add fancy displays, links to the
370, etc.).
¬
∂03-JUL-76 1810 REM via AMET KL10 timing for R SPINDL
To: JMC, REM
CRUNCH.PLN[1,REM] used in all four figures to get timing data.
TIM1.DO[1,REM] used as dummy file to get second point for linear form
in crunch and uncrunch figures.
Spindle -- .18 seconds per 1000 words
Unspindle -- .10 seconds per 1000 words
Crunch-by-pages-and-spindle -- 1.01 seconds plus .46 seconds per 1000 words
Unspindle-and-uncrunch-by-pages -- .73 seconds plus .23 seconds per 1000 words
No error analysis was done, only one sample of each was done, so reliability
of data is not established. -- The system TIme command, increment, was used
to determine runtimes. Realtimes were not surveyed at all.
I presume the number of words refers to the original file, and I presume it
is packed 5 characters to the word so that the uncrunch rate is
a little less than 22,000 words per second. I suggest you include these
results in SPINDL.REM.
∂03-JUL-76 1825 JMC
I presume the number of words refers to the original file, and I presume it
is packed 5 characters to the word so that the uncrunch rate is
a little less than 22,000 words per second. I suggest you include these
results in SPINDL.REM.
[REM -- Well, .23 seconds per 1000 words is about 4,400 words per second,
perhaps you meant 22,000 CHARACTERS not WORDS per second... ]
Yes characters.
∂04-JUL-76 1305 SYS Your message file exceeds 10K words. Unless you delete some messages,
To: SGK, RWG, RHT, AIL, ALS, JP, TVR, RH, JMC
or BURP your mail file, or move your mail into your own disk allocation, your
entire message file will be purged.
∂04-JUL-76 1339 REG
To: LES, JMC, REG, JBR
For your information, since we put up the new disk drive, we've been using
additional space at the rate of approximately 430 tracks (about 1 megaword)
per day. This rate will (be forced to) abate in about 2 weeks.
∂04-JUL-76 1506 REG
To: LOTS.DIS[P,DOC]:;
Is anyone interested in an informal PDP-10 assembly language programming course?
If taught this summer, of necessity this course would include the A.I. Lab system
calls instead of TOPS-20 system calls, but that might be considered as only an
insignificant drawback. People interested in attending such a course (say
2 or 3 three hours two nights per week) should reply by MAIL to REG.