COMMENT ⊗   VALID 00005 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	∂16-Dec-82  2112	Lee Erman <LErman at SRI-KL> 	RLL wallowing  
C00017 00003	∂Tue 15 Feb 83 19:10:29-PST	<CSD.LENAT@SU-SCORE.ARPA> [MSIMS@RUTGERS: A question about Eurisko/AM.]
C00029 00004	∂29-Mar-83  1121	CSD.GREINER@SU-SCORE 	[Gavin Hemphill <hemphill@NRL-AIC>: rll-1 paper]
C00036 00005	∂17:15 21-Apr: kwh@mit-mc, agre@mit-mc/cc, csd.lenat@score/su RLL info
C00040 ENDMK
C⊗;
∂16-Dec-82  2112	Lee Erman <LErman at SRI-KL> 	RLL wallowing  
To: DLENAT at SRI-KL, RDG at SU-AI

Doug and Rus,
	This is to reiterate my conversation with Doug this evening:

I would like to bring up at Teknowledge some RLL system that I can use for
my educational benefit.  I particularly would like to have a reasonably
extensive example use, to rummage in and play with.  I would expect other
Tek folks might also want to play.

Doug and I agreed I would get back to you two the week after Christmas, and
that would be a good time to prevail on you for some help.

		thanks,
			Lee
-------

∂TO LErman@sri-kl, dlenat@sri-kl/cc 11:34 29-Dec-82
re: RLL (s)wallowing
Lee -
	It's not clear just what RLL's state *really* is these days.
The most up-to-date version had been at USC-ISIB, on the <MANCOM.RLL>
directory.  Unfortunately there is no guarantee that that system is working;
the ISI "liason" left the project rather abruptly (for personal reasons,)
in the middle of a bundle of alterations.  If you still have an account there,
and ISI's good will, perhaps you could play with it there for a bit...
	A prior implementation was up and running at Rand, but, as
Rick can tell you, not only did that project all-but-die,
(at least with respect to its RLL association,) but Rand-Ai TOPS20 itself left!
	And as for me, personally: I have gone on to work with MRS, and will
eventually bring up RLL's good features within that system.
	However, if all of this doesn't dissuade you, do give me a call
at 497-1863, and I'll try to help (... "to bring up an implementation", not
"to further discourage you").
	By the way, how are you finding picturesque, pastoral, pleasant
Palo Alto?  Are you still recorder-ing?
	Russ
∂Tue 15 Feb 83 19:10:29-PST	<CSD.LENAT@SU-SCORE.ARPA> [MSIMS@RUTGERS: A question about Eurisko/AM.]
To: CSD.GREINER@SU-SCORE.ARPA

Please send Michael Sims, at Rutgers CS Dept, a copy of RLL.
Thanks
Doug
                ---------------

Mail-from: SU-NET host SAIL rcvd at 15-Feb-83 1632-PST
Received: from SU-DSN by SU-AI with PUP; 15-Feb-83 16:30 PST
Received: From RUTGERS by SU-DSN.ARPA at Tue Feb 15 16:27:14 1983
Date: 15 Feb 1983 1925-EST
From: MSIMS@RUTGERS
Subject: A question about Eurisko/AM.
To: Lenat@SU-AI
cc: Msims@RUTGERS

Doug,

Saul Amarel said that you will be sending along some listings and a
copy of RLL. I am most grateful.  I feel that I have a much better
chance of doing interesting work if I can learn from your experiences.

There is one point about Eurisko/AM that is puzzling me.  I have
gotten from your writings that when Eurisko attacked the domain
investigated by AM, that it did not perform better than AM.  Of course
there are things which Eurisko does and AM does not do, eg., the
discovery of heuristics and the flexibility of representation and
control.  Since Eurisko does so much more than AM it is natural to
expect Eurisko to be less efficient.  Also it seems most reasonable to
me that Eurisko did not discover new heuristics there because of the
extensive past investigations of number theory, as you have suggested.
However my impression is that the encoding of AM's knowledge base in
the more general framework of Eurisko didn't offer any advantages.

This issue of power is important to me, because my system will
probably sit closer to AM than Eurisko.  I am hoping that I can trade
some of the flexibility and generality of Eurisko for power in a
specific application domain.  Ignoring issues of the learning of
heuristics (a very big if to be sure), do you think this kind of trade
off is possible?  Practically this sort of trade off seems my only
choice if I want to pursue this line of research.  Attempting to build
a system with the scope of Eurisko seems out of the question with the
kind of time scale I have for getting a degree.

Let me know what you think.  Thanks.

-Michael
-------

∂26-Feb-83  1742	AGRE@MIT-MC 	[Nth attempt to send this, apologies for any lossage.]   
Received: from SUMEX-AIM by SU-AI with PUP; 26-Feb-83 17:19 PST
Received: from MIT-MC.ARPA by SUMEX-AIM.ARPA with TCP; Thu 24 Feb 83 08:51:28-PST
Date: Thursday, 24 February 1983  02:31-EST
Sender: AGRE @ MIT-OZ
From: AGRE @ MIT-MC
to:   greiner @ sumex
cc:   AGRE @ MIT-OZ
Subject: [Nth attempt to send this, apologies for any lossage.]
In-reply-to: Msg of 24 Feb 1983  00:50-EST from AGRE

Russell (Russ? whatever), I was speaking with Doug Lenat a couple weeks ago
about Eurisko, and he mentioned that there is a paper that describes RLL in
fair detail.  (I've got the AAAI-80 paper.)  I'd much appreciate it if you
could send me a copy, and a pointer to the source code too if that's OK, and
a copy of the CORLL paper and code if that's OK too, and anything else that
I might find useful in writing RLL-like and Eurisko-like things myself.
Thanks alot.

   Phil Agre
   MIT AI Lab
   545 Tech Sq Rm 711
   Cambridge MA  02139

   net: AGRE@MC works best

∂TO csd.lenat@score 16:13 3-March-83
Favors
Doug -
	I sent Phil Agre a copy of the RLL papers, together with some MRS
stuff.  Mailing the code is a bit more difficult: ISI's machines automatically
archive all files which have not been accessed for X amount of time,
which means that essentially all the RLL source code is currently "offline".
	I could ask Bill Mann to retrieve it, but that will cost us a favor.
Is it worth it?  Or should I point Agre and Sims to Mann directly?
Finally, I could simply (naively) send them my ISIB account and password,
and let them "learn" about off-line files, and methods of retrieval.
Other suggestions?

----
On other fronts: I talked to Johann the other day, and came away
pondering/discouraged/confused.  Could we meet sometime soon to discuss it?
One particular issue is how to encode the domain facts:
whether in my current, rather procedural (production rule based) manner,
or in a more declarative (deKleer-ative?) manner.
... and does any of this make any different, for a analogizing system which
strives to be impervious(sp) to idiosyncrasies of representations?

Anyway, my schedule is open, as always.  When is a good time for you?
	Russ

∂03-Mar-83  2356	CSD.LENAT@SU-SCORE 	Re: Favors     

Come over 2:30 on Tuesday.
Point Phil to Mann.
We will disucuss proc/decl on Tues. -- don't get depressed!
Doug
-------

∂TO 16:42 4-MAR-83
CSD.LENAT@SCORE/SU Re: Favors
Doug -

What should I do wrt Micheal Sims, of Rutgers?
Point him to Mann as well?

Also, Johann deposited my copy of MI3 the other day; did you pick it up?

See you Tuesday, 8/March, at ParcPlace.
	Russ

∂04-Mar-83  1909	CSD.LENAT@SU-SCORE 	Re: Favors     

Yes, same ptr.
I picked up MI3 and it is on my desk in MJH

See you Tue
Doug
-------

∂14:56 7-Mar
MSIMS@RUTGERS, AGRE@MC, CSD.LENAT@SU-SCORE/cc/subj RLL: ↑Code & ↑Documentation

Michael, Phil,

You should soon be receiving the (outdated) RLL documents you requested,
by snail mail.  That bundle has been augmented with miscellaneous MRS
documents, for reasons which will be clear later.  Providing the actual
code has proven a bit more difficult.

Some background: About a year ago, in the interest of graduating, my time
and energy shifted from RLL czar to thesis-ing.  Others, including Dr
William Mann and some of his coworkers at USC/ISI, continued to develop
and augment RLL.  For that reason the most up-to-date version had been at
USC-ISIB, on the <MANCOM.RLL> directory.  Unfortunately there is no
guarantee that that system is working; the ISI liason/principle-updater
left the project rather abruptly (for personal reasons) in the middle of a
bundle of alterations.  Furthermore, much of the needed code has been
archived (and deleted).  To access it will require either competence with
the TOPS20 retrieval command, or some ISIer's assistance; or more likely,
both.  MANN@ISIB may be able to provide the latter, or at least point you
to the right person.  As this letter implies, you have both my blessing
and encouragement to examine/use/update whatever code you can; and my
permission (subject to Dr Mann's approval) to use that ISI account.

I personally am coding in MRS these days, and eventually hope to bring up
(all and only) RLL's good features within that system.  As the documents
show, the two systems are fairly similar in theme and capabilities; the
major differences (wrt me) lying in the facts that someone else is
responsible for its adminstrative and technical support, promotion, etc.;
and that MRS, having a larger "constituency", will probably be maintained
(if not for perpetuity at least for a while).

However, if all of this doesn't dissuade you, do give me a call at (415)
497-1863, and I'll try to help ("... to bring up an implementation", not
"... to further discourage you").

Russ

RDG@SAIL (or GREINER@SUMEX)
∂29-Mar-83  1121	CSD.GREINER@SU-SCORE 	[Gavin Hemphill <hemphill@NRL-AIC>: rll-1 paper]
Subject: [Gavin Hemphill <hemphill@NRL-AIC>: rll-1 paper]
To: rdg@SU-AI.ARPA
cc: csd.lenat@SU-SCORE.ARPA

Return-Path: hemphill@nrl-aic
Received: from nrl-aic by SU-SCORE.ARPA with TCP; Tue 29 Mar 83 07:40:07-PST
Date: 29 Mar 1983 08:57:39-PST
From: Gavin Hemphill <hemphill@NRL-AIC>
To: csd.greiner at su-score
Subject: rll-1 paper

			Defence Research Establishment Atlantic, 
			Box 1012, 
			Dartmouth, Nova Scotia, 
			B2Y 3Z7

			Tel: 1-902-426-3100 ext. 189

Dear Russell Greiner, 

	I have just finished reading your interesting paper, 
"RLL-1: A Representation Language Language", which is an expanded 
version of the paper published in AAA-80.  I'm keen to learn more about 
RLL.  

	A version of the UNIT PACKAGE (see 1 below) is being used in 
our research to build an expert system to interpret acoustic data.  
(We've taken to calling it UNIT*.)  Last year, I implemented an object-
centered approach for inheritance roles in UNIT*.  I concluded that more 
of the package would have to be changed to an object-centered implementation 
before it would be truly extensible.  Imagine my delight to find that you 
had already designed another language to do just that!  

	What is the current status of the RLL project? Can you tell me 
more about the implementation details of RLL-1?  How different is RLL-1 from
the UNIT PACKAGE?  I would appreciate answers to any of these questions and 
any other information available about RLL.  I may be reached at the above 
address or by sending NET mail to G.HEMPHILL at SU.SCORE.  
I look forward to hearing from you.  Thanks.  

					Sincerely, 
					Ann Dent

1) R.  G. Smith and R. E. Friedland, Unit Package User's Guide, 
DREA TM 80/L, Defence Research Establishment Atlantic, Dartmouth, 
Nova Scotia, Canada, December 1980.  (Also published as HPP-80-28, 
Heuristic Programming Project, Stanford University. )

-------

∂16:02 30-Mar: G.HEMPHILL@SCORE, CSD.LENAT@SU-SCORE/cc/subj RLL: ↑Code & ↑Documentation

Dr Dent,

	It's not clear just what RLL's state *really* is these days.
About a year ago, in the interest of graduating, my time
and energy shifted from RLL czar-ing to thesis-ing.  
I am personally coding in MRS these days, and eventually hope to bring up
(all and only) RLL's good features within that system.  
These two systems are fairly similar in theme and capabilities,
(as both grew out of the same meetings, several years ago;)
the major differences (wrt me) lying in the facts that someone else is
responsible for its adminstrative and technical support, promotion, etc.;
and that MRS, having a larger "constituency", will probably be maintained
(if not for perpetuity at least for a while).
Let me know if you would like me to send you the set of MRS documents.

Returning to RLL, proper, a form of its code does still exist, sorta.
Dr William Mann and some of his coworkers at USC/ISI continued to develop
and augment RLL even after I left.  For that reason the most up-to-date
version had been at
USC-ISIB, on the <MANCOM.RLL> directory.  Unfortunately there is no
guarantee that that system is working; the ISI liason/principle-updater
left the project rather abruptly (for personal reasons) in the middle of a
bundle of alterations.  Furthermore, much of the needed code has been
archived (and deleted).  To access it will require either competence with
the TOPS20 retrieval command, or some ISIer's assistance; or more likely,
both.  MANN@ISIB may be able to provide the latter, or at least point you
to the right person.  As this letter implies, you have both my blessing
and encouragement to examine/use/update whatever code you can; and my
permission (subject to Dr Mann's approval) to use that ISI account.

Anyway, if all of this doesn't dissuade you, do give me a call at (415)
497-1863, and I'll try to help ("... to bring up an implementation, and/or
to discuss the ideas of RLL/MRS", not "... to further discourage you").

Russ

RDG@SAIL (or GREINER@SUMEX)
∂17:15 21-Apr: kwh@mit-mc, agre@mit-mc/cc, csd.lenat@score/su RLL info

First, I just spoke with Wes, who told me he was going to use Doug's
rll system, rather than my hard-to-access ex-system.

Second, below is a list of the KBs I had, as of 1982:
(Do realize I've not examined these in years -- and my memory may be faulty...)

[Kept on RAND-AI: <GREINER.RLL>]

    RLL		- held the basic system
    USERS	- facts about each user
    LISPFNS	- some facts about some LISP functions
    SLOTS	- stored facts about each slot in starting RLL system

ALL of the above were needed for RLL to run.
---

    HEURS - list of heuristics, which EURISKO could use
    GENLINFO - overview of the world -- VERY sketchy
    EURISKO - basic agenda based control scheme.

The above three were nice -- still very fragmentary.
---

    SPILL - see "Building Expert System" book
    RATALE - beginning of rationalization, I think???

[Kept on SCORE:  <CSD.RLL>]

<<none of these were ever really used.  Ask Doug about their actual contents>>
    BIOLOGY - basic facts about animals, etc.
    SETS - facts about sets, again coarse 
    NUMBER - facts about number theory
    MATH - gen'l facts re: mathematics -- required by SETS and NUMBERS
---
    RESOL - a resolution-theorem-prover, to be. 
	[Begun by Larry Hines, of UofTexas]


***---***
The naming scheme (as we discussed on the phone)
    *.KB - file which only CORLL knew how to use
    *.XKB - human readable version of a knowledge base
	-- I think in the format  unit1(p1 v1 ... pn vn)unit2(p1' v1' ... pn' vn')
	where unit1, unit2 are the names of units, and p-i's are properties
	(slots) and the v-i's are the values of these slots.
    *.. - is the InterLisp file, associated with this KB
    *.COM - ILisp compiled code, of course

-----

Third, the following people have each,
at one point or another, expressed interest in using this RLL system:
    Caroll Johnston - JOHNSTON@SUMEX ?
    Lee Erman - LErman@sri-kl
    Michael Sims - MSIMS@RUTGERS
    Gavin Hemphill - hemphill@NRL-AIC or G.HEMPHILL@SCORE

----
Good luck.
Any word from Bill Mann wrt copying that tape?
Let me know if there is anything else I can do.

	Russ