COMMENT ⊗   VALID 00004 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	See all RLL.BBD[rdg,dbl].
C00003 00003	∂05-Nov-80  1342	SASTRY at USC-ISIF 	RLL applications    
C00005 00004	∂10-Nov-80  1401	KEIRSEY at USC-ECL 	Details   
C00009 ENDMK
C⊗;
See all RLL.BBD[rdg,dbl].
∂05-Nov-80  1342	SASTRY at USC-ISIF 	RLL applications    
Date:  5 Nov 1980 1330-PST
From: SASTRY at USC-ISIF
Subject: RLL applications
To: greiner at SU-AI


Hi,

I obtained a copy of the paper describing RRL-1. Thankyou. Mentioned in it is 
a reference to some work on applications of RLL to VLSI layout. I am also 
interested in this although I have not done any real work along this path. 
What I would like to know is what problem is being looked at and what has 
been done. 
Could you send me any info concerning this or direct me as where I amy obtain 
such information ?. Thanks again.

Sarma Sastry. (Sastry@USC-ISIF)



-------

Mailed to SASTRY, STEFIK@PARC&HBOWN@SCORE/cc  16:49 5-Nov-80
VLSI/RLL interconnections
	Mark Stefik (STEFIK@PARC) and Harold Brown (CSD.HBROWN@SCORE) are 
currently building a system which will help plan and design VLSI layout.
(According to Mark, ISI's Danny Cohen knows all about it.)
	I sent them a copy of your message.
Russ

∂10-Nov-80  1401	KEIRSEY at USC-ECL 	Details   
Date: 10 NOV 1980 1401-PST
From: KEIRSEY at USC-ECL
Subject: Details
To:   RDG at SAIL
cc:   keirsey

I have read your paper on the Details of RLL.  On the whole I thought
it was good.  Unfortunately, I had the naive notion that RLL-1 would be
simpler in conception and implementation.  As I read it, I got a feel
for how you represented some things, but it is clear to me that real
understanding of a particular knowledge base will only come through
use.  One important issue to me is how one can understand another
person's knowledge base (given that they are large).  Do you have any
thoughts on this?

I have a very detailed question.  On Diagram 1 of Details of RLL, what
is the significance of the AND-node notation (AnyConcreteThing&Unit,
AnySlot)?
-------

Mailed to Keirsey@ECL 17:37 11-Nov-80
Possible Answers
First, to answer the easy question:
In general the subclasses form a DAG, as opposed to a tree.
The "AND-notation" you asked about means that those particular subclasses
are disjoint.

As to your other inquiries:
You are correct in expecting the ideas of a rll to be simple.
The document you received described a particular implementation, slanted
towards things which I thought were relevant. Furthermore, many of the
details presented reflected facets which seemed
appropriate a while back; and which have not yet been updated to a better,
cleaner form.
Prof Genesereth, in writing up his description of MRS 
has stressed the simplity of its design. I recommend that paper
(HPP-80-24) to confirm the accuracy of your intuitions.

Yes, Prof Lenat & I have worried quite a bit about the question of how
to insure that each KB is understandable and unambiguous.
We considered using a predefined formal specification,
(such as predicate calculus),
and rejected these as too hard to understand, and not perspicious enough.
The sub-section 4.4 of the AAAI paper (on pages 167-168) describes another
approach - using a "Semantics" slot for each slot-unit, telling what this
slot "really" means.  The "Meta-Description & Modifiability" memo (HPP-80-18)
by Lenat & Genesereth, elaborates these ideas.

	If this doesn't answer your questions, let me know and I'll try again.
Russ