COMMENT ⊗   VALID 00004 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002		Agenda of Discussion Items and Things to Consider
C00010 00003	GET CORLL.LSP←<CSD.GENESERETH>CORLL.L
C00012 00004	CORLL.LSP,CORLL.MACROS,MACROS.FASL,MACROS.LSP,MRS.LSP,MRS.INFO,MRSDEMON.LSP,MRSINDEX.LSP,MRSLOGIC.LSP,MRSMETA.LSP,MRSOPT.LSP,MRSPL.LSP,MRSPL.INFO,MRSPN.LSP,MRSREP.LSP,MRSRESIDUE.LSP,MRSS.LSP,MRSSETS.MRS,MRSSUBWORLD.LSP,MRSWHY.LSP,MUSER.LSP/l[lisp]/d[dover]/c/n/z/u
C00015 ENDMK
C⊗;
	Agenda of Discussion Items and Things to Consider

					Items 1-6 : LGC, 3 Oct 81
					 (addenda 7 Oct 81)

1.  Introduction of some good system of (at least partially)
    automated bookkeeping and documentation to keep us from
    drowning in complexity and incomprehensibility nightmares
    as MRS grows and changes over time in our hands.
    (Masterscope for MACLISP; how do you guys handle this problem
    in the work on S-1 NIL, etc.?)

2.  Introduction of free variable formulas as general concepts
    to permit implementation of JMC's 1958 idea about storing useful
    information at appropriate levels of abstraction of formulas (SIP, p. 409).
LGC--Perhaps lambda-binding would be better than free variables for this.
LGC - 15 Oct 81
    Actually, ordinary propositions can be used for this purpose, once it is
    seen that the mere presence of a propositional representation in memory is
    not to be confused with belief in (or any other specific attitude toward)
    the proposition in question.  Just as a prop. can be believed, desired,
    etc., so can it have useful information stored with it that is relevant to
    those attitudes.

3.  Introduction of sortal information for all variables to make both (2), and
    quantification generally, more useful and natural.
LGC--The new variable-free notation will handle this by giving the lambda-
     operator the same general form as a quantifier.

4.  Consider using explicit quantifiers instead of Skolem terms in the
    internal representation (how would this affect matching processes?)
LGC--The new version of quantifiers will probably help matching --
     check this out with rpg.

5.  Shouldn't the PATTERN property of compound statements contain a
    fully expanded list representation for ease of matching, instead
    of the current practice of including propositional atoms as
    constituents of PATTERNs?  Observe:

(setq ifce (car units))
|(IF (COLOR CLYDE GREY) (ELEPHANT CLYDE))|    ; (a propositional atom or unit)
(get ifce 'pattern)
(IF |(COLOR CLYDE GREY)| |(ELEPHANT CLYDE)|) 
(sprinter (plist ifce))


(MYWORLD (GLOBAL)
	  DATA
	  (|(IF (EXPRESSION $X) (MYTOUNASSERT $X DEMON-UNASSERT))|)
	  PATTERN
	  (IF |(EXPRESSION $X)| |(MYTOUNASSERT $X DEMON-UNASSERT)|))
;;  full list representation:  (IF (COLOR CLYDE GREY) (ELEPHANT CLYDE))
---------------

6.  Consider normalization of variables to simplify indexing and to
    reduce MEMSAMEP and its complex subordinates to MEMBER.
    (see MRS.LSP[MRS,LSP]/3p/49l)
LGC--A better solution: redesign the representation language to eliminate
     variables and thus remove the NEED for normalization, etc...

LGC -- 9 Oct 
7.  The question of indexing the propositional data base needs consideration.
    The present approach in MRS is to distinguish the input language from the
    internal language, to compute for each input proposition a normalized
    internal version, and then to use the normalized internal formulas
    themselves as p-names of atoms interned on the obarray.  In the first
    place, the present criterion of identity for propositions is faulty -- the
    normalization process is logically incorrect in some ways and insufficient
    in others.  Secondly, there is the fact that the "index" of propositions
    for each object and/or relation concept is a simple list; some provision
    should be made for adding more tree structure to this scheme.

    Also, there are some concerns that RPG has about using the obarray in the
    way described.  [What exactly are these concerns? ]

LGC -- 11 Oct
8.  How hard would it be to implement a facility in AT/MRS for user
    declaration of a data-type of CONSTANT ATOMS which:
	a) have no value cell (a saving in space),
	b) are never evaluated (like numbers, T, and F),
	c) and cannot be SET or SETQ'd (like numbers, T, and F).?

LGC -- 15 Oct 81
9.  What about the use of structure-sharing in the storage of propositions?
    Would this be worth the extra effort required to implement it?

10. Keep in mind that the propositional representations are to be used as
    the objects of all the attitudes, not just beliefs (whether this can be
    fully carried out for plans and procedures is an interesting theoretical
    question).  Thus, the mere presence of a propositional representation in
    memory signifies only that it plays some role there, but not which roles
    (e.g., belief, desire, etc.) those happen to be in a given case.
GET CORLL.LSP←<CSD.GENESERETH>CORLL.L
GET CORLL.MACROS←<CSD.GENESERETH>CORLL.MACROS
GET MACROS.FASL←<CSD.GENESERETH>MACROS.FASL
GET MACROS.LSP←<CSD.GENESERETH>MACROS.LSP
GET MRS.LSP←<CSD.GENESERETH>MRS.L
GET MRS.INFO←<CSD.GENESERETH>MRS.INFO
GET MRSDEMON.LSP←<CSD.GENESERETH>MRSDEMON.LSP
GET MRSINDEX.LSP←<CSD.GENESERETH>MRSINDEX.L
GET MRSLOGIC.LSP←<CSD.GENESERETH>MRSLOGIC.L
GET MRSMETA.LSP←<CSD.GENESERETH>MRSMETA.L
GET MRSOPT.LSP←<CSD.GENESERETH>MRSOPT.LSP
GET MRSPL.LSP←<CSD.GENESERETH>MRSPL.L
GET MRSPL.INFO←<CSD.GENESERETH>MRSPL.INFO
GET MRSPN.LSP←<CSD.GENESERETH>MRSPN.LSP
GET MRSREP.LSP←<CSD.GENESERETH>MRSREP.L
GET MRSRESIDUE.LSP←<CSD.GENESERETH>MRSRESIDUE.L
GET MRSS.LSP←<CSD.GENESERETH>MRSS.L
GET MRSSETS.MRS←<CSD.GENESERETH>MRSSETS.MRS
GET MRSSUBWORLD.LSP←<CSD.GENESERETH>MRSSUBWORLD.LSP
GET MRSWHY.LSP←<CSD.GENESERETH>MRSWHY.L
GET MUSER.LSP←<CSD.GENESERETH>MUSER.LSP

(load '|<csd.genesereth>demo.lsp|)
(load '|<csd.genesereth>mrs.demo|)
CORLL.LSP,CORLL.MACROS,MACROS.FASL,MACROS.LSP,MRS.LSP,MRS.INFO,MRSDEMON.LSP,MRSINDEX.LSP,MRSLOGIC.LSP,MRSMETA.LSP,MRSOPT.LSP,MRSPL.LSP,MRSPL.INFO,MRSPN.LSP,MRSREP.LSP,MRSRESIDUE.LSP,MRSS.LSP,MRSSETS.MRS,MRSSUBWORLD.LSP,MRSWHY.LSP,MUSER.LSP/l[lisp]/d[dover]/c/n/z/u