Ramesh Patil <ramesh@ISI.EDU>
Message-id: <199402152224.AA24748@quark.isi.edu>
To: phayes@cs.uiuc.edu, macgregor@ISI.EDU, cg@cs.umn.edu, interlingua@ISI.EDU
Reply-To: ramesh@ISI.EDU
Date: Tue, 15 Feb 94 14:24:17 PST
From: Ramesh Patil <ramesh@ISI.EDU>
>>So, from the standpoint of translating such things as definitions, types,
>>and propositions, it makes little difference which semantics KIF
>>chooses to specify.  The important thing is that each of these 
>>is given a standard "name" that a translator can look to when translating
>>in and out of KIF.  If that name has an elegant semantics associated
>>with it, so much the better.
>>I can't imagine a standard being approved that doesn't specify a semantics.
>>But for practical purposes, it matters little what that semantics happens
>>to be.
>Bob, maybe you just intended to be rhetorical, but if you really mean this
>then we should all stop now. You can use the integers as a suitable KIF. It
>supplies as many names as you could possibly want, and if the semantics
>doesnt matter, what else is there?

 I believe Bob was expressing a level of frustration with the KIF as it
currently stands.  I agree with you as well as Bob: let me explain why.

I think having a well defined semantics for types and propositions is absolutely
essential for knowledge representation for the reasons you and others on this
mailing list have described.  The arguments have been regarding how exactly to
specify these.  On the other hand, as it stands KIF does not, in practical
sense, allow us to expresses these notions, that is, as a translator I cannot
easily state that the fillowing expression is a proposition.
This is a SERIOUS handicap for developing even approximate/practical translators.

Thus, even when we disagree exactly what the foundations for propositions and
such should be, there is a common understanding that we can build on.  We agree
enough to allow us to discuss these issues in EMAIL, and SIMILARLY in building
practical translators.  I am well aware of the fact that translators built on
such shakey underpinnings will not be sound/correct or have any provably nice
propoerties.  But they fill a need, and can be built while the exact semantics
is being developed.

Thus, let us agree to agree that KIF needs to have types and propositions (among
other things).  Let us agree on the syntax and reserve names for these so that
translation work can go on in parallel with the semantic specification/debate.