[SCL] minutes of telecon 8 April 03

[Jim Hendler]

Pat - sotrry I couldn't make the call - a couple of quick responses


>
>2. Core language. We agreed that the core should allow restricted 
>quantification as useful syntactic sugar, but should not attempt to 
>include types or more advanced concepts.
>
>One issue we did NOT discuss, by the way, is whether the core should 
>allow function symbols and full terms (I vote yes), and whether or 
>not it should restrict relations and functions to have a fixed arity 
>(I vote no*); and if so or not, whether or not a user can assert 
>that a given relation or function has a certain arity. Certainly in 
>the full syntax, arity could presumably be a property of a relation 
>, and perhaps we should provide a logically defined syntax for 
>asserting those properties. Comments?
>
>(* To clarify: not including an arity constraint into the syntax, 
>that is: it would be too much trouble to take it out again for the 
>generalized syntax, and I don't think it is really necessary even 
>for conventional FO syntax. In the conventional language, two uses 
>of the same symbol with different arities are essentially treated 
>semantically and in proofs as different symbols, and does no harm.)


While I agree with you, in most web languages (incl a lot of the XML 
derivatives) arguments are essentially "tagged" (like the 'modern' 
convention of having optional arguments in LISP) -- so a binary v. 
n-ary relation w/same name would be confusing to most web language 
types.  I think you make the right decision, but let's be careful to 
explain it.

For those who didn't follow - in XML or RDF/XML or similar, I would 
say something like (I'm cheating on details)

<person :John>
   <name>John</name>
   <size>large</size>
</person>

<person :Mary>
   <name>Mary</name>
   <size>Medium</size>
   <inLovewith :John />
</person>

and this would just be the same n-ary person relation in both cases, 
with different arguments.  Few if any web languages prefer
   Person(name, size, inLoveWith, bestFriendOf, shoeSize, ...)
although XML DTDs allow it and make use in some cases (like Peter's 
abstract syntax of OWL, which several XML folks have criticized for 
not using tagged arguments).



>
>-------
>
>3. Built-in logical names and externally defined namespaces
>
>The core should include equality.
>The full syntax should include an 'is-a-Relation' predicate.
>(and Arity predicates?)
>Special name domains should provide a built-in logical predicate 
>true of entities in the domain, so that an atom formed by applying 
>that predicate symbol to any special name with the appropriate 
>syntax will be true.

like the idea, will need to be careful ion the web.
>
>We need a way to specify the syntax of names in a special-name 
>domain. We did not reach agreement on how to do that.  Here are some 
>options:
>
>1. Do not provide one, but instead restrict ourselves to a small set 
>of built-in special name categories, eg numerals and quoted 
>character strings, or maybe a set based on the XML schema built-in 
>datatype domains. This would be quick and dirty but might be worth 
>doing even as an exercise for ourselves :-)
>
>2. Simply require that whoever defines 'an SCL language' provide a 
>BNF for their particular syntax of special names, and write their 
>own parsers.

If SCL intended for the web, this is a sure fire loss.  Perhaps 
instead of BNF we could use XML (DTD or schema) and then we'd get 
something similar -- wouldn't be quite as general, but would make it 
easier to build.

>
>In my view, neither of these is really adequate. The second option 
>basically makes it impossible to write a general SCL parser, since 
>particular SCL languages may have arbitrary syntactic complexities 
>which might need to be parsed, and the reversion to BNF allows too 
>low-level and detailed a possible relationship between special name 
>syntax and the rest of the language.  I would be happier if we could 
>provide a more coherent mechanism which defines the syntactic 
>interface between special name domains and the rest of the language 
>more clearly. However I confess I have no clear idea of how to do 
>that in an appropriate way.
>

agreed


-- 
Professor James Hendler				  hendler at cs.umd.edu
Director, Semantic Web and Agent Technologies	  301-405-2696
Maryland Information and Network Dynamics Lab.	  301-405-6707 (Fax)
Univ of Maryland, College Park, MD 20742	  240-731-3822 (Cell)
http://www.cs.umd.edu/users/hendler