[SCL] Proposed additions to the SCL draft standard

[Chris Menzel]

On Sun, May 25, 2003 at 11:19:35PM -0400, John Sowa wrote:
> Chris,
> 
> I agree with most of your comments on Tanel's note, but I would
> think that we can do more to accommodate comments (or annotations).

I'm all for annotations, properly accommodated! :-)

> TT>>- Annotations (might be called also comments or documentation
> >> strings). Free text (strings) may be attached to terms
> >> and formulas.
> >
> CM> I see the need for annotations in SCL implementations, but the suggested
> >approach won't work.  You propose including them as terms.  But the
> >underlying logic/semantics of SCL is classical, and in a classical
> >semantics all terms must have a semantic value.
> 
> All terms, yes.  But the abstract syntax can specify things which
> are not terms and which do not have a semantic value.  

True!

> I think that it would be better to adopt the principle that for every
> syntactic rule there is a corresponding semantic rule.

Hm, not what I was expecting, since I thought the idea was to add a
syntactic rule for a class of things that would NOT have a corresponding
semantic rule.
 
> For an example using BNF rules, a language might have a rule that
> defines a constituent X as an A followed by a B:
> 
>    X -> A B
> 
> Then if we wanted to annotate X, we could extend that rule by saying
> 
>    X -> A B [ Comment ]
> 
> This says that X could be annotated with an optional comment
> following the constituent B.  But the semantic interpretation would
> be identical to the interpretation of the rule without the comment.

Ok, now I see what you mean (though I don't think what you said was the
best way to say what you meant! ;-)  I have no serious problem with
this; though I really do wonder if it's a part of our ABSTRACT syntax to
legislate where people put their comments.  The current syntax
seems to me to accommodate them just fine simply by not preventing
anyone from putting them pretty much wherever they want.

> CM> If some sort of annotating facility is going to be added, it must be
> >thought of as some sort metalinguistic device.  Hence, if we agree some
> >notion of annotation is important, then we need to have a section of SCL
> >that describes such a device, and offers counsel about how to implement
> >it.
> 
> I think that comments could be supported by the SCL abstract syntax
> by just adding a category called "Comment" to the rules that define
> those categories that can be annotated -- by an approach similar to
> the one sketched out above.

Yes, agreed; but my concern above remains.

> I agree with you on the following point:
> 
> CM> I don't think the semantics for predicate-bounded quantifiers will be
> >genearalizable to a semantics for numerical quantifiers.  But it will be
> >easy enough to implement an appropriate semantics.
> 
> I also agree with your comments about identity.  But there are other
> features of the "wild-west syntax" that illustrate points that are
> easily confusing to people who are familiar with more common notations
> for logic.  So we do need a definition of a subset language with
> restricted quantification.

Well, the current implementation of restricted quantification is
actually pretty civil-east rather than wild-west, it seems to me.
Though I agree a *lot* more can be done to carve out particularly useful
SCL language types (e.g., traditional first-order languages, languages
with/without restricted quantifiers, etc).

-chris