Re: FLD review

Thank you Igor,
I have incorporated your suggestions. A reply to some of your
comments and also some questions below.


	--michael  

> I would suggest to add a short note at the end of
> Syntactic framework about the relation to sorted logics.
> It seems that by combining symbol spaces and signatures
> (and allowing for explicit representation in the language)
> one can easily specify sorts.

Done.

Signatures actually generalize the standard notion of sorts in logic. The
difference with respect to the earlier use of sorts in BLD is twofold:

    1. There are no sorts in BLD, only in FLD.
    2. Sorts (under the name of signatures) were decoupled from symbol
       spaces. This gave us more freedom and eliminated confusion. Symbol
       spaces are close to data types and people thought that the purpose of
       sorts is to define data types (which it was not).  It is the tie-up
       between the sorts and the data types that was the cause for all the
       previous troubles.


> In Semantic framework you introduce semantic structures.
> I like the term and it seems that it is broader (eg, explicit
> parametrization of truths values?) then interpretations 
> (as used in LP and DL). Some short note on the distinction 
> would be helpful.

This is a standard term in logic. For instance, Enderton's classic book
uses it.  If you search the web then you will find both, although
"interpretation" is more common.

I prefer "semantic structure" because "interpretation" is an overloaded
word.  I added a note that this is the same notion as "interpretation" in
some books.

> In the last paragraph there is a distinction between assumed
> set of semantic structures in DL and LP. This is very interesting
> and relevant, but I don't grasp it. In LP one deals with Herbrand
> interpretations only (corresponding to S), and then shows the 
> relation between inference procedure (SLD) and minimal Herbrand model. 
> If you could slightly expand and clarify the last sentence it
> would be helpful.

We are not talking about the proof theory (SLD) -- only about logical
entailment.  What you said about the Herbrand interpretations is true and
confirms what is written in the overview. So, I do not understand what you
see as being unclear here.

> XML serialization statement could also be extended to a
> paragraph (mentioning classes and roles, XSD and simple
> translation between presentation syntax and XML).

This will be extended when that XML framework section will gain some flesh.


> Do you mean that support for all the subtypes of xsd: string and
> xsd:decimal has to be provided by RIF compliant systems?
> Shouldn't we make them explicit, then?

I am maybe looking at a wrong spot, but where does it say that all (or
some) subtypes must be supported?

> Notes on RIF-compliant support...
> A RIF-consuming system
> Here it is mentioned what it is _not_ required to support.
> What is required?

The list of the mandatory symbol spaces.

> Logical entailment.
> Specifying the set of intended models - I find this a great
> idea to parametrize dialect, but some more explanation or
> examples would be helpful.

This is quite standard now, and I provided a reference to Shoham's paper
where this idea originates. What kind of examples do you think are needed?
It already has 4 examples: FOL, Horn, Stable, and WFS.

Received on Thursday, 14 February 2008 06:25:28 UTC