Re: BLD review

Hi Igor,
Thanks again. I fixed most of the problems.
Here are replies to some issues that you raised.


	--michael  

> In RIF-BLD I like the dual approach: instantiation of FLD,
> and self-contained definition. I think it should be kept that way.
> Examples in the presentation syntax are difficult to read
> (see below for a possible improvement).
> 
> Regards,
> Igor
> 
> 
> 2.0.1. The syntax
> 
> * Supported formulas
> 
> * RIF-BLD condition
> Here I would be explicit that a condition can be used to form queries.
> I don't see any need to form another working group which will
> specify RIF queries :-)

done

> * RIF-BLD rule
> I would explicitely state that equality can appear in the head
> as well.

done

> 2.0.5. Formulas
> 
> I would move (and reformulate) the statement: "Formulas using
> the above definitions are RIF-BLD conditions" in front of the
> four items.

done

> 2.0.6. EBNF Grammar
> 
> I would omit rif:local (and make it default) in the presentation
> syntax. This would make the examples much more readable.

We could do that, but we decided that the presentation syntax should not
have sugar and other attributes of a concrete language. This example does
look a bit verbose, but, on the other hand, we keep the presentation syntax
at bay.


> 2.0.7. XML serialization
> 
> "Positional information is optionally exploited only for the
> arg role elements"
> I guess this is not optional, but required for the positional terms.
> Shouldn't we also allow optional positional information to
> be exploited by the formula and rule roles?
> Otherwise I don't see how can one ensure roundtripping
> of Prolog rules.

I'll let Harold deal with that.


> 2.0.9. Subdialects of RIF-BLD
> 
> I wonder if it were not useful to also define 'syntactic' equality
> (eg, unification) which can be used only in rule conditions, and
> not in the heads? This would make more sense in Core.

This is not necessary. I fixed the problem by just saying that the core
will not allow equality formulas in conclusions, while still allowing them
in rule premises.

> * Supported type sof terms
> Compared to RIF-FLD...
> 1st item, last sentence:
> a variable ... -> a variable cannot range over atomic formulas

Actually, the current formulation is correct. It says that vars are not
atomic formulas. "Ranging" is an implementational notion. (Note that in FLD
vars *are* aromic formulas.)

Received on Thursday, 14 February 2008 07:26:29 UTC