Re: RIF+XML data ready for publication

Hi Gary,

Gary Hallmark <gary.hallmark@oracle.com> wrote on 21/10/2010 18:25:36:
> 
> the following suggests that the slot name may only be a literal string:
> 1. For all the [nodes]XDM e ? E, 
Itruth(Iframe(IDM(e))(IC("expr"^^xs:string), 
> RIFValue(e, expr))) = t (true) for any well-formed XPath expression,expr
> , for which RIFValue(e, expr) is defined; 

I see: "Ic" makes the difference, in the clause.

I mean, the clause is about the interpretation of "expr"^^xs:string, 
Ic("expr"^^xs:string), not the (literal) constant itself.

By definition of fn:concat, Ic(fn:concat("ex" "pr")) = Ic("expr") (for 
instance).

> Another problem with using strings for xpaths is that any typo in 
> the xpath is still a string, and is therefore syntactically valid, 
> but will give unexpected results. Therefore I suggest again that we 
> use a new symbol space, e.g. rif:xpath, rather than overloading 
xs:string.

I see your point. 

On the other hand, under the translation paradigm, that may not that much 
a problem: if there is a typo, either your original rule was wrong or your 
translator has a bug; in either case, it is not RIF's problem. I mean: 
garbage in, garbage out!

Do we want RIF to allow this kind of checks? Personally, I do not think 
so. I fear a dangerous slope (remember the discussion about rule language 
VS rule interchange format?): if we want to go in that direction, using 
XPath/XSD-CD expressions as strings is only a little part of the issue 
(e.g. the slot in your frame is an IRI and there is a typo: syntactically 
valid, unexpected results. Oh, ok, that's a slightly different problem 
:-).

In any case, changing the spec, at this point, to introduce a new symbol 
space such as rif:xpath (actually, we would need two: rif:xpath and 
rif:xsd-cd) would not be practical.

I propose that we add the following editor's notes in section 2.1 (I added 
them already, actually):
Editor's Note: Alternatively, an additional symbol space, rif:xpath, could 
be added to RIF built-in data types: in that case, XPath expressions would 
be represented as rif:xpath constants. 

and
Editor's Note: Alternatively, an additional symbol space, rif:xsd-cd, 
could be added to RIF built-in data types: in that case, XSD-CD 
expressions would be represented as rif:xsd-cd constants. 

Would that be enough at this point?

Cheers,

Christian

IBM
9 rue de Verdun
94253 - Gentilly cedex - FRANCE
Tel./Fax: +33 1 49 08 29 81

Sauf indication contraire ci-dessus:/ Unless stated otherwise above:
Compagnie IBM France
Siege Social : 17 avenue de l'Europe, 92275 Bois-Colombes Cedex
RCS Nanterre 552 118 465
Forme Sociale : S.A.S.
Capital Social : 612.509.964 ?
SIREN/SIRET : 552 118 465 03644