Re: Review XML Data in RIF

Adrian, Gary, Michael,

(Adrian, thanx for your review; see discussion below)

(Gary, I also answer you earlier comments, below)

(Michael, any chance you would have the time to have a look at the spec?)

I am almost done updating the RIF+XML data document [1], now: remains only 
to add a couple exmaples in the section "Combined interpretation of RIF 
BLD non-document formulas and XML data", and to complete the Glossary. I 
do not intend to include any text (except the current Editor's notes, in 
the conformance section, nor the appendices B and C. I propose to leave 
them in the document, nonetheless, as place holders, and a clear 
indication that the document is only a working draft. I will complete the 
work in the coming hours or days,

[1] http://www.w3.org/2005/rules/wiki/XML-Data

However, I would like the three of you to have a last look at the 
document, and give your imprimaturs. I think the current state of 
completion is sufficient for you to do that: there is no way we can make 
major changes at this stage, but we can add Editor's notes wherever you 
think one is needed.

I left a couple such notes myself, both as indications that thi sis still 
a WD, and to mark places where I think the proposed spec needs more 
discussion.

Below are responses to Adrian's and Gary's reviews.

Cheers,

Christian

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


Adrian wrote on 07/10/2010 17:59:42:
> 
> The document is clearly in the state of a first working draft and has 
many
> editors? notes left. Several sections and details are missing, like 
e.g.
> the conformance definition. I would propose to make this explicit in
> ?Status of this Document? section in the beginning, as readers might
> wonder why the RIF working group is finished while this document is 
still
> very much a working draft.

I completed the doc quite a lot, and I removed many of the editor's notes. 
But I left some, and some parts will remain missing.

But we resolved to publish it as a WD, anyway, so that it will be clear 
(in the status etc) that this work in progress (although it is less clear 
how we will manage to progress it any further, of course :-)

> Section 2 Definition (RIF+XML data combination)
> ?E is a, possibly empty, set of [data model nodes]XP that contains the
> information that is represented in the XML data
> 
> Q: Does it make sense to have just an XML schema attached with a RIF
> document without having any concrete XML data, i.e. an empty XML data 
set?

Consumer-side data. I reworked section 3: is that any clearer, now?

> C: The constraints on the combination are missing.

I fixed that, quickly and dirtily; but that should be enough for the WD. 
Or isn't it?

> C: Since the complete working draft has been written for using Strings 
to
> represent XPath expressions, I propose to remove the editors notes about
> rif:IRI as alternative.

Yes, I agree. I wondered what I should do about them. I left it to allow 
Gary and Michael to give an opnion without having to dig into a past 
version.

Gary, Michael, do you agree we should remove the note (in section 2.1).

> ?Another rule example, below, shows how another kind of XPath 
expression,
> used as an xs:string constant: "@xml:lang"</nowiki>>?
> 
> C: </nowiki>> should be removed

Done.

> A consequence is that such values have to be cast into the required 
types
> when used as arguments to RIF buit-in functions and predicates.
> 
> C: RIF built-in functions

Why not predicates as well?

> to add a frame formula to the condition, to check that the variable ?x 
is
> bound to an element that is, itself, named ex:Customer":
> 
> ?x["ex:Name" -> ?y]
> ?x["self::ex:Customer"->?x]
> 
> 
> C: This imposes an ordering of the conditions, since ?x first needs to 
be
> bound. So you cannot write:
> ?x["self::ex:Customer"->?x]
> ?x["ex:Name" -> ?y]

Disagree. sub-formulas are not ordered in an And (nor in an Or, for that 
matter).

> 2.2.2. Combined interpretation of RIF BLD non-document formulas and XML 
data
> 
> 
> C: [component-kind()]CD; [component-name()]CD etc. has not be defined 
before

As mentioned in the intro, the notation indicates that they are defined in 
XSD-CD, and that the definition is repeated in the glossary (but, of 
course, they not yet in the glossary, which may be why you were confused 
:-)

> Definition (RIF BLD+XML data combined interpretation)
> Itruth(Iframe(IDM(e))(IC("expr"^^xs:string), RIFValue(e, expr))) = t 
(true)
> 
> Q: Is it actually an interpretation of a frame as defined in the 
semantics
> of BLD or something which syntactically looks like a frame but 
semantically
> it is different (e.g. order dependent, based on an external schema, at 
least
> for the XPath functions etc.)?

No, it is actually an interpretation fo a frame as defined in the 
semantics of RIF BDL, see section 2.2.1.

I re-structures the whole section 2.2 to try to make that clearer. And I 
added explicit references to BLD definitions, wherever I thought there 
might be a doubt.

Does that work better?

> (T1, T2) ? Classes(S)2
> 
> Q: Why Classes(S)2?

I replaced Classes(S)^2 with Classes(S) x Classes(S).

Was that your question?

Gary Hallmark <gary.hallmark@oracle.com> wrote on 21/09/2010 17:03:49:
> 
> A medium-severity issue is that the ebnf in section 2 does not show an 
> abbreviated syntax, even though it is referenced in the following couple 

> of paragraphs. Also, an example contains "?x[".ex:customer"]->?x which 
> is not valid syntax according to the ebnf (nor xpath, I think). Maybe 
> you meant "../ex:customer", although that is not valid by the ebnf, 
either.

Corrected. I added the missin gparts in the EBNF.

> implementation-dependent: isn't this the same as "undefined"? And why 
> not define whether a singleton sequence should be a list, or not?

Right. I removed the definition :-)

> It is unclear whether PRD actions can modify an XML document (or an RDF 
> graph, for RDF combinations).

Yes, rules can modify the input data. In PRD, you can then use the Print 
action to serialize the state of the data after execution of the rules, I 
guess.

> Shouldn't consumer-side input, if allowed at all, also apply to RDF/OWL 
> data?

I am absolutely sure that it has to be allowed in the combination with 
XMLdata/schema. I do not know for sure about RDF/OWL graph: depends on the 
use cases, I guess.

> This version represents a "radical" change from previous versions, and 
> although I found nothing obviously broken, more eyes need to look at it. 

> And I need more time to think about it. But I think I like it.

Yes, I think I like it too :-)

Thanx to all for the constructive comments that brought us there.

Cheers,

Christian

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