Re: Text for FAQ re rdf:datatype="&rdfs;#XMLLiteral"

Patrick.Stickler@nokia.com wrote:

> 
> 
>>-----Original Message-----
>>From: ext Frank Manola [mailto:fmanola@mitre.org]
>>Sent: 11 August, 2003 16:13
>>To: Stickler Patrick (NMP/Tampere)
>>Cc: w3c-rdfcore-wg@w3.org
>>Subject: Re: Text for FAQ re rdf:datatype="&rdfs;#XMLLiteral"
>>
>>
>>>Perhaps it remains an open question whether we require RDF
>>>parsers to check the validity of lexical forms explicitly
>>>specified with rdf:datatype="&rdf;XMLLiteral". I (perhaps
>>>mistakenly) thought we didn't want to impose such a requirement
>>>on RDF parsers. Perhaps we do.
>>>
>>>Patrick
>>>
>>>
>>
>>I'm not necessarily suggesting that we *ought* to impose such a 
>>requirement, just pointing out that it seems to break the 
>>model that we 
>>describe for other datatypes. 
>>
> 
> I don't think it breaks any model. RDF is completely ignorant
> about any datatype other than rdf:XMLLiteral. Thus, it simply
> is unnable to say anything about lexical forms other than
> those for rdf:XMLLiteral.
> 
> rdf:XMLLiteral is RDF's own datatype, so it seems reasonable
> that RDF parsers would pay it special attention -- but that
> doesn't mean that the datatype itself is in any way different
> from any other datatype or is getting treatment that is
> contrary to the MT.
> 


Just to clarify:  the "model" I'm referring to is that software that 
purports to support a given datatype is responsible for being able to 
determine the validity of lexical forms for the datatype and, given a 
lexical form, determine the proper value according to that datatype (at 
least, that's what I understand the "model" to be).  What I'm suggesting 
breaks the model in the case of rdf:XMLLiteral isn't that RDF parsers 
would "pay it special attention" (the "model" *requires* that it be paid 
special attention), it's that it would be *optional*.  In other words, 
if RDF software works for "its" datatype (rdf:XMLLiteral) the way I 
understand other software is supposed to work for "their" datatypes, RDF 
software ought to check the validity of lexical forms for rdf:XMLLiteral.

Conversely, would we consider it legitimate according to our model for 
RDF datatypes if software that purported to support xsd:integer got the 
typed literal "pumpkin"^^xsd:integer for John Smith's age (to quote an 
example from the Primer) and failed to complain (or even check)?


> 
>>And I think it'll be natural 
>>for people 
>>to wonder exactly how (or when) the validity of an instance of 
>>rdf:XMLLiteral gets determined, if RDF software doesn't do it.  
>>
> 
> I agree with you there, and upon considering it, I think we
> *should* in fact require (or at least very strongly encourage)
> RDF parsers to check explicitly specified lexical forms
> for rdf:XMLLiteral for validity.
> 
> 
>>If we 
>>continue our litergy that says we're not defining a 
>>processing model, I 
>>don't see how this can be dependent on whether we imagine an RDF/XML 
>>parser on top of an XML parser (for example).  Even though we define 
>>RDF/XML as the standard language for serializing RDF graphs, that 
>>language is defined in terms of how the RDF/XML generates 
>>triples.  
>>
> 
> Well, I'd view validity checking of rdf:XMLLiteral lexical forms
> to be a syntax check, though I could conceive of many valid
> arguments that such checking would not be completely generic
> (but so what, what does it hurt?)


It *is* a syntax check, but it seems to me there are (at least 
potentially) two syntaxes involved.  One is the lexical form defined for 
rdf:XMLLiteral.  This is the lexical form defined at the graph level 
(whether the original source syntax that produced the graph was RDF/XML 
or not).  The second is what you have to write in RDF/XML (since it's an 
XML markup language) to carry that lexical form through the XML to the 
graph.


> 
> 
>>It's 
>>at the level of triples that this validity checking would have to be 
>>done.  Presumably it's at this level that whether subjects and 
>>predicates are valid URIrefs is determined, and whether any literals 
>>specified are valid literals.
>>
> 
> Fine, so an RDF parser SHOULD (not necessarily MUST) check the
> validity of all lexical forms for rdf:XMLLiteral, however they
> were defined, and SHOULD issue a warning for any illformed
> lexical forms.
> 
> We can treat it as an optional, but recommended, step following
> the purely generic, datatype agnostic parsing phase.
> 
> 
>>I'd note that an alternative would be to treat rdf:XMLLiteral sort of 
>>like we treat other datatypes, except that we (RDF) have defined a 
>>built-in name for it.  That is, we don't define XML (any more than we 
>>define xsd:integer) but we know people are presumably going 
>>to want to 
>>use XML literals, so we provide this datatype, which is likely to be 
>>widely available (but if it is made available in a particular RDF 
>>implementation, the implementation has to provide the appropriate 
>>software to support the RDF-defined datatype model, and the datatype 
>>semantics).  I don't really care, except if we're going to break the 
>>model, I think we need to say so.
>>
> 
> I don't think any special attention given to rdf:XMLLiteral by
> RDF parsers breaks any model, but I agree that we should be clear
> about the what's and why's.
> 


I repeat, it's not special attention given to rdf:XMLLiteral by RDF 
parsers that seems to break the model, it's that this attention would be 
optional.

--Frank


 


-- 
Frank Manola                   The MITRE Corporation
202 Burlington Road, MS A345   Bedford, MA 01730-1420
mailto:fmanola@mitre.org       voice: 781-271-8147   FAX: 781-271-875

Received on Tuesday, 12 August 2003 08:59:40 UTC