Re: Possible solution to XMLLiteral issue (was Re: [RDFa] rdf:XMLLiteral (was RE: Missing issue on the list: identification of RDFa content))

Elias Torres wrote:
> Mark Birbeck wrote:
>> Hi Dan,
>>
>> anything, can you comment on whether the problem you foresee would be
>> addressed by modifying the parsing model so that either a typed
>> literal or a plain literal is created, depending on whether the
>> element being parsed has element children or not?
>>
> 
> I think we almost missed this in the thread and I wanted to make sure we
> give it proper attention. On our RDFa TF call, we discussed with Mark
> the XMLLiteral issue and (hopefully I'm not misquoting him) he thinks
> that we do have a problem with our current default to XMLLiteral because
> we lose our language information. This in fact is a consequence for the
> lack of ability in RDF concepts to add language to datatyped literals,
> but, nonetheless, we have to deal with it.
> 
> Mark proposes something that was thought sometime ago by him and Steven
> that in order for us not to lose mark-up, we change the spec so that
> parsers can make the decision based on the element's children whether
> it's a plain literal or an XMLLiteral.
> 
> I like the scenario because it's a double-win. Neither plain literals or
> XMLLiterals have to add a datatype property. It's stated by the author
> simply by the fact whether she used mark-up or not in the element's content.
> 
> I also brought up the point about plain ol' HTML. HTML mark-up content
> is not XMLLiteral so I think that's this conditional rule could be of
> benefit in allowing documents to switch formats and later on adding a
> rule that in HTML there be some sort of HTMLLiteral.
> 
> What does everyone think? Any major issues with the approach? We don't
> want to decide on the issue until we have input from the community, so
> please let us know.

I have to say I'm skeptical, if the design is what I think it is.

A single coherent off-Web dataset might contain names with markup in, 
eg. foaf:name of "Dan <foo>br</foo>rickley", or without.

If published in RDFa, ...and the above rules are followed, some but not 
all of the literals become plain, others become datatyped. Amongst other 
things, the lang-tag handling of these two idioms is radically 
different. I'm not sure it's healthy to fragment data this way, based on 
its incidental internal characteristics.

Sorry, I have to run for a plane, this isn't a very well articulated 
concern, I'll try to come up with something more precise.

Dan


> -Elias
> 

Received on Monday, 19 March 2007 15:47:48 UTC