Re: OWL XML Literal constraining facets

Responding only for myself.

Hi Paul,

XML Literals are a bit controversial (not least because of the  
canonicalization requirements which means, for example, that XSLT  
cannot be reliably round tripped through XML Literals...basically, you  
lose namespace declarations if the prefix is only used in content).  
This is without reference to OWL, by the by: For a long time,  
triplestores and apis didn't do such a great job with XMLLiterals and  
there's even some issues, if I understand it correctly, with RDFa and  
XMLLiterals.

Given the poverty of implementation and use, plus some warts in the  
current datatype, plus the overhaul of the datatype system,  
plus...well, you get the picture. I have to say that I didn't find  
your obvious picks all that obvious (not that they aren't, perhaps,  
sensible, but I sure didn't think of them), which suggests that it  
would have been quite difficult to get consensus on these.

(Plus, there's some sort of reasonable concern about how to even  
conceptualize such structured types inside an ontology language from a  
modeling perspective. People seem willing to lift XML data into RDF  
and OWL rather than comment on that data. That maybe due to a poverty  
of imagination or tool access. But again, it makes it unobvious.)

For me, the obvious thing to do is to try to integrate XML Schema  
complex types (perhaps suitably restricted). Assuming we're a bit  
careful (e.g., with keyrefs) this should be a straightforward  
extension from the semantic and (conceptually) the implementation pov  
(i.e., it should preserve the datatype oracle modularity). The issue,  
of course, would be how easy it is to do a sat checker for  
conjunctions of such types (I didn't see any obvious existing  
libraries). Another issue would be how used that would be in practice.  
(It's not like complex types are so very widely used.)

My personal advice is to start developing a specification for the  
extended datatype you desire. (That's certainly my plan.) The datatype  
system is extensible and there's at least one example of an extension.  
For engaging the interest of OWL implementors OWLED or ISWC are  
reasonable venues, as are the various lists.

Hope this helps.

Cheers,
Bijan.

Received on Saturday, 2 January 2010 12:50:28 UTC