Re: ACTION-267 DONE

On 22 Jan 2009, at 09:52, Ivan Herman wrote:

> HI Bijan,
>
> thanks. This indeed includes my thoughts of yesterday. Actually,  
> reading
> it again, I think that we should refer to the RDF spec(s) and not  
> to the
> XML ones. Even if the XML specs get updated, the issues you describe
> would still be around (at least formally) because, as you say, RDF
> normatively refers to Unicode 3. It may be cleaner if OWL refers to  
> RDF
> in some open ended fashion so that if RDF is updated at some point in
> the future to refer to Unicode 5 (or Unicode 25:-), OWL would be fine.

I think this is the *least* we should do. Hardcoding a Unicode  
version into OWL just ups the number of places that need updating.

> Having said that, we would have backward compatibility issues in the
> sense that we may have ontologies that _today_ are not kosher but  
> would
> become so in future. But I guess that should be fine, it fits our open
> world view...
>
> Sigh:-)

Here's another possibility: Leave it to be tied to the latest Unicode  
but point out that serializations and parsers (and apis) would need  
to be updated. That way, we do not discriminate against people who  
need the new characters. In the conformance document, we could  
discuss this and even suggest (require?) that *implementations*  
indicate which version of Unicode they support. This would add some  
implicit pressure to keep up to date wrt Unicode.

Indeed, I rather suspect that RDF should update, esp. given the  
(widely loathed) XML revision 5. Certainly the *model* should track  
Unicode latest....perhaps this could be considered an errata? Given  
the right conformance description, such a change would have no impact  
on implementations.

(Actually, there's an interesting question about RDF now. RDF/XML  
refers normative to XML revision 2! Wacky. That means that RDF  
parsers should reject XML documents that use revision 3-5 features?  
Are there any 3-4 features which make a difference?)

BTW, I'm in favor of making rdf:text/xsd:string whatever have finite  
alphabets. Discussions with Birte about her datatype implementation  
seem strongly in favor (mostly in the reuse of existing automata  
libraries). There's a number of ways to handle this including  
parameterizing text/string (i.e., with unicode version).

Cheers,
Bijan.

Received on Thursday, 22 January 2009 15:56:44 UTC