Re: rel=CURIE in RDFa, but rel=URI in Link:

> If HTML5 later on decides to allow URIs in link/@rel (and I think  
> I've seen several
> proposals that go into that direction), then we have another conflict
> between the RFDa community and the WHATWG crowd.

HTML5, as currently drafted, already allows URIs in link/@rel to the  
same extent @xmlns takes a URI. That is, using a URI as a rel token is  
allowed (if the extension designer registers it on the wiki), but the  
token is compared as a string--not as a URI.

> This one could be easily avoided by requiring a safe CURIE.

Or by dropping CURIEs and using fully spelled out absolute URIs instead.

Doing this would avoid the problems related to relying on the  
namespace mapping scope from a different layer. The TAG has already  
found qnames-in-content problematic but said that "It is simply not  
practical to suggest that this usage should be forbidden on  
architectural grounds." However, not being able to forbid pre-existing  
usage doesn't mean that it is a good idea to introduce new usage  
(CURIEs) with the same problems.

Henri Sivonen

Received on Friday, 12 September 2008 08:37:29 UTC