[whatwg] Ghosts from the past and the semantic Web

Henri Sivonen hsivonen at iki.fi
Thu Aug 28 06:06:25 PDT 2008

On Aug 28, 2008, at 15:00, Russell Leggett wrote:

> I actually think that using custom microformat-like conventions with  
> classes or tags is really not as robust a solution as what is being  
> attempted with RDFa (I honestly did not know much about RDFa before  
> this conversation). However, while people keep suggesting classes, I  
> have yet to hear anyone suggest the data- attributes. Maybe it was  
> said or implied elsewhere, but it seems like a good fit here.  
> Instead of requiring the addition of "about" or "property"  
> attributes, just use "data-about" or "data-property". It may not be  
> ideal, but it fits with the existing spec.

As Anne and Julian have pointed out, that's not a use of data-*  
attributes permitted by the spec.

> Beyond that, you have the issue of CURIEs. I can see how they make a  
> good fit, but it really is just piggybacking on  something else  
> convenient. It's an abuse of namespace syntax. That works fine for  
> XHTML, but there is no way you are getting namespaces put into HTML,  
> so figure out another way. Why not something like "data-curie="dc:http://purl.org/dc/elements/1.1/ 
> "?

That would be an abuse of data-* attributes as well. The data-*  
attributes are for scripts included by the page itself. The data-*  
attributes aren't for communication with other parties.

Having something-other-than-data-curie="dc:http://purl.org/dc/elements/1.1/ 
" or somesuch would work around the layering problem of that qnames-in- 
content have (and CURIEs have, too, when using the namespace mapping  
context). It leaves the problem that making URLs shorter (in the  
amortized sense) by introducing a supposedly insignificant prefix  
confuses people and makes stuff brittle under copying and pasting.

Why not property="http://the.entire.full/uri/here/if/you/really/want/uris/as#identifiers 

Henri Sivonen
hsivonen at iki.fi

More information about the whatwg mailing list