Re: RDFa and Web Directions North 2009

On Feb 16, 2009, at 13:50, Mark Birbeck wrote:

>> Why was RDFa specified to use xmlns:foo when it was obvious that  
>> people
>> would want to deploy it as text/html and you should have known that  
>> HTML
>> parsers handled xmlns:foo differently from XML parsers?
>
> Ah...now that is a fair question. :)
>
> I am no great fan of XML namespaces either. In my view they were
> created for a world of interlocking/interchangable documents that
> simply never happened.

Here we have agreement.

> However, given that many W3C standards rely on XML namespaces, we felt
> that we would face problems if we tried to progress RDFa using a
> prefix/mapping mechanism that did *not* use XML namespaces.

It seems bad if syntax with technical issues is chosen for political  
reasons.

This reminds me of the ARIA naming debate. Towards the end of the  
debate, it felt like proponents of the colon were more interested in  
keeping up the syntactic appearance of Namespaces than actually  
protecting a Namespace-wise architecturally sound programming model.

> So we took a pragmatic decision; we realised that at some point we
> would use a simple attribute to specify token mappings, but for the
> first pass through of RDFa, we thought we should provide a less
> 'controversial' mapping mechanism.

Evidently, the mechanism you chose isn't exactly uncontroversial.

Note that one option would have been not having a mapping mechanism at  
all: Using full URIs. (After all, if full URIs as identifiers aren't a  
problem, why have a mapping mechanism?)

> We did however, make some important steps away from XML namespaces,
> though. Of course, we didn't get a smooth ride, but we started by
> making the move away from QNames. We devised CURIEs, which are strings
> with a simple substitution mechanism, and we said that XML namespaces
> are simply *one* possible source of tokens for doing that
> substitution.

I think having many sources of prefixes introduced over time is not an  
improvement.

> By saying that we are not using XML namespaces, but that @xmlns is one
> way to provide prefix mappings

A problem with using @xmlns as opposed to using Namespaces it is that  
strictly correct XML+Namespaces APIs down to expose @xmlns as an  
attribute at all.

> We strongly believe that once this mechanism is available, people will
> not bother using @xmlns.

Who is "we" in this case?

> Although it took time, and had to be done in a couple of steps, we  
> will have then completed the transition away from XML namespaces.

That sounds promising. :-)

-- 
Henri Sivonen
hsivonen@iki.fi
http://hsivonen.iki.fi/

Received on Monday, 16 February 2009 13:56:46 UTC