Re: Meeting with SVG, XHTML, WAI people to move forward on ARIA as a cross cutting technology

On Oct 19, 2007, at 12:46, T.V Raman wrote:

> I dont believe we need go down the ugliness of inventing a second  
> namespacing mechanism because some dont like the current one;

I agree we shouldn't invent a second namespacing mechanism. I think  
we should do ARIA without a technical namespacing mechanism at all  
and only use a naming convention for spec organization without the  
parsing and DOM layers knowing anything about it.

It's not about not liking the current XML namespacing mechanism. It's  
about the current XML one being unsupported in already shipped text/ 
html implementations. It follows that introducing it now to text/html  
would create a scripting discrepancy between legacy and new browsers.  
We don't want that, so we are stuck with no Namespaces in XML-style  
namespaces in HTML. We want DOM and scripting consistency between  
HTML and XHTML, so the legacy leaks to XHTML as well. We want  
browsers to be able to use the same DOM reading code for SVG, too, so  
the legacy leaks to SVG as well.

The legacy is very powerful and I don't think there's a payoff from  
fighting legacy that would outweigh the cost of discrepancies with  
the legacy.

> --- the current namespacing mechanism wont go away, and the world  
> will be left to contend with twice the ugliness that pervades at  
> present

The Namespaces in XML processing layer can't be taken out from  
between the XML 1.0 layer and the DOM layer. It is too late for that.  
However, we can let the Namespaces in XML layer remain latent and  
avoid using it for new attributes.

No actual technical doom or gloom in Web UAs will arise from putting  
the ARIA attributes in the per-element no-namespace partition of  
elements on the XML side. We know that XHTML and SVG don't have pre- 
existing attributes that start with "aria-", so there won't be actual  
collisions as far as Web languages are concerned. In theory, a non- 
Web language out there could collide with this scheme if you tried to  
use the "aria-" naming scheme on a theoretical non-Web language.  
However, since we are working on Web specs, I think we should  
optimized for actual Web language usage instead of theoretical non- 
Web language usage.

(This is exactly where xml:id went wrong. The practical approach  
would have been declaring that the attribute id is always and ID and  
if someone out there had a non-Web problem with that, it would have  
been a non-Web problem. Now the fallout from xml:id is seen in Web  
specs and implementations.)

Henri Sivonen

Received on Friday, 19 October 2007 10:42:10 UTC