Re: [xml-dev] xlink 1.1

[David Carlisle <davidc@nag.co.uk>]

On 01/03/2010 23:08, Simon St.Laurent wrote:
>
> It's still not clear to me that the IRI support does anything but
> explicitly require something that was possible before.


I think the IRI update is just a general cleanup as specs are updated to 
use that terminology rather than each spec effectively defining the same 
thing as an extended URI syntax allowing strings that need to be 
%-encoded to make a valid URL.

The real point of xlink 1.1, dropping the mandated xlink:type attribute, 
  is pretty useful (if you are using xlink at all) If you are only 
interested in simple links (which is most people, most of the time) 
xlink 1.0 mandates that every link element has xlink:type="simple".

For something like mathml2 where any element can be used as a link, this 
is very costly. Originally the idea was that a DTD could default the 
xlink:type attribute, but that only works if the dtd is read and worse, 
if the dtd is read then it defaults the element on every element that 
could _potentially_ be a link, which in a markup heavy language like 
mathml has a non trivial cost in size of the DOM. So MathML2 changed its 
definition so that type attribute isn't defaulted and the end user had 
to add xlink:type="simple2 whenever they use xlink:href=.

Switching to Xlink 1.1 would drop this requirement to use xlink:type.

However MathML3 is suggesting that an unprefixed mathml-specific href 
attribute is used instead.

David