Re: SVG <title> (was: SVG Feedback on HTML5 SVG Proposal)

On Wed, 11 Mar 2009 13:38:51 +0100, Robin Berjon <robin@berjon.com> wrote:

> On Mar 11, 2009, at 02:04 , Ian Hickson wrote:
>> On Tue, 10 Mar 2009, Doug Schepers wrote:
>>>
>>> The SVG equivalent of <span lang=""> is <tspan xml:lang="">.  We
>>> considered making the content model of the <title> and <desc> elements
>>> match that of the <svg:text> element, but also wish to allow X/HTML
>>> content for document semantics like lists and such.  Up until this
>>> point, the SVG+X/HTML story was unclear, but with browsers natively
>>> implementing SVG, we now have an opportunity to sort this out.  (Do  
>>> note
>>> that there are SVG-only UAs, so any solution there would have to only
>>> optionally use HTML.)  Any thoughts or comments along those lines?
>>
>> One option would be to have SVG say what it does now, and to have the
>> HTML5 spec explicitly say that the content model of <title> in SVG in
>> text/html is limited to what HTML5 calls "phrasing content". This
>> basically excludes what HTML4 calls "block-level elements", and includes
>> things like <span> and <ruby>.
>>
>> I don't really have an opinion on exactly what the right solution here  
>> is.
>
> I think that's the right approach. Basically, the limitations that Tiny  
> 1.2 has in making it text only are (as you point out) bad for I18N and  
> in effect entail that there's no need to use an element as an attribute  
> would suffice. Since a) there is no specified SVG rendering for this  
> element, and b) the cases in which it can be involved with the rest of  
> the (notably with <tref>) are well defined, using phrasing content seems  
> sensible.

But...

SVG to date only allows text in title.

XHTML 1.x and XHTML5 only allow text in title.

text/html HTML does not and cannot allow elements in title.

If SVG <title> in text/html does not use RCDATA parsing then it's pointless to make SVG <script> and <style> use CDATA parsing.


Performance with speculative parsing (which needs more research) and consistency in syntax and content models should be given consideration here, too, IMHO.

-- 
Simon Pieters
Opera Software

Received on Thursday, 2 April 2009 15:43:20 UTC