Re: 'role' should be property

2007/5/31, Mark Birbeck:
> On 31/05/07, Matthew Raymond <> wrote:
> >
> > Bruce Boughton wrote:
> > > Dmitry Turin wrote:
> > >> This demand is anachronism,
> > >> first, since _behaviour, binding, pronouncing_ are already specified in it;
> > >> second, since sense has appeared to abolish it.
> > >
> > > Could you give examples to support your assertion that behaviour and
> > > binding are described by CSS?
> >
> >    Only example I can think of is the BeCSS[1], which was introduced for
> > languages like XBL 2.0 and Microsoft's HTML Controls (HTCs).
> You say "only example", as if it's inconsequential. The point being
> made is that the mechanisms associated with CSS--what I've called the
> dynamic infoset--have already been used for non-presentational
> features. One of these features is 'behaviour'.

Yes, but adding behaviors shouldn't (theoretically, if used correctly)
change the semantics of the markup. The document should still be
usable without the bindings.

If you put semantic data (because I think "roles" belong to the
semantic facet of documents, the one which allows accessibility for
everyone) in a stylesheet (or "rolesheet"), the document could
potentially not have the same "meaning" depending on whether the sheet
is loaded (supported by the UA) or not. Even worse, you could change
the semantics of a document by just changing the sheet being used!

For example (not a really good one re. the proposed role= attribute;
just to say that behaviors add value to what's already there and
doesn't "redefine" everything), the new <menu> and <output> elements
and the new <input> type= values can quite easily be implemented with
behaviors (in browsers which support them but do not support HTML5,
such as Internet Explorer and Firefox). This doesn't change the
meaning of the document, which can be processed by behavior-unaware
UAs (such as, I suppose, most AT).

Thomas Broyer

Received on Thursday, 31 May 2007 12:50:38 UTC