Re: [Role Module] Namespace for 'role' Attribute

Jim Jewett wrote:
> On 10/30/07, David Woolley <forums@david-woolley.me.uk> wrote:
> 
>> Given that a namespace can be a single letter and a colon

As noted, I read this particularly in the context of SVG.

> 
> And declaring the namespace in the first place.

That's boiler plate.

> 
> You may believe the cost is trivial; the html5 editors seem to
> disagree.  They have said that browser implementors consider

I think they are working according to the following principles:

- source compatibility with IE is required and whilst IE has some
   special case support of namespaces, it doesn't, for example, give
   proper style sheet access to them.  Consequently they don't want to
   use namespaces in the HTML5 serialisation.

- source compatibility with HTML5 in XHTML syntax is required, so,
   they don't want anything that can also be used in text/html mode to
   use namespaces in application/xhtml+xml mode;

- they are dealing with authors who unsophisticated with respect to
   formal languages.

The IE thing is worrying, because it means that IE can poison the use of 
almost any true XML feature.  Adobe SVG Viewer also has a poisoning 
effect, because it doesn't handle namespaces properly - an example of 
how vendors can frustrate standards when something is easier for them.

As against the reasons for HTML5:

- languages like SVG are XML namespaced languages already and, except
   for the need to use fixed namespace names for ASV, there is no problem
   of maintaining backwards compatibility;

- anyone using role manually, is either relatively sophisticated,
   because they are doing something for the benefit of a minority
   (in a whole that is targetted at the normal web site market),
   or are working in a disciplined environment, and including it because
   the local standards require it;

- most non-HTML XML languages also require reasonable understanding of
   formal languages.

> 
> So again, someone believes there is a cost.  What is the benefit of
> the namespace requirement?

The benefit is that it gives a message that the attribute is not the 
private property of the particular language.  This is not foolproof, but 
I think it will dissuade a few developers from taking liberties.

(In a sort of way, the ASV namespace problem is a counter example, in 
that they didn't respect namespaces enough to implement them properly in 
a language for which they were an important part of the original concept.)

-- 
David Woolley
Emails are not formal business letters, whatever businesses may want.
RFC1855 says there should be an address here, but, in a world of spam,
that is no longer good advice, as archive address hiding may not work.

Received on Tuesday, 30 October 2007 22:37:07 UTC