Re: Identifying D-links with "rel"

 
> > IMHO, the Dardailler-Ragget correspondence did not address specifically 
> > the notion of "reserved class name" to be used by user agents. 
> 
> What do you mean exactly ?
> To me, the reservation comes with the added wai- prefix.

No it doesn't. Take an author who studied the HTML4.0 spec. 
and invented rightfully class names of his best imagination
some of which have the wai- prefix. The spec. never said you shouldn't 
use this prefix as "it is resevered for future use" or something of 
that sort.

If you want to to have classes that _user agents_ understand without 
style/script instructions from the author, you have to change 
the HTML spec. and define a different syntax (e.g. reserved names 
start with a _ ).  
This is necessary for backward compatibility with HTML4.0. 
It should be assumed that authors may have used all valid class 
names with their own invented meaning.

The fact that it is unlikely that people chose wai- prefix to there
classes is just a guess. 

WAI guidelines must be backward compatible with HTML4.0.

The rel attribute is a mechanism that is designed for predetrmined names.
The spec. doesn't say what they are but makes it clear that rel is useful 
only if its value is _commonly_ understood by user agents. 
  
> > Actually, Ragget suggested that he wouldn't stick to an identical 
> > interpretation of rel for LINK and A. Taking Ragget (and not the spec.)
> > as the expert on the meaning of rel, it seems that rel is better since
> > "conventions" on the meaning of its value are recognized (also 
> > by the spec.)
> 
> I'd still maintain that CSS1 support i.e. begin able to do 
>  
>  .wai-dlink { display: none }
> 
> (which you can only do is class=wai-dlink is used)
> 
> is more important than more focus semantics of rel vs. class.


This is OK if you want to give an _authoring_ tip, assuming
_user agents_  aren't supposed to do anything special with 
"hard coded" D-links.

If you want to give _user agent_ requirements, rel is better. User agents 
can show and hide D-links if identified in a predetrmined way without 
supporting stylesheets at all.

If you want advice for both authors and user agents then use rel.
Authors who want to be compatible with CSS1 can use 
class in _addition_. 

In my view the best thing to do is to recommend to authors to use 
_only_ longdesc, and leave user agents alone. 

Writing both D-link and longdesc is in the first place motivated 
by backward compatibility with  HTML 3.2 and lower who don't 
recognize longdesc.  Authors should recognize that in HTML3.2 
it is _impossible_ to provide long descriptions of images, and should 
be happy that with HTML4.0 it is possible. If they want to do hacks
in order to get the impossible in HTML3.2 they should do these hacks 
on their own.

User agents should put effort to support alt in AREA which is a part of 
HTML3.2, and title in A which is a part of HTML2.0, and not deal with 
backward compatibilty hacks of a very small group of authors who want 
to hard code D-links. 

Regards,
Nir Dagan.
http://www.econ.upf.es/%7Edagan/

Received on Friday, 5 June 1998 06:36:34 UTC