Re: deprecated tags in Wilbur & Cougar -Reply -Reply

Paul Prescod (papresco@calum.csclub.uwaterloo.ca)
Wed, 07 Aug 1996 20:22:35 -0400


Message-Id: <1.5.4.32.19960808002235.00856530@csclub.uwaterloo.ca>
Date: Wed, 07 Aug 1996 20:22:35 -0400
To: marc@ckm.ucsf.edu, www-html@w3.org, www-style@w3.org
From: Paul Prescod <papresco@calum.csclub.uwaterloo.ca>
Subject: Re: deprecated tags in Wilbur & Cougar -Reply -Reply

At 02:02 PM 8/7/96 -0700, Marc Salomon wrote:
>Mike Wexler <mwexler@mv.us.adobe.com> wrote:
>I agree that there is a need to allow for a way for the "semantic classing" of
>generic elements that is independent from style classing:
>
><DIV CLASS="sidebar" ROLE="section">
>
>This construct uses the proposed CLASS attribute to invoke the CSS scheme for
>rendering according to class "sidebar" on this structural element that is
> semantically equivalent (ROLE=) to any other section.

I don't see why these roles (if you'll excuse the pun) must be separated. If
you make CLASS just a pointer into stylesheets then you are just turning
HTML into a once dereferenced presentation language. Objects of the same
semantic classes are supposed to look the same. Objects that look different
should have different semantic classes.

Where you need to make some objects look different from other objects
without semantically classifying them differently (a dubious practice), then
you should use the STYLE attribute. 

I can see that you might need a finer subdivision than whatever standard you
are adhering to allows. In that case, you provide both:

<DIV CLASS="section diversion"> The "section" would be the standardized
semantic name and "diversion" is your idiosyncratic subclass. (in some
mediums that might be a sidebar, in others it might be a different tone of
voice ...)

>|If we need a much larger set of standard elements, than lets create them.
>|Maybe the solutions is to get the UAs to start handling SGML documents with
>|DTDs specified by a URL. Than any organization could create a DTD that matched
>|the needs of its members. Some DTDs might be shared by a large number of
>|organiations. Others might be specific to a certain author.
>
>Sure would be nice.

I think we all agree. Browser writers, are you listening?

Using traditional SGML DTDs is a big step forward. It allows us to
decentralize the development of the tag soup, so that we can better encode
our documents. But it is still tag soup in the sense that the standard is
weakend. You can't very well validate to anyone else's DTD anymore. SGML
architectural forms offer an even more advanced way of solving the same
problem. Not only can we all make our own tags and our own DTDs, we can
implicitly build documents that conform to other dtds, and VALIDATE them
according to those other DTDs.

 Paul Prescod