Re: Generic Markup [was:Re: deprecated tags in Wilbur & Cougar]

Gavin Nicol (gtn@ebt.com)
Sat, 10 Aug 1996 00:53:37 GMT


From: Gavin Nicol <gtn@ebt.com>
Date: Sat, 10 Aug 1996 00:53:37 GMT
Message-Id: <199608100053.AAA16377@wiley.EBT.COM>
To: marc@ckm.ucsf.edu
CC: www-style@w3.org, www-html@w3.org
In-reply-to: <9608091635.ZM7630@gaia.ckm.ucsf.edu> (marc@ckm.ucsf.edu)
Subject: Re: Generic Markup [was:Re: deprecated tags in Wilbur & Cougar]

>But not if those GI's don't yet exist.  There seems more intertia towards
>standardizing on CLASSES than on a set of GI's.

You don't get it do you? Neither a standard set of CLASSES nor a
standard set of GI's will ever work for all cases. HTML is living
proof of it (though it's doing pretty well overall). Why do you think
SGML was invented in the first place?
 
>Look, I don't see using attributes to express structure as anything but a
>bridge technology towards a "real" SGML future.  I want to encourage/enforce
>the separation of structure from presentation as much as is feasible.  But
>until there are mature proposals for using GI's to interoperably express
>structure on the WWW, I am concerned to see CLASS, as specified in
>Cougar, used or at least intrepreted for both, as harmful towards
>continuing that separation.

I don't think that separation is *always* important, but I think
representation *is*. If people can support CLASS in a meaningful way
(especially indexing engines) then they can be made to do the same for
GI's. 
 
>|In both cases, I cannot see why anything that can actually *use* the
>|structure shown by the CLASS attribute for anything beyond style
>|selection, cannot also use a form that uses GI's.
> 
>Because the choice has been made, for the sake of mass consumption
>generalization,  to keep HTML simple.  I see a clear path towards
>your more ideal but not yet mature solution in standardizing on a
>small set of structural attributes aside from presentation classes
>and then fleshing out the problems involved in establishing
>equivalencies between arbitrary tagsets. 

I haven;t really seen a choice made yet, and when I talk about "GI's"
I'm not really talking about HTML anymore. HTML is one representation
structure, I want the freedom to choose my own. If cougr really get's
going, I can imagine a lot of people who dynamically generate content
using something close to SPAM instead of tags. It'd be a lot less work
than trying to do a meaningful conversion.
 
>But the real conflict comes in that browsers are supposed to be
>liberal in what they accept, rendering any tagset that isn't
>catestrophically broken.  But  the value of producing conforming HTML
>documents is common wisdom not to be questioned.

Really. I question the wisdom of any supposedly almighty tag set. The
idea of creating valid HTML is fine, except that very few do it. One
reason why people do not do it is because they try to use HTML to
accomplish things that are better done in a different
representation. The ither reason is because vendors have an interest
in locking people into proprietary formats (extended HTML tags).

>Do you claim that validation is only needed to know about omittag?

Certainly not. I claimed that you only need a DTD for omittag in
*non-validating* systems. Validating data has many benefits,
especially for enterprise situations. Even then though, you'll find
that if you introduce a standard DTD, or even two or three of them,
people have a tendency to abuse them, because the DTD model may not
model the data in the way a user wants. DTD's *need* to be dynamic.

>How are we to resolve this dissonance in philosophies (or at least
>common wisdoms)?  And are is it practical at this time to propose a
>global registry (yet another registry) of GI equivalency mapping
>tables? 

I think that we must accept that the representational structures need
to be infinitely expandable. There are 2 ways to do this:

  1) Via GI's, which offer some simplicity and whatnot to indexing
     etc.
  2) Via attributes indicating structuring.

From my experience, *both* are actually necessary in most cases. If
you have a standardized DTD in a company, you really need an escape
route for people to use (attributes), and go back later an modify the
SGML, or the DTD, depending on frequency analysis etc.

In the case of the WWW, we have such a huge number of people, and such
a huge variety of uses, that a single DTD can *never* be enough, and
can never even be *close* to enoguh to represent data in a meaningfull
manner (ie. you could do presentation well, but you could never
capture the structural semantics with a single DTD). 

BTW. I don't think I have ever asked for *full* SGML to be supported
on the WWW. I think something like the TEI subset is enough.