whither LONGDESC?

The good news:

Daniel gave us a good review of how far we have come in 

  Re: Dave Raggett: Re: IMG ALT attribute in HTML 4.0 (fwd)
  http://lists.w3.org/Archives/Public/w3c-wai-wg/1997JulSep/0264.html


The bad news:

I don't think we are ready to "call the question" and try to 
pick a specific strategy as _our proposal_.  Why?

1.  We haven't looked at all the criteria.

Daniel, as I read it, evaluated options based on assuming
conforming browsers, i.e. the assumed browser is up-to-date with
HTML 4 and CSS 2.

I personally think we should also consider (not to the exclusion
of the new-browser evaluation) how different strategies play with
existing browsers.  That evaluation hasn't been brought up to the
level that Daniel gave us for the new-browser case.

I also don't think that we should try to decide by ourselves how
much weight to give to old-browser evaluation vs new-browser
evaulation as we prepare a position to take into discussions with
HTML and CSS communities.  This is the sort of value-setting
where we should try to get the full community to express itself
through discussion on the IG list.

2.  We haven't looked at all the solution possibilities.

Particularly if we care about how our solution plays with existing
browsers, I think that we want to find HTML techniques which will
insert text so that it will

	- display all the time in old browsers
	- display conditionally in new browsers

One downside of the LONGDESC proposal as I understand it, is that
it will

	- display never in old browsers
	- display conditionally in new browsers
	
For this reason I have been thinking about constructs that would
show, as opposed to hide, the contents in legacy browsers.  The
examples I have been thinking about resemble
conditional-compilation constructs.

The language techniques are 

	a new container for optional material with control attributes
		or
	a predefined class of DIV with the same effect.
		or
	addition of control attributes to many/all block-level containers
		and/or
	a new contentless element GUARD containing control attributes and
		a TARGET=id attribute designating the element to which
		the precondition has been added.


	control attributes such as:
	  IF=bool
	  IFNOT=bool
	  IFSHOWN=id		-- for "id" read "id | name" throughout
	  IFNOTSHOWN=id
	  ALTFOR=id
	etc.

The semantics of ALTFOR is like IFNOTSHOWN except in addition it
says that if the element whose ID is cited as the value of ALTFOR
is not shown, then the content of this element assumes the layout
assignment otherwise used for that element.

If we had these language resources, we could write something along the
lines of a 
	
	- figure legend with link to textual description
	- guide to the linked resources in the case the image is a 
		sensitive map

and have it disappear or be used under ALT conditions in new, smarter
browsers but appear (even at the foot of the page) in legacy browsers
with the links to the cited resources intact.

This is just my fevered brainstorming.  But I wanted y'all to
consider if there aren't better options that could be
hypothesized, from the old-browser evaluation perspective.

Note that by having the element with the text refer to
the non-textual element alternatized, and not vice versa, this
technique can be applied to all manner of bad actors that can bear
an ID.

--
Al

Received on Friday, 19 September 1997 15:24:38 UTC