- From: Ian Jacobs <ij@w3.org>
- Date: Thu, 26 Aug 1999 17:46:17 -0400
- To: w3c-wai-gl@w3.org, w3c-wai-ua@w3.org
UA/GL teleconference to coordinate issues
26 August 1999
Present:
Gregg Vanderheiden
Chuck Letourneau
Wendy Chisholm
Ian Jacobs
Al Gilman
Jon Gunderson
Charles McCathieNevile
Gregory Rosmaita
Jason White
Marja Koivunen
Dean Denman (The Lighthouse, joined after 1 hour).
Next scheduled meeting of GL WG: 9 September.
Note 1: We will try to schedule a meeting for 2 September.
If we can schedule this meeting, we will invite
UA to join the GL call to discuss outstanding
issues on 9 September.
Note 2: Meetings will henceforth last 90 minutes.
Agenda [1]
[1] http://www.w3.org/WAI/GL/meetings/19990826.html#agenda
----------------------
1) Review of action items
2) Navigation bars
3) Long description media types
4) Specificity of checkpoint 1.5
5) Metadata [Not covered]
----------------------
1) Review of action items.
(Previous minutes)
http://lists.w3.org/Archives/Public/w3c-wai-gl/1999JulSep/0266.html
CL: Talked to Coodination Group about a standard set of
questions for Web development "instructors".
Judy said good idea, and falls within EO, but no
time to address this today.
JW: Two other points made as well:
a) EO has clarification of guidelines on their agenda
b) Concerns about "scope creep" for GL.
WC: Rob sent a note to GL with questions.
2) Navigation bars
JW: Al's suggestion [1] regarding a mode in which DIV/title
can be regarded as a container that can be skipped
is a valid suggestion. I do have reservations about
the use of MAP to contain navigational links. Don't
want to use this unless in a spec. Could propose to
HTML WG, but don't want to endorse without making
it part of the specification.
Furthermore, future technologies (schemas) will allow
us to do this correctly.
[1]
http://lists.w3.org/Archives/Public/w3c-wai-gl/1999JulSep/0293.html
CMN:
a) For DIV/title: It's a backwards step. It's the navigational
equivalent of using FONT everywhere to create structural
illusions.
b) For MAP: The suggestion comes directly from the HTML 4.0
specification [2]. Shows how to use MAP with straight text.
Similar situation as we were in with longdesc. We recommended
that people used d-links until longdesc was supported.
Similar with MAP: use of MAP as block-level container
is supported - it is rendered. HTML 4.01 has this.
[2] http://www.w3.org/TR/REC-html/struct/objects.html#h-13.6
c) The "skip navbar link" is a hack that is similar to
the d-link. Useful if understood by humans, but not
machine recognizable.
GV: (Scribe missed comments)
JW: About DIV - it's a container element for style sheets
and other purposes. When it does this, in an outline
reading of the document, it should appear as part of
outline.
WC: DIV with a human-readable title is a bridge to XML.
Are there other elements that need to be grouped (e.g.,
images)?
CMN: But using DIV this way takes you out of the "standard"
hierarchy. So it's like the XML approach, but it's not
XML. So use xhtml instead.
IJ:
a) For medium term: use schemas. Don't focus on HTML
but move towards xhtml + schemas.
b) For immediate: use scripts (for example).
What problem are we trying to solve?
GV:
a) We can add additional functionalities to existing
browsers (plug-ins). So don't focus only on short-term
solution.
b) What strategies apply today?
JW: I see DIV as the only short-term generic solution
for adding structure that we're looking for. I think
the long-term solutions should be schemas.
IJ (to CMN): How does DIV/title strike you in the short term?
CMN: Marginally inferior to MAP/title solution.
HTML 4.0 doesn't say explicitly that you don't
need an image. CMN notes: MAP content is rendered
by all browsers I've tested but Amaya.
IJ: HTML 4.01 language of 13.6.1 section on rendering
of block-level content of MAP is new. Since HTML 4.01
in PR now, good time to ask an AC member to add
a phrase that would be helpful to address this
issue (e.g., no image necessary, "should" render
content).
JW: Two times to render content:
a) Image map not represented in display medium.
b) No image associated with MAP element.
So I propose adding to HTML 4.01 a comment about
rendering MAP block content along the lines of
"The user agent should render MAP content, notably
when there is no associated image." (Note
use of "should".)
[Scribe note: Suppose the user were able to
configure the browser to render or not to
render block content in a MAP element. UAs would
not be able to render a document progressively
in "don't render block content" mode since that
would imply having to wait for all images (in
fact, all content) to know whether the MAP was
associated with any of them.]
CMN: What about OBJECT element? If you want to do what
Jason suggests, do this:
<OBJECT data="image.gif">
<MAP name="foo">
...block content...
</MAP>
</OBJECT>
Doing this, you get links instead of the image (if
MAP content is rendered).
GV: Are we trying to push understanding of MAP to
far? Is this just a loophole we are exploiting?
AG: HTML 4's intention was to allow authors to
specify URLs one time only.
CMN: Also to allow image maps or text links, both based
on a single element content. The spec allows this, but
doesn't say that you can do this even if you never
intended to use an image.
GV: What about case of bulleted list of links? You
wouldn't want to put that in a MAP.
CMN: Users approach these lists differently than
navbars.
JW: I'd like to separate
a) How MAP is defined
b) Technique of when it should be used for
grouping links. Author discretion enters here.
MK: There can be a lot of navbars in a document
(e.g., CNN site). Users want to be able not
just to skip, but to return to navbars.
WC: For DIV/title, if author puts information
in "title", UA can present this information to
the user.
CMN: MAP/title will allow the same thing.
JW: With appropriate wording in HTML 4.01,
I'd agree with Charles' proposal. Otherwise,
I'd opt for DIV with title.
IJ: What about scripts?
AG: Problem of Lynx, which doesn't support scripts.
CMN: Then include a "skip nav bar" in the block of
links. But this pushes the problem from one
of listening to physically tabbing through a list
of links. The "skip nav bar" approach is still
the solution here. Or use a browser with a structure
view (e.g., Amaya).
WG: So first link inside of MAP is to jump over the
others?
CMN: So this will work today (e.g., with scripts, CSS)
and tomorrow. Also, use an image map up front and
text links at the bottom.
Conclusion:
a) For today, use MAP/title. Put a link in the MAP
to skip over the others. Use CSS or scripting
to hide it. Also, put image map at the top and
text links at the bottom.
WC: Some bugs using tabindex within frames in
at least IE 5. Don't believe supported in NN 4.0
b) For tomorrow: Use MAP with semantics as written in the
spec.
JW: Ask PF WG to consider this scenario and other
structural navigation issues when considering
schemas.
Actions:
1) CMN: Raise awareness with AC Members about enhancing
accessibility in HTML 4.01.
JW: I will dissent use of MAP unless the update
is made to HTML 4.01.
2) WC: Propose techniques based on this discussion.
JG: Does this discussion represent consensus between UA and GL?
WC: No. I will write up the GL side of this discussion, send
to GL for comments, then send to UA.
CMN: UAs should expect to find MAP if they expect to
skip blocks of links.
MK: What about getting to the central content of the page?
AG: You may find repeated MAP elements, but you won't, from
this proposal, know how to find the primary content.
There are undoubtedly refinements to be made.
3) Long description media types
See Al's proposal. [3]
[3]
http://lists.w3.org/Archives/Public/w3c-wai-gl/1999JulSep/0297.html
GV: longdesc was designed to lead to something really
accessible. Other than "why limit it if we
don't have to", what's the benefit for making
this more than text?
AG: We're trying not to make it different than a standard
URI reference, which doesn't limit to a specific
media type.
IJ: But no "type" attribute with longdesc.
GV: But putting audio at the end of longdesc
raises barriers.
AG: But to meet WCAG, you'd need an auditory
description to accompany the audio clip.
JW: I don't think it's approprite to limit content
types for longdesc in an HTML spec. I think
the WCAG should have jurisdiction.
MK: "longdesc" also used in SMIL.
GV: I'd like to register nervousness on this point.
I fear people will put content out their that
will impose too many layers before getting to
accessible content.
IJ: Drop longdesc from SMIL and use schemas.
CMN: Drop IMG entirely.
/* Chuck leaves the call */
JW: In response to Greg's concern, I'd be willing
to have non-normative language in an HTML
spec that longdesc generally points to text.
GV: Use the Techniques Doc to address the longdesc
issue.
CMN: Having text is nowhere near as good as having
full markup: Put HTML at the end of a longdesc.
GV: Why use a movie at the end of a longdesc?
CMN: E.g., instructional videos.
GV: Why not put it on the main page?
CMN: No particular way to do it. The primary version
may be images plus text. The decision of
what is main and what is secondary is up to the
author.
IJ: This sounds like a cognitive technique: put text
as primary content, video as complement.
JW: Longdesc should not designate plain text alone,
but markup.
CMN: Yes, at a minimum, refer to markup.
JW: But this should not be normative in the HTML spec.
GV: We need to talk about the purpose of longdesc:
should limit it to a description of the image, not
supplemental information about the context in which
the image occurs. Similar to using alt text just
for the purposes of a tool tip. Warning: if the
function goes beyond a description of the image,
this could lead to problems.
AG: What about content negotation? You get text
or video depending on preferences. Refer also
to CC/PP.
IJ: Distinguish
a) Function of longdesc
b) Content types.
GV: But users may not know how to adjust clients
to get different types.
WG: How used in SMIL?
MK: Every media object has it (audio, video, textstream, ...)
CMN: Working from the HTML 4.0 spec [4]
[4] http://www.w3.org/TR/REC-html/struct/objects.html#edef-IMG
WC: Conclusion - we need to get consensus on the list about what's
the intended use of "longdesc".
4) Specificity of checkpoint 1.5
1.5 Until user agents render text equivalents for
client-side image map links, provide redundant
text links for each active region of a client-side
image map.
Action IJ: Send proposal to GL list about broadening
language to refer to device-independence, not
just text.
Received on Thursday, 26 August 1999 17:46:59 UTC