Re: Request for PFWG WAI review of Omitting alt Attribute for Critical Content

[distribution note: I have dropped all the cross-posting.
caveat: personal opinions.]

At 8:04 AM -0500 23 10 2007, Laura Carlson wrote:
>The HTML 5 working group is questioning and debating the need for the
>alt attribute on critical content. In fact, the current HTML 5
>Editor's Draft allows instances where critical content is allowed to
>have no alt attribute on the img element.

** points of likely agreement

Let's see if there are some candidate points that we think might
be agreeable.

* good ALT is good practice

I think that HTML WG and the WAI would all agree that images which
are critical content SHOULD have text, at least an appropriate value
of @alt and in some cases a further long description.

*  function and need for @alt=""

I think that the WAI would say an example where an image is
an interpretation of the accompanying text material is an
inappropriate use for @alt="".  This obscures the point that
there are cases where *no text equivalent* is the right
text equivalent.  Conventionally, this has been encoded

WCAG2 doesn't say that this encoding has to be preserved,
but the 'cowpaths' principles of the HTML WG suggest that
to change it would be regarded as unnecessary thrashing of
the authoring community.

** points of confusion

* critical content images, or all images

The issue has been stated as "must critical content <img> elements
have an @alt attribute set?" whereas the present and plausible
candidate rule in the syntax would be "must *all* <img> elements have
an @alt attribute set?"

The point here is that "critical content images" is not a syntactic
category.  Whether the image is critical content or not is a matter
that requires human judgement -- it lies outside the realm of markup
aside from the convention that @alt="" is used to signal that the
image is best ignored in voicing.  But that's a semantic distinction,
not syntax.

* the issue at hand doesn't impact use of @alt=""

Conscientious authors can use @alt="" the way we want in either
case.  The markup coming from authors who care is unaffected.
What is changed is the relationship between the markup coming
from authors who don't care and the syntax rules of the specification.

* the role of validation in gaining uptake of accessible markup

This is an area where there are likely to be divergent assumptions
among the people involved in the discussion.  We need to surface
these assumptions.

One perspective is that the web has exploded with a muddle-through
level of conformity to spec, but not strict conformity to spec.  When
we got onto this topic during the Plenary session, Dan Connolly
volunteered that the fraction of spec-conformant HTML on the web
was "zero to several decimal places."

The other perspective is that the language specification is the W3C
authoritative statement on what is right and "nothing happens unless
it is a MUST."

I think we have some talking to do before reaching an agreement on
this topic.

One possible outcome is that the syntax does not require @alt but
that the specification says that all critical content <img> elements
SHOULD have a meaningful @alt set and that decorative <img> elements
(steal exact wording from WCAG2) SHOULD have @alt="".

This could be viewed as too weak a statement of principle, but on the
other hand, the HTML WG is trying not to write a specification that
is almost-universally violated. We in accessibility need to hear them
out, and see if our campaigns are actually made more difficult by the
absence of the MUST in the syntax, or if the hard stuff is there
anyway, independent of the syntax rule.

The HTML Working Group is attempting to come up with a specification
that pretty well captures the actual fail-soft behavior of HTML
processors that are in use.  And that HTML5 conformant processors
will process and render essentially all of the existing content on the
web.  Maybe we could get them to raise an informative message
when some accessibility rules are violated.  But I seriously doubt
that we will get them to say browsers should fall over and fail to
process a page where there is one <img> with no @alt.

[maybe I should put that in terms parallel to the above "the WAI
would agree...]

I expect that the HTML WG would consense that such a "hard-fail
on no-ALT" rule is out of the question.  This could still go as an issue
to the Director's meeting, but let's see if we can't avoid that.

** summary

I think that we can combine syntax and semantics to get the
following case tree:

1: @alt is absent -- one may infer that the author didn't think about it
the user agent may attempt repair from context, filename, etc.

2a: @alt=""  -- pretty good hint that the author thought about it and
.. thinks that the image is ignorable in speech (could still be wrong)

2b: @alt="something appropriate" -- when the author did think about
it and pen a good text.

2c: @alt="random garbage" -- when the authoring tool inserted something
that accomplishes a repair to meet syntactic conformance but the
net result is to dilute the audible experience with garbage.

The differences in estimation of author behavior has to do with how
much making the attribute a syntactic requirement moves authors
from case 1 to 2b or to 2c.

The browser still isn't going to validate per the HTML5 specification.

What checks the authoring tools perform is up for promotion.  With
a more realistic spec, more authors may accept/tolerate spec conformance
checking in their authoring tools.  But to get to case 2b, it take actual
though by the author, even with good hints from the authoring tool.

To get to case 2a we need to get the author's attention and
cooperation. I'm not sure we can expect the HTML WG to carry that
water for us. Checking on the author side will be done where we have
convinced content producers they need to do some checks for
accessibility. Will HTML WG move the interoperation realities so far
that they need to for interoperability? We have yet to see. Certainly
not overnight.

The 'jaundiced' perspective could be that "It doesn't matter
whether 'use @alt on all images' is a syntax rule or an
accessibility guideline -- the same subset of all authors will
apply and follow this rule in either case."


>Alternate text is essential for accessibility. There needs to be a
>markup solution to indicate whether or not the alternate text of an
>image is critical to understand the content - omitting such an
>important attribute is ambiguous, and doesn't help anyone. The problem
>is differentiating between ignorant and intentional lack of text.
>The issue is detailed at:
>In order for this debate to reach a satisfactory resolution, review of
>this issue and advice from the PFWG and WAI on the potential
>accessibility impact of omitting alt attribute for critical content in
>HTML 5 would be appreciated.
>Thank you.
>Best Regards,
>Laura L. Carlson
>Steve Faulkner
>Gregory J. Rosmaita
>Joshue O Connor
>Philip TAYLOR
>Robert Burns
>HTML WG Members

Received on Wednesday, 14 November 2007 21:52:43 UTC