Checkpoint 5.4 and user agent dependencies

Dear WCAG WG,

After a discussion with Wendy about Lee Roberts' recent email [1]
about checkpoint 5.4 (user interfaces), Wendy suggested that I
share my thoughts with the WCAG WG. [Thanks again to Wendy
for reviewing this email.]

This email focuses on questions of conformance, and
relies fairly heavily on the conformance mechanisms used by
UAAG 1.0 [2].

I'd like to discuss three questions in this email:

  1) What user agent features does WCAG 2.0 assume are
     implemented?

  2) How far should WCAG 2.0 go in requiring as part of
     conformance to WCAG 2.0, that the author be able to
     show implementation of those features?

  3) Is there any reason to distinguish content that
     creates a custom interface from other content?

The WCAG WG faces a number of challenges on the subject of
user agent dependencies.

First, the WG must identify which user agent features are
expected to be implemented in order for the content requirements
of WCAG to work. For instance, WCAG requires authors to provide
text equivalents under the assumption that the user agent will
make them available to the user.

Since WCAG is format-neutral, it's difficult to state generically
what user agent features must be implemented. It would be easier
to be able to say something like "Since we are only talking about
SVG, the SVG viewer must be able to do the following..."

Second, since WCAG is format-neutral, and the Web is a diverse
environment with old and new formats, the WCAG WG faces the
challenge of ensuring that for a given piece of content in a
format (call it "F"), there really is an accessible user agent
out there that implements F.

[1] http://lists.w3.org/Archives/Public/w3c-wai-gl/2003JanMar/0114
[2] http://www.w3.org/TR/2002/REC-UAAG10-20021217/conformance

-----------------------------------------------------------
1) What user agent features does WCAG 2.0 assume are
    implemented?
-----------------------------------------------------------

Since UAAG 1.0 is a Recommendation, it should be possible to
identify required user agent features using "UAAG 1.0 conformance
profiles." We didn't have UAAG 1.0 when writing WCAG 1.0, and so
we ended up with the "until user agent" provisions. I hope that
WCAG 2.0 will not include any software requirements other than in
terms of UAAG 1.0 profiles. This means, for example, that the
"level 2 success criteria" of checkpoint 5.4 in the 8 Jan 2003
WCAG 2.0 WG draft should be deleted:

<delete>
   2. device-independent access to functionality is provided

   3. accessibility conventions of the markup or programming
   language (API's or specific markup) are used
</delete>

UAAG 1.0 includes both of those requirements, so they can be
safely dropped from WCAG 2.0.

If UAAG 1.0 does not include something required by WCAG 2.0,
that's unfortunate, and the WCAG WG should let the UAWG know.  In
that case, WCAG 2.0 might need to include a software requirement.

I think it's more likely that UAAG 1.0 has more requirements that
WCAG 2.0 needs. And therefore, the two WGs should work together
to establish a suitable UAAG 1.0 conformance profile that meets
the needs of the WCAG WG. That work should start with a statement
from the WCAG WG of which user agent features are assumed.

Once a profile has been defined, checkpoint 5.4 should be
deleted.

Then, the WCAG conformance chapter can say:

  (1) In order to conform to WCAG 2.0 for a format F,
      content in F must satisfy WCAG 2.0 checkpoints.

  (2) As part of a conformance claim, the author must
      also make a UAAG 1.0 conformance claim (which
      includes a conformance profile).

Or, informally, the author has to:

  - Ensure that the content is accessible, and
  - Ensure that a proper delivery context exists.

WCAG 2.0 checkpoints should only be requirements related to the
characteristics of content. Requirements related to other
conditions in the world (e.g., the existing of UAAG-conforming
user agents) belong elsewhere. Aside from being a simpler model,
this separation promotes automatic conformance checking.

Assertions by the author belong in conformance claim:

   - "This is the version of WCAG 2.0 I'm referring to..."
   - "My content conforms to WCAG 2.0 requirements as follows..."
   - "I am assuming the following UAAG 1.0 profile. ..."

And so on.

-----------------------
On conformance profiles
-----------------------

WCAG 2.0 is format-neutral, and that makes the construction of a
UAAG 1.0 conformance profile difficult. It would be much easier
to say, for example, "If you are implementing SVG, then your user
agent has to satisfy the following UAAG 1.0 profile." When the
format is not known in advance, the profile is harder to build.

One way to help is to use the UAAG 1.0 content type labels.
Thus, WCAG 2.0 could say "If your format F is a format for
creating images, then your user agent must satisfy the following
UAAG 1.0 profile for images..." That would help for the handful
of content types listed in UAAG 1.0.

This suggests to me that WCAG 2.0 conformance profiles might be
useful.

Conformance profiles allow flexibility in conformance claims but
increase complexity. Conformance profiles were perhaps the
hardest issue the UAWG faced for UAAG 1.0. I am happy to discuss
this further.

WCAG 2.0 could allow authors to indicate different conformance
profiles. However, for the sake of simplicity, there should be a
few simple default profiles for common authoring practice. In
WCAG 1.0, there were three options: levels A, double A, triple
A. In UAAG 1.0, there are several more. We don't have enough
experience with UAAG 1.0 conformance claims yet to know whether
we've hit the sweet spot in trying to find a balance between
flexibility for claimants and simplicity for those who read
claims.

----------------------------------------------------------
  2) How far should WCAG 2.0 go in requiring as part of
     conformance to WCAG 2.0, that the author be able to
     show implementation of those features?
----------------------------------------------------------

A conformance claim is a set of assertions. In UAAG 1.0 terms, a
claim is "well-formed" if it contains all of the proper
information (e.g., a conformance profile). A claim is "valid" if
the user agent ACTUALLY satisfies the identified requirements.
Today, W3C does not validate claims. This is relevant to the
question of whether the author should be required to show that a
conforming user agent actually exists (or, for that matter, that
the content ACTUALLY satisfies the content requirements of WCAG).

WCAG 1.0 did not require the author to show implementation of
user agent features as part of conformance to WCAG 1.0.  I have
the sense that the WCAG WG wants to go further for WCAG 2.0, but
that is somewhat problematic since W3C does not act as an
assuring party today.

There are other challenges to including a requirement to show
real implementation, such as How many user agents must be
demonstrated? On how many platforms? What if the user agent is
not available for free?

Some of that work is done in Candidate Recommendation period: the
WG advances once it shows sufficient implementation. But, for
instance, the evaluation reports of the UAWG include big
disclaimers that they are not validity assessments.

----------------------------------------------------------
  3) Is there any reason to distinguish content that
     creates a custom interface from other content?
----------------------------------------------------------

In my opinion: No.

The desired outcome is that the user be able to use all content
via an accessible user agent. If the author creates content
in a format F, then (based on what we said above), the author
had better be sure that there are accessible implementations
of F out there, whether F is HTML or something less widely
deployed.

-------
Summary
-------

a) Delete checkpoint 5.4
b) Make it a conformance claim requirement that the
    author include a UAAG 1.0 conformance claim.
c) Challenge: Design WCAG 2.0 conformance profiles.
d) Challenge: Identify appropriate UAAG 1.0 conformance profiles.
e) Challenge: How to assess the validity of both the WCAG and UAAG
    conformance claims.


-- 
Ian Jacobs (ij@w3.org)   http://www.w3.org/People/Jacobs
Tel:                     +1 718 260-9447

Received on Wednesday, 29 January 2003 13:07:34 UTC