Re: A Proposal To Not Establish "Minimal Requirements"

Eric Hansen wrote:
> 
> Version: 23 June 2000, 01:03 hrs
> To: UA List (w3c-wai-ua@w3.org)
> From: Eric Hansen
> Re: A Proposal To Not Establish "Minimal Requirements"
> 
> I like the way the Ian has organized the document "Determining Conformance
> to the User Agent Guidelines" [1]. It helps show why further clarification
> of certain checkpoints may not be necessary and helps focus our attention on
> a certain checkpoints for which further work may be needed.
> 
> Proposal
> 
> I propose we do NOT establish "minimal requirements." This may seem like a
> radical proposal, but I will explain why I take this position. This memo
> provides a rationale for this proposal and suggests other ways to resolve
> the reviewer's concerns. The memo also provides examples of the proposed
> resolution; I would like to get feedback on this proposal before trying to
> "fix" more checkpoints.

For the most part, I agree with Eric's proposal (and in fact,
our changes in the last two versions of the document have gone
in this direction). However, while there are "obvious" cases
such as the examples Eric provides below, I think there are difficult
where we might not be able to capture the abstract requirement and
the minimal requirement in a compact and effective form. One risk
of pruning checkpoints as one would trees is that the forest may
disappear. Yes, the checkpoints should be as precise as possible,
and verifiable, but if the expression of the requirement is lost,
I might prefer abstracting slightly and expressing a "clear idea"
of what's expected. Maintaining a degree of abstraction affords us
some flexibility in interpretation, which is both a blessing and
a curse. 

Also, I think there are some checkpoints where we cannot 
enumerate all the possibilities (e.g., configuration of
keyboard input) and therefore must settle for a forest-like 
approach.

Recall how we modified the structural navigation checkpoint, from
"Allow the user to navigate according  to structure" to a form
that reflects the author's need more clearly: "Allow the user 
to navigate efficiently to and among important structural 
elements identified by the author." It might be beneficial
to refine it further as the minimal requirements, something
similar to:

   "For markup languages with known semantics, allow the
    user to navigate forward sequentially through important 
    structural elements. For other languages, allow forward    
    sequential navigation through the document object."

However, I think there are some checkpoints where we cannot 
enumerate all the possibilities (e.g., configuration of
keyboard input) and therefore must settle for a forest-like 
approach. 

I look forward to reactions from others in the Working Group.
I've sprinkled a few more comments below.

  - Ian
 
> Characteristics of a Good Checkpoint
> 
> I think that the checkpoints in the WAI guidelines documents should meet
> several characteristics. The first three points have been discussed before.
> I'd like to propose a fourth point.
> 
> 1. Comprehensibility. The meaning of the checkpoint should be clear.

Yes. 

> 2. Feasibility. A checkpoint should be practical from a technical and
> practical standpoint.

Yes. 

> 3. Verifiability. One should be able to verify conformance.

Yes, except sometimes this is hard, even when the requirement of
the checkpoint is clear.

> 4. Minimalism. The checkpoint itself should express the minimal requirement.

Yes.
 
> I think that these characteristics are related and somewhat dependent on
> each other. We might think of these four characteristics as different facets
> of a unitary concept that we might call "validity" of the checkpoints. (A
> unitary concept of "validity" is also found in the area of educational
> testing.)
> 
> Minimalism
> 
> If the working group succeeds in clearly expressing minimum requirements
> within the checkpoint itself, then there is no need for separate minimal
> requirements. 

I agree. And I would note that some of our efforts to establish minimal
requirements have, thus far, led to modifications that have shrunk
the scope of the checkpoints to minimal requirements.

In some cases, however, I haven't seen a way to express some
abstractions
AND have that phrasing capture minimal requirements.

> A checkpoint that is already minimalistic is not susceptible
> to further reduction (unless justified by new information).

I don't agree. For instance, it would not be possible, nor 
desirable, to list all the possible keyboard configurations that must
be available to the user. Therefore, we say "allow the user to
configure the keyboard configuration." That's the minimal form
that meets our needs. 
 
> Pitfalls of Establishing Separate Minimums
> 
> There would be several negative consequences to establishing minimal
> requirements that are separate from the checkpoint statements themselves.
> 
> 1. Indication of Padding. To establish a minimal requirement that is
> separate from the checkpoint statement itself may be seen as an indication
> that we have "padded" the checkpoints, i.e., made requirements that go
> beyond what is justified by their actual impact on people with disabilities.

I agree that we must be cautious here.

> 2. Confusion About the Number of  Priority Levels. To divide or parse a
> checkpoint (especially a Priority 1 checkpoint) into a "minimal" portion and
> "beyond minimal" portion would, in effect, add a new level of priority,
> perhaps one that we might call Priority 0 (zero). By having separate minimal
> requirements, we would be saying, in effect: "Even though we rate this
> checkpoint at Priority 1 (meaning that some disability group would find it
> "impossible" to access content if this were not followed), there is yet
> another level of priority -- Priority 0, for which failure to conform would
> make content "REALLY impossible" to access." The non-sensical nature of such
> an implication would add confusion about the meaning of the priority levels.


> 3. Indication of Failure to Make Checkpoints Comprehensible, Feasible, or
> Verifiable. I think that since the four facets of comprehensibility,
> feasibility, verifiability, and minimalism are related to each other, then
> failure to ensure that checkpoints statement themselves are minimalist may
> indicate that there are problems in the other characteristics as well.
> 
> In summary, I think that, to the extent possible, we should make the minimal
> requirements obvious in the checkpoints themselves, at least to the extend
> possible, given the necessity for comprehensibility, feasibility, and
> verifiability.

Yes, I agree with that.
 
> Solution
> 
> I think that the solution is to simply clarify the checkpoints themselves as
> necessary but not go down the slippery slope of providing minimal
> requirements.

I agree, but haven't found it "possible" in some cases to state the
requirement and the minimum in a compact or useful form.

>  If we want to add clarifications, advice, guidance, or other
> information beyond the checkpoint itself, that can be put in a note or in
> the techniques. In some cases, new checkpoints might need to be added. I
> think that the "expected solutions" need to be treated this way as well.
> 
> Examples
> 
> Checkpoint 2.2:
> 
> <OLD, 10 June 2000 UAAG as cited in 20 June Determining Conformance
> document>
> "Checkpoint 2.2 For presentations that require user input within a specified
> time interval, allow the user to configure the time interval (e.g., to
> extend it or to cause the user agent to pause the presentation automatically
> and await user input before proceeding). Minimum: Pause.
> </OLD>
> 
> <NEW>
> Checkpoint 2.2 For presentations that require user input within a specified
> time interval, allow the user to configure to cause the user agent to pause
> the presentation automatically and await user input before proceeding.
> 
> [If the working group wishes have the user be able to "extend" the "time
> interval", then that should be expressed either in a technique, a note, or a
> new lower-priority checkpoint.]
> <NEW>
> 
> ====
> 
> Checkpoint 4.16:
> 
> <OLD, 10 June 2000 UAAG as cited in 20 June Determining Conformance
> document>
> Checkpoint 4.16 Allow the user to configure the user agent to limit the
> number of open viewports. Minimum: One.
> </OLD>
> 
> <NEW>
> Checkpoint 4.16 Allow the user to configure the user agent to limit the
> number of open viewports to as few as one viewport.
> <NEW>
> 
> [1] http://www.w3.org/WAI/UA/2000/05/ua-minreqs
> 
> ________________________________________________________________________
> Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com

-- 
Ian Jacobs (jacobs@w3.org)   http://www.w3.org/People/Jacobs
Tel:                         +1 831 457-2842
Cell:                        +1 917 450-8783

Received on Friday, 23 June 2000 04:25:00 UTC