Summary of thoughts on conformance

Hello,

I'd like to share my current thoughts on questions of conformance
to the User Agent Guidelines. (I apologize if I have
not expressed my thoughts clearly enough).

1) Conformance should be defined as satisfying certain
   techniques as they are defined in the guidelines document.
   [Henceforth, to avoid confusion with the techniques
   document (and to use a term already being employed
   in other guidelines), I will refer to the techniques
   as defined in the guidelines document as "checkpoints".]

2) Conformance should be defined as satisfying
   one or more subsets of the Priority 1 checkpoints. Subsets
   are necessary since the WG considers that it is too
   much to ask *all* UAs to satisfy *all* of the Priority
   1 checkpoints. 

   How are subsets established?

   We have the option of allowing other companies/organizations
   to establish subsets, e.g., by using their common sense ("Our UA
   doesn't support this feature" or "Our UA doesn't render this
   type of content natively", therefore we don't have to
   satisfy a certain checkpoint) or having the WG do the dirty work.
   I think the WG should define conformance as clearly as
   possible by establishing useful subsets itself. Since
   this implies creating a finite number of subsets,
   the conformance mechanism will be less flexible, but 
   also more reliable. As a compromise between the two, 
   I propose that the WG define some useful subsets, and then allow
   flexibility within those subsets (see point 5).

3) Each subset of checkpoints is given a name for easy reference.
   We should also provide an informative description
   of each subset so that people understand why the
   subset was identified. [Note that no UA has to be
   perfectly described by the informative description. Conformance
   is in no way related to the informative description; it
   is only based on satisfying predefined subsets of checkpoints.]

4) I think there should be a "core" subset of checkpoints
   that all UAs must satisfy to claim conformance. In addition,
   they may claim to conformance to other subsets (that probably
   shouldn't overlap the core subset).

5) The current challenge to the Working Group, in my opinion,
   is to choose useful subsets (and to establish criteria 
   for choosing them). There are many "axes" that one might 
   use to choose subsets, including:

    a) Based on the target device (e.g., to a braille device,
         to a graphical monitor, etc.)
    b) Based on the type of content the UA handles (e.g., supports
         sound, video, images, etc.)
    c) Based on language features (e.g., supports scripts, tables, etc.)

   I think that trying to create subsets for more than one axis will
   be difficult unless the axes are completely orthogonal.
   By this I mean that, for example, a subset of checkpoints named 
   "table" (meant for UAs that support tables, whatever 'support' means)
   must not depend on whether the UA claims conformance to another
   subset such as "sound" or "speech" or "video". In order to
   avoid enumerating all possible combinations, the subsets along
   different axes must be independent. Since I believe this will
   be difficult to achieve, I suspect we will have to stick
   with one axis. I believe Jon Gunderson is leaning
   towards "target device support" as the primary axis. If this
   is the one chosen, the WG could allow flexibility within 
   each subset by including statements such as the following
   in the document:

      "User agents not supporting or not configured to
       support a particular language feature are not required 
       to satisfy checkpoints related to this feature."

   Thus, a UA could support one or more target-device subsets
   but within a given subset, may not have to satisfy all
   checkpoints. 
   
6) Note that the same checkpoint may "participate" in 
   several subsets (e.g., it may be included in the "core"
   subset but also the "tty" subset). 
   I think a checkpoint's priority should be allowed
   to vary from subset to subset. Thus, a checkpoint related
   to table navigation may be less important for visual user
   agents (if "visual" is a chosen subset) than for speech
   synthesizers rendering tables.

To summarize my summary: 
 - Conformance is defined as satisfying a subset (or
   several subsets) of Priority 1 checkpoints.

 - The WG has to establish these subsets in the simplest
   possible manner.

 - The priorities chosen for a checkpoint may very
   from subset to subset.

It is my opinion that the WG should try to establish
meaningful subsets as quickly as possible, then move
on to settling, once and for all, priorities for
the checkpoints.

 - Ian

-- 
Ian Jacobs (jacobs@w3.org) 
Tel/Fax: (212) 684-1814 
http://www.w3.org/People/Jacobs

Received on Monday, 4 January 1999 21:09:42 UTC