type enforcement (recapitulates disability adaptation)

This post speaks to the contentious aspects of the current Proposed TAG Finding discussed in the thread at

 http://lists.w3.org/Archives/Public/www-tag/2002May/thread.html#99

Unfortunately, two issues have been confounded both in the language of the Proposed Finding, and in the ensuing discussion.  I would distinguish these as

- when to recognize a type error, and

- how to handle a type error

** when has a type error occurred?

In other words, there is a distinction to be drawn between "optimistic upgrading" of the "process-as" type, and genuine repair or "pragmatic recovery" involving changing the "process-as" type.

The TAG should not be taking a position against empirical type determination.  By this I mean attempting processing a data collection under the constraints of a type and seeing if it conforms.  This is a primitive and cannot be totally eliminated from the system.  Real systems run with finite error rates, including errors in type marking, particularly the way the web resource food chain runs today.  It is one thing for XML to say that content that fails to be well-formed XML is to be thrown away forthwith.  XML can enforce that, and it works.  This is not true across the gamut of MIME types in use on the Internet.  Not yet.  Don't go there.

Empirical type determination is particularly appropriate in cases where processing is first attempted using the announced type and this processing fails.  Here we have a bona_fide type error.

On the other hand, switching the "process-as" type on the detection of suspected upgrade opportunity without so much as a by-your-leave to the author (no longer available) or customer (presumed luser) is indeed wrong.  But we need to describe this misbehavior with more surgical precision, to make it clear that the terms of the finding are reasonable.

** How to recover from a type error

In terms of how to recover, the user should be in command, but help is helpful so long as it only helps.  This is territory that we have ploughed and re-ploughed in the WAI, so let me try to give some context and history.

Look at how the adaptation control is staged in the User Agent Accessibility Guidelines, currently in Candidate Recommendation:

First, the interaction takes place at the optimum nominated by the author.

Next, systematic tuning adjustments built into the format and applicable across the type of resource representations are used.

Lastly, adjustments specific to the specific user and specific representation component instances are possible.  But note, only as a last resort.

The user has total control, tools from the technology, and hints from the author.

Compare this summary with Guideline 2 in the UAAG 1.0

http://www.w3.org/TR/UAAG10/guidelines.html#gl-content-access

I would suggest that in the user agent implementation of user control over process-as type for inteaction with web-delivered data, that all of the above layers and considerations carry over verbatim.

What relates here can be recapitulated as follows:

- The user's terms of engagement with the content of a resource representation is subject to adjustment
 + under necessary constraints required for comprehension of the content
 + under the user's command

- Automation to assist the user in exercising this control is very helpful.  At times it is essential, because a welter of twiddle factors can become a serious barrier to access in its own right.

- But when this automation steps over the line from assisting to supplanting user command of this process, it has stepped over the line into misbehavior.

The systematic tools in the technology are supertypes in the family of types.

Processing as a supertype is graceful degradation and is compatible with what the author has indicated about her content.  Processing as a new, incompatible type, is not known to be consistent with a valid dialog transaction, but may still be attempted given the de_facto error rate in communicating type information from the author.

Supertypes loom large in the future prospects for defining the limits of delivery-context adaptation that is consistent with the sense of the content.  Until we have a stable practice of multi-level partial understanding that has distinguished definitions of the sense of the communication and the default realization of that sense in physical user interface media, we should hold a place in the architecture for supertypes as one of the tools for achieving this.

Note that this story is also summarized in a message to the DI group which has consensus support from the Protocols and Formats WG in the WAI.

 some WAI comments on Device Independence
 http://lists.w3.org/Archives/Public/www-archive/2001Nov/0069.html

Al

Received on Thursday, 23 May 2002 10:04:28 UTC