- From: Jason White <jasonw@ariel.ucs.unimelb.edu.au>
- Date: Fri, 15 Mar 2002 11:05:38 +1100
- To: Web Content Guidelines <w3c-wai-gl@w3.org>
As discussed at today's teleconference, a frequently recurring issue in this working group is that of how to address the changing capabilities of user agents and other software within the framework of the guidelines. To appreciate the full context of this proposal I suggest reading the minutes of today's meeting. Disclaimer: the following proposal is put forward as an idea for discussion; it does not necessarily have my personal or unqualified support. Nevertheless, elements of it have appeared in working group meetings from time to time. As I remember, Matt was one of the first to suggest it at a teleconference last year. Nevertheless, here is an attempt to crystallize the ideas as a concrete proposal. The essence of the proposal is that the conformance scheme for WCAG 2.0 would incorporate a distinction between what I shall here call "semantic" and "operational" conformance. These may be characterized as types or levels of conformance that apply to the techniques and/or (possibly) the success criteria under each checkpoint. Thus, these conformance types do not distinguish between checkpoints; they only designate the way in which each checkpoint with respect to which a conformance assertion has been made, is implemented. 1. Each checkpoint specifies an access requirement which, if met, will make the content more accessible to one or more groups of users. These are functional, end-user, requirements. 2. Semantic conformance is reached if the necessary information is provided, for example in markup or a data model, to allow a tool to satisfy the relevant end-user need reliably. The test of reliability is that a well understood, practically implementable, deterministic method exists whereby this information can be extracted and processed in software so as to satisfy the end-user need. By deterministic we mean that the method does not rely on heuristics or probabilistic analysis, but is guaranteed to give accurate results. By "implementable" we mean that it has actually been implemented in tools, or that it relies on techniques and algorithms which are widely understood and are regarded as amenable to implementation in the proposed context. 3. Operational conformance is reached if the content is so designed that the end-user requirement expressed in the checkpoint has been met using techniques which are well supported by user agents, assistive technologies or other applicable software. By "well supported" we mean that the pertinent technologies and features have been implemented and available in relevant software for a significant period of time (the exact time period would need to be specified more precisely), are available in internationalized/localized versions of such software, and do not exhibit serious interoperability problems that would preclude implementation by content authors. These requirements can of course be fine-tuned. This conformance proposal would affect our work in at least the following areas: a. In the techniques documents, individual techniques would be labeled as satisfying the "semantic" or "operational" conformance levels (possibly both). In the case of techniques which are claimed to allow implementation of a checkpoint at an operational level (see the foregoing definition), the working group would require that testing have been carried out to show that the technique is practically useful and supported by user agents/assistive technologies. Details of which techniques were relevant to the "semantic" or "operational" conformance levels, would change over time as new tools were released and existing software updated. Thus, whenever the techniques documents were revised, the label associated with each technique would need to be reconsidered in view of the state of deployed technology at the time. b. In our success criteria, the situation is somewhat more problematic. We could either (1) maintain the generic nature of the guidelines by stating the success criteria in terms that do not depend on particular methods of implementation that are likely to change as technology evolves; or (2) specify each of the available approaches as alternatives in the success criteria, then provide guidance in the techniques documents as to which solutions would entitle a developer to claim "semantic" conformance, and which would justify an assertion of "operational" conformance. Both of these potential approaches would be compatible with our goal of keeping "until user agents" qualifications out of the guidelines document itself. c. In our conformance scheme, we would allow conformance claims to specify whether the content met the relevant checkpoints at the "semantic" or "operational" level. Note: one consequence of this proposal is that content developers would not need to consider the subtleties of the distinction between "semantic" and "operational" conformance, since the specifics regarding which techniques counted toward the satisfaction of each of these conformance types, would be prescribed by the working group and stated explicitly in the techniques documents. In effect, the task of determining whether "until user agents" requirements had been met, would be primarily the responsibility of the working group, to be carried out as part of the task of writing and revising the techniques document corresponding to each technology (in contrast with having a separate "user agent support" page).
Received on Thursday, 14 March 2002 19:05:47 UTC