- From: Al Gilman <asgilman@iamdigex.net>
- Date: Wed, 04 Dec 2002 11:23:01 -0500
- To: "Ian B. Jacobs" <ij@w3.org>, DPawson@rnib.org.uk
- Cc: wai-xtech@w3.org
At 08:36 AM 2002-12-04, Ian B. Jacobs wrote: >> c) in the absence of preferences from the user, the User Agent >> should follow author preferences > >UAAG 1.0: > > 2.1 Render content according to specification. (P1) The UAAG requirement here depends on the format specification to say the right thing. The XAG tries to guide that format spec in what the right thing is to say. In this case the UAAG provision does not assure that the XAG provision is met; if both provisions are met the user will have the access we seek. >> d) the User Agent may display more than one, or one other than the >> first feasible choice, should this be appropriate from what the >>User Agent knows of the delivery context opportunities and preferences. > >UAAG 1.0: > > 2.3 Render conditional content. (P1) > 2.9 Render conditional content automatically. (P3) Here is a case where I don't think that the UAAG requirements cover what XAG needs to say. The key, once again, has to do with the discriminants between when you follow case c) or case d). Formats have to indicate (in the semantics bound to their syntax) a distinction between behavioral suggestions which have the semantic weight of author proposals or preferences, and the more limited set of propositions or constraints on User Agent rendering and otherwise binding of the experience which qualify as bona_fide constraints of the application. The concept of a notice that must be presented if a related service is to be exercised is an example of a bona_fide grouping or atomic transaction requirement. This is a keep-together constraint which reflects intrinsic and unavoidable requirements of the application. On the other hand, the customary widows-and-orphans-curing "keep together" annotation may be violated for good reason for example in tiny[1] delivery contexts. Large fonts and screen magnifiers (i.e. adaptation strategies typical of the visually impaired or low vision user) may make a tiny context of almost any text presentation frame, regardless of the total pixel size of the physical medium. The formats further have a burden to attempt to express the constraints in as repurposable way as possible. This means eliminating dependencies on the response time of the user wherever possible, and providing alternates where not possible. Yes, that is a burden on the application designer, but the format designer needs to discover where the patterns of application expression in the format can systematically reduce any such dependencies, and support authoring tools in alerting application instance creators in recognizing the need for an alternate path through the interaction logic where the need arises. Al [1] 'tiny' as in "tiny display" See for example the family of -Tiny profiles from W3C. >DPawson@rnib.org.uk wrote: >> >>I like the sense of these comments Al; >>Could we extract some of it for XAG? > >Is this really for XAG? As I see it, yes; there is stuff that belongs in XAG, but it may not be obvious until one adds back into the change proposal a few points that Dave missed, and I may have not spelled out enough. Suffice it to say that the things you have recapitulated are the instance requirements that must be supported, and these are traceable to existing writ, but that the feasible range of "how to do this in a format" expressed in the part of the message <quote> In other words, there is a declarative assertion of a bag of roughly-equivalent resources, an author's expected ordering of preference for selection of one, and an understanding that some other optimization to select one or more than one may be appropriate in some circumstances. So a select-first-match-from-list default structure with a permissive escape to processing as the superclass select-multiple-from-bag is a rough summary of the format structure. </quote> gives evidence of a more analytical view of the format design problem that interprets the instance requirements in terms of XML dialect constructs. I actually believe that what I see as appropriate here is more clear when we apply the remaining XAG guidelines in this scenario. The point with the multiple images is that the differences between the alternatives have to be made clear to both person and machine, so there is an information structure that has to be captured in the format which goes beyond just a bag of image objects. This is required under XAG Guidelines 2 and 4 as a response to the guidelines "1. support WCAG" and "3. support UAAG" [I paraphrase heavily]. The other hard part is that the Accessibility requirement does not resolve whether the format specification will bind similarity semantics *implicitly* through collecting the similar items in one container element as in html:object or smil:switch or indicate the associations through *explicitly* marked relationships (attributes of type IDREF | IDREFS) or even more indirectly through relationship assertions outside the primary syntax tree as in smil:metadata (see also SRGS). So we may need to be getting into writing queries against infosets as the medium of expression for what it is that is required. But there is content in there, and simply identifying SMIL as the first place to look for prior art for content control and SVG as the first such reference for grouping and annotation within the XML syntax, we will have helped format designers a lot. How can I put it? I think that the fact that Paul felt moved to ask shows that there is a use for something beyond the operational requirements on the instances that are captured in the WCAG and UAAG. Al >Looking at Al's comments below (Al, please correct me >here if I've chosen the wrong checkpoints): > >>In order to offer the user multiple images where >> a) the author believes any of these will do the job > >WCAG: provide equivalents. > >> b) the User Agent can display at least one if there is one that >> meets delivery context constraints. > >UAAG 1.0: > > 2.3 Render conditional content. (P1) > >> c) in the absence of preferences from the user, the User Agent >> should follow author preferences > >UAAG 1.0: > > 2.1 Render content according to specification. (P1) > > >> d) the User Agent may display more than one, or one other than the >> first feasible choice, should this be appropriate from what the >>User Agent knows of the delivery context opportunities and preferences. > >UAAG 1.0: > > 2.3 Render conditional content. (P1) > 2.9 Render conditional content automatically. (P3) > > >>Each image should be provided with an appropriate textual alternative. > >WCAG, 1.1. > >I don't see any format requirements in the above >statements. For XAG, the requirement is to allow >the author to specify equivalents. See a long list >of suggested XAG requirements starting "Allow the author to..." >in: > http://lists.w3.org/Archives/Public/wai-xtech/2002Sep/0028 > > _ Ian > >-- >Ian Jacobs (ij@w3.org) http://www.w3.org/People/Jacobs >Tel: +1 718 260-9447
Received on Wednesday, 4 December 2002 11:19:46 UTC