W3C Workshop on Metadata for Content Adaptation - October 2004

Position paper prepared by

Al Gilman

As an individual.  This material has not been reviewed by nor consensed on by the Protocols and Formats Working Group or anyone else representing the W3C or WAI.

Mapping the demand and supply properties domains  to the point of interoperation

Take-Aways

Meta-Negotiation: sharing decision capability information, negotiating binding responsibilities

Mark Birbeck, speking at the 2004 W3C Technical Plenary, raised the idea of "Defer decisions when you know downstream processors can handle them."  This leaves us with two questions:  How can the upstream processor know that the downstream processor can handle this decision?  How much does the upstream processor need to know to trust that the downstream processor can handle this decision better?  There is clearly risk, from the perspective of the upstram decision process, in deferring.  So there has to be a material expectation of gain (better results) to offset this.

Data thesaurus: compare and contrast of terms and scales incumbent in demand (UI, preferences) and supply data domains

In principle we could compare and contrast and reduce until we had accomplished a definitive mutual data normalization for the property terms used in user interfaces and their adjustments and in media objects and their transforms.  But not fast.  For the near terms we need to begin to recognize where scales touch aspects in common and roughly how they map.  Because initially we will have much more oppotunity to adapt at the server out of information entered for other reasons at the client, and to rescue a failing presentation on the client with the aid of gross choices and rough characteristics available from the server, not everything you ever wanted to know.

What's Old

Author Proposes, User Disposes

In the Device Independence Concepts, a distinction is drawn between an optimized presentation and a functional presentation.  The point is made that the user should be given the option of re-binding UI effects so as to restore a functional presentation if what someone had thought would be an optimized presentation is not going to work at all.  Giving the user control over presentation options has been developed in the User Agent Accessibility Guidelines [UAAG10], and providing the assets needed to populate these alternate views in the Web Content Accessibility Guidelines [WCAG10, WCAG-WG].  In the development of W3C Format Specifications, where a format was about to violate these principles the Protocols and Formats WG has generally been able to gain remedies by appealing to these principles [someWAI].

A Rationalized Development Process Aids Multi-Channel Delivery

Let's just hit the high spots:
This applies to serving uses with different devices, and it applies to serving users with different needs and preferences.  This is recognized in the work of the Device Independence Working Group and of the Authoring Tool Accessibility Guidelines Working Group.

What's New

Disaster Management adds a Market

Recent attention has focused on infrastructure for public information in emergencies and particularly disasters - emergencies affecting large numbers of people and quite possibly affecting the communications infrastructure as well.  Here is an application where a command-and-control communications and intelligence infrastructure has to be rapidly configured out of assets of opportunity - what terminal devices people actually have and pay attention to - and what communication network assets survive.  A functional presentation is what is desired, not spit and polish.  In its requirement for "readiness to operate in alternate modes" it is almost exactly like accessibility in its push to build Device Independence capability into a broad range of information dissemination systems.

Education has a Proposal for us

Since the last time I got to discuss this with many of you at WWW2003 [DevDay], the education community has made considerable strides in bringing forward proposals for accessibility metadata [IMS ACLIP].  This material has to be considered and compared with the spectrum of needs across different applications and particularly the decisions involved in optimizing a presentation to the level of being commercially competitive.  But the bottom line is still "Here's a way to do it.  Best it or use it."

Analysis

Decisioning: Think Globally, Act Locally

Cascading Style Sheets is set up on the premise that 'rendering' takes place one 'canvas' at a time.  I have been trying to assert that decisions about styling, what presentation effects to use to represent which content distinctions, should be decided from a broader perspective to meet WAI needs.  I finally think that I have the 'toy example' it takes to explain why.  This is the sheaf of cards that is sent to a WAP1 browser, if I recall correctly.  Here there are a tuple of views.  Always the same kinds of views.  Part of the user interface is knowing what aspects of a situation are addressed in what views.  In such a delivery context, there are things that have to be emphasized or distinguished to the user.  But not in all views.  Some distinctions will be prominent in one view and suppressed in another, because the aspects emphasized in each view are comprehended by the user after a few dialogs.  A screen reader session works much the same way.  There is a screen-review view, a list-of-links view, a tables mode and a forms mode.  Each of these views has its own listen and feel.  It doesn't make sense to make un-coordinated presentation policy decisions in each view; the criticality of a given distinction for the aspect at the heart of each view should be considered in allocating vocalization effects to different qualities of the content.

That's fine in principle.  What could be done about it in practice?  Perhaps the user-defined test variables in SMIL give us a clue.  The global analysis could create an allocated climate of decision-conditioning parameters for the styling of each canvas, or a profile of content-model properties to be exposed (with emphasis gradations) in each sub-view.  This is a wild guess, but that is an example of how global conditions are conveyed to local decisions in one W3C technology.

Warning: The term 'delivery context' could get us talking past one another

A naieve or demand-side reading of that term would mean the conditions under which the user interaction takes place.  Yes it includes devices, user, and environment.  But it includes all aspects of these that could fuel a better adaptation, and to enough precision so that more information would not lead to better results.

If I have it straight, the way this term is used in the Device Independence documents, it referes to a collection of parameters which contribute some information about selected aspects of the 'delivery context' as conceived in the first paragraph above.  There is no warrantee as to how precisely, how accurately, or how thoroughly.

For the purposes of this workshop, I think we need to keep in mind that there is an approximation relationship between these two, and it is the proper business of the workshop to enquire whether there are aspects that should be better covered in order to yield a more effective adaptation end-to-end or a more cost-effective negotiation of what adaptation gets done where.  So somehow we need to devise a provisional vocabulary that will allow us to talk about either of the above concepts and not get confused about which we are addressing.

Optimization uses finer characterization of [supply, demand] properties than survival does

Accessibility is an 'ility' and tends to ask the questions: "Is there a way to achieve a functional presentation?  Can the user steer the process into such a configuration?  On the other hand, the commercial customers of Device Independence techniques are generally interested in having their services presented in a way that is competitive with other services hand-crafted for each of the several channels addressed.  [With important deviations, such as enterprise functions, hospital infrastructure for patient orders, etc.]  However I will persist in the generalization.  To do the Device Independence job, you prepare a variety of image options and you are seeking to select the best one for the current delivery context.  So you need to know enough about each to understand the tradeoffs down to the point of "which is best."  This is more information that what is generally needed to discover "is there one there that I can make do?"

My prejudice would be that we should go ahead and use the more fine-grained scales that address the commercial demand where they exist, and infer functional success or failure from abstractions of these finer properties.  This still may leave the accessibility community seeking coverage for aspects, such as interpretation of words, that aren't yet covered by commercial practice in multi-channel delivery of services.  [But globalization is a tide floating translation and internationalization boats.]

Bits and Pieces

User runs out of 'contrast' adjustment

This may be apocryphal, but somewhere I think I heard that the DI concept (or is it the current core presentation properties schema) doesn't address environmental factors.  Here is a little use case.  Suppose a user is using a graphical PDA in a situation where there is a lot of glare from the environment and the screen is hard to read.  No, the device doesn't have an ambient glare sensor.  But if the user exercises the contrast control repeatedly until running out of adjustment capability, we might reasonably infer that a content adaptation would be appropriate.  An upstream event asking if the upstream processes could format the material to be less dependent on device contrast capability might bring down a high-contrast version that would save the day.

'Usable at' size range

At present we have content asserting a fixed size for itself, in the area of images.  In actual practice even pixel-array images are zoomed by screen magnifiers and in other situations as well.  Vector graphics are inherently scalable in terms of device capabilities, but still aren't usable below a given size.  The accessibility problem -- can we get to a working adaptation -- would be advanced if we could get images rated as to the min and max size over which range they constitute a 'functional presentation.'  This doesn't have to keep us from setting a 'best at' size property, but it would help to understand what range of transformations they would tolerate reasonably well, and what transforms could be know ahead of time to be useless.  It could help in optimizing page layouts as well, such as guiding the rollover between full image and line-drawing morphs.

Even checkpoints (success criteria) can be too coarse for some uses.

There is a use in the educational arena as follows.  The instructor in a course is using SCORM [SCORM] educational modules as assignements for her students.  One of the students in the course has a disability.  One of the modules the instructor would like to use has problems for this individual.  How can the instructor find an alternate assignment to reach the same educational objectives?

Here it may be overly broad to know "this module is missing text equivalents for some non-text media objects."  For a specific student, it may make all the difference in the world whether the media objects missing their alternatives are visuals or audio tracks.

When there is no course-level sequence that a) follows the instructor's tailored syllabus and b) meets the needs of all students for all lessons, then it is reasonable to try to rescue off the scrap heap module-by-module enough lesson plans to put together an alternate course of study that works for this specific student.

Readings

[UAAG10]  User Agent Accessibility Guidelines 1.0

[WCAG10] Web Content Accessibility Guidelines 1.0

[WCAG-WG] The Web Content Accessibility Guidelines Working Group is working toward a second-generation WCAG.

[someWAI] some WAI comments on Device Independence

[DevDay] collected presentations from the WWW2003 Developers Day Track 6

[IMS ACLIP] IMS AccessForAll Meta-data Specification

[SCORM] Shareable Content Object Reference Model Initiative ( SCORM )