24 March 2002 WCAG WG F2F Minutes


Summary of resolutions and actions

Morning meeting with EO

Minutes were taken by the EOWG.

Issues discussed:

General thoughts about reorganization proposal

/* Andi Snow-Weaver scribed this session */

Bob: If you have questions you think need to be addressed in guidelines, send to Bob.

Lisa: How does this relate to the FAQ?

Gregg: Feeds into FAQ and into WCAG guidelines

Cynthia: Goal is to have guidelines answer the most common questions. e.g. Gregg's re-write of guideline 1 is trying to answer the "what about the Mona Lisa?" and "what about the symphony" type of questions.

Gregg: Want to explicitly cover the exceptions that we intend so that people don't use them to cover exceptions we don't intend

[Gregg introduces discussion of reorganization of guidelines]

Kevin: Didn't see reference to color blindness

Gregg: Goes in "separate content from presentation". will add.

Cynthia: Also nothing about making sure you maintain contrast

Gregg: Will clarify under "emphasize structure through presentation".

Wendy: We have an issue logged gainst WCAG 2 on this (may not be reflected in issues list, however).

Paul: Like the words "perceivable" and "operable" at beginning of the guideline. would like to see re-wording that eliminates one word and replaces with really short phrasing of guideline.

Gregg: People take our stuff and shorten it. Did it this way so that one word becomes "handle" but guideline itself is complete thought. If shorten the guidelines further, they "become" the guidelines and important things may be lost.

Jason: Difference between function and information is described in the details of 1.x checkpoints just not in the guidelines or checkpoints themselves.

Gregg: Should we stick to new proposed checkpoint groupings for discussion or should we stick to what was in last draft?

[group agrees to use new groupings but still need one for 5th group (cross-cutting issues)]

Cynthia: What about old guideline 4 - Technology considerations. Design for compatitibility and interoperability.

[various suggestions made - agreed to wait until we work on the checkpoints and then high level wording for guideline will suggest itself.]

Matt: Looking for "interoperability over time both forwards and backwards"

Gregg: Technology considerations (durability, robustness, and timelessness).

Bob: Interoperability of other technologies is as important as interoperability with AT.

Paul: How about a one-sentence summary of guidelines such as "ensure perceivability, operability, .... ". Something that covers all the principals.

Lisa: Something that would look good on a t-shirt.

Carlos: Be careful in choosing wording that doesn't translate well to other languages. "Technology considerations" translates pretty well but "durability, robustness, and timelessness" may not. Other words could be used in explanatory sentence where you can explain what you mean by them.

Gregg summarizes: Use new structure but work to be done on wording for guideline 5.

Gregg: Will preserve the old numbers in square brackets for a while because some people identify strongly with the old numbers.

Wendy: We have a short id for each of these checkpoints. Might be useful to use those for a while.

Gregg: Good idea.

Conformance and minimum browser

Gregg: Moving on to conformance and minimum browser. Can either have guidelines that describe everything that can be done and people can pick what they want to do or could have minimum set. Can't have checkpoint for which you have no success criteria. How do we get someone to at least try to meet the requirement? Introduced new idea of success criteria. (See new cp 3.1) Authors have done as much as they think is possible and at least the two things. Then added list of "advisory recommendations".

Kevin: Governments will only take the top layer of requirements and forget the rest because that's the only thing you can argue about in court. UK legislation uses word "reasonable". As technology expands, idea of what is reasonable gets easier to do. Be careful if we layer things in a hierarchy because regulators tend to take top layer. Only thing people will be thinking about at the end of the day is what appears on the web sites and what they can get resolved in courts.

Gregg: If set minimum layer, really set a ceiling. Some will argue that if you don't set a ceiling, companies will set their own ceiling lower.

Gian: Someone can take you to court if they are disabled and can't access your site but can't take you to court if they are using IE 3. doesn't fall under disabilities legislation.

Kevin: Important legal point that what you are trying to legislate for is delivery of whole solution.

Gregg: Speaks to backward browser sort of thing. if someone argues that site is not accessible with technolgy they like to use, may not hold up in court. what if it requires someone to purchase a $1K.

Cynthia: In defense of hierarchy, we are not writing law. we are writing technical specification for a group of people for which hierarchy is essential

Jason: Agree with cynthia. Relationship between guidelines and policies varies greatly depending on jurisdiction you are in. Guidelines are set of technical requirements that indicate their effect on accessibility. Can be used to develop laws and policies. Guidelines have to support that kind of usage. Now that we have a draft of success criteria, may be time to revisit conformance.

Gregg: Industsry has two different needs around accessibility - have to know what they have to do to comply and have to be able to innovate using new technologies.

Kevin: EU lifted WCAG 1 and made it a schedule to a piece of legislation requiring level 1 conformance.

Cynthia: Have discussed the possibility of a policy toolkit or a scheme that lets you conform at a checkpoint level. Organizations implementing have freedom to choose which checkpoints they will conform to. More concerned that people know what they're supposed to do and how to do it. we may not have legal skill to determine policy considerations.

Mark: Agree with kevin's points. on section 508 page, it says WCAG 1 was not developed with law in mind. WCAG 2 is being developed with this in mind. Hope to make Section 508 refer directly to WCAG 2.

Gregg: Add some evidence both for and against. In US, used WCAG 1 as "picking list". priority 1 minus all cognitive, added some priority 2s and 3s. Reworded some a little bit. Some differences accidental, some are on purpose. Some international companies faced with two standards. Couldn't have anything in a law that says "if you do something, it will be more usable". have to just tell them to do it. Can't have anything in a law that nobody can do yet. Standard doesn't have to say what you have to do. regulation says you have to do it but standard must be implementable and testable. Once we get agreement on draft, need to get feedback from regulation writers.

Jason: Some regulations being written as "recommendation". interpreted as "good practice" by people interpreting disability accommodation laws.

[Kynn joins]

Bob: Different standards that industry has to comply with. e.g. Germany and Denmark have different standards.

Gregg summarizes: What we generate now has to be robust enough that WCAG 3 doesn't look radically different. have to make sure that people implementing WCAG 1 can easily proceed to WCAG 2.

Bob: Multiple implementations even in countries that have not tried to rewrite the guidelines.

Bob: Seem to be worrying a lot about how this will play out in different regulatory environments. Can we get some of the people who develop regulations to participate in the working group?

Wendy: EO group is working with regulatory groups (or representatives) adopting WCAG 1. We should coordinate with EO on this. Concerned that Judy is not here for this discussion.

Gregg: Maybe Judy could come into this meeting when EO takes a break.

Jason: Need to be thinking about possible conformance schemes without thinking only about what regulators will do.

Gregg: Another consequence of having the regulations be the same as our guidelines is that all training materials can be shared. In US, there have to be two sets because 508 is different from WCAG 1.

Katie: Agree that we need to keep our eye on international regulations as they are developing.

Gregg: Moving to discussion on old 3.3 and 3.4. Proposes no minimum "set" but minimum implementations within checkpoints.

Discussion of supplementing text with illustrations

/* Bill LaPlant, US Census Bureau joins us for this discussion.

Cynthia Shelly scribed this session. */

WAC: There are 2 aspects to this: Illustrating text, symbolic language Wendy presents her document, which she will post to the list, and shows a printout of the News to You site, which uses symbolic language to present news stories.

GV It will be a real challenge to build a concept code system that is compatible with the various symbol languages. Many of these systems have gaps, so this may cause some of them to fill in their gaps. Have they looked at the Esperanto work?

LS Will concept code system be compatible with RDF?

WAC, Carlos: don t know. A member of WWAAC has joined the WG.

LS: Multitasking between reading and comprehension. Sometimes you re concentrating on reading the words and the content gets lost. Perhaps we could define different types of content, and for different types of content, define different levels on conformance. Something like an instruction you should always supply information. Same might be true of a concept.

WAC: What do we think about this info?

GV: in the past we ve had screen-readers etc, but we have not had the ability to translate into the symbol system (or sign language). If you have the ability to go down from a symbol system to a standard and up again, you can also translate to sign. Perceivable could have a new meaning, that it could be translated to a symbol system. Augmenting with symbols isn t practical, but replacing might be. Translation research. Could also be used to allow things to be used in the future that aren t available now by building an interface layer.

Mark How credible is this? Is there work in scientific journals to support symbol languages? How about towards other languages? Is this just for cognitively disabled, or could it be used to communicate with someone who doesn t speak English?

WAC: Yes, there are many people using symbol languages. It s primarily for people with communication difficulties. There s also Bliss, which is sort of a symbolic Esperanto.

GV: Bliss designed for international communication. There s a great video called the symbol man about the man who designed bliss. It s interesting to immerse yourself in the symbols and the power they have to communicate. They were using them with children.

[I didn t catch the name of the movie]

JW: It might be possible to establish a conformance scheme to the guidelines where every checkpoint could be implemented at different levels. Such a scheme might apply very well to these checkpoints (old 3.3 and 3.4). Can associate metadata to any web content to describe how abstract, concrete, complex etc it might be. It could also express levels of annotation, etc. People could recommend different levels it would be good to reach for different types of content and different situations. A general metadata requirement seems to be supported by existing technologies and isn t particularly onerous. Could start looking at concept languages and metadata, and other types of interactions and demonstrations, etc. All of this discussion of concept systems reminds me a of a lot of work that s being done in philosophy to try to boil down and make language unambiguous, for example Frager wanted to establish a logical script rather than natural language. Layers of conformance works particularly well in this instance.

Carlos: WWAAC has an email client that uses symbolic language. Translates symbols to some kind of code, and translates to symbols for other countries. There are some feasible technical solutions. Shorter text is better for some parts of population, perhaps it could be customized with server-side techniques. ATAG defines conformance minimum set, should we adopt that too?

LS: This kind of language and tools will need some things that we re doing anyway, like perhaps the guidelines to use literal text, and what we were hoping to include so that words could be pronounced correctly (using the vowels in Hebrew text so screen readers are more accurate). These tools kind of increase the scope of the need and the benefits of these existing suggestions and requirements. If you can t guess the word correctly through an electronic means, then your symbol translation will be inaccurate and useless.

WAC: I think that what you re saying is that if we have concept codes in the content, then it could be translated to literal text, or it could be read by a screen reader if the vowels aren t there? Screen readers and people who need literal text could maybe make use of the concept codes.

GV: most people who are doing web sites wouldn t be able to put in concept codes. In order to generate the concept codes, you need the vowels and literal text.

LS: What Greg said.

KB: The news to you thing reminded me of ruby backwards.

Greg, Wendy: agree

Bob: Would authors have to learn the symbol language?

Wendy: no. They would use concept codes.

Bob: some sort of XML language?

WAC: i assume so.

Carlos: Don t know yet. 2-3 months away from xml app.

Bob: at least a year away from authoring environment? Big companies would look at this too, but& Are we too far out in front??? Perhaps this group should make the decision if the technology is mature enough for a reasonable implementation

WAC: yes we have to figure that out. Users have to adopt it, and we have to decide when it s ready. This is the until user agents issue. Also want to tie together Lisa and Jason s concept: Hierarchy of conformance and levels/types of content.

Kevin: It s nice to look forward instead of back for a change. Can tie level of conformance to the author s interpretation of importance. Throws the responsibility for accessibility to the providers of goods and services, not the providers of applications.

GV: WWAAC work is in many ways is like Braille. One way is that if you have concept codes, and I used a tool to translate my site to concept codes, I would have no idea if it did a good job. One approach is to translate and then reverse translate to see if it did a decent job. If you have a good set of concept codes, you should be able to run it out to concept code and back, it should be the same. If you go all the way to the symbol system you will loose some complexity and shades of meaning. If the tool will generate the codes, why not do the text-to-concept codes-to-symbols at the display end, not the content end? The author doesn t know how to write in symbols, so the tool can do it. If the save-to-code is good, then the display end can be improved over time without requiring changes to authoring software or sites. Those symbols could also be used for language translations too. Need to make sure the language is non-ambiguous.

Carlos: This project is only European, so it might be even longer for US and other countries to integrate. Perhaps an interim step would be to have a guideline to provide a page summary like you do a table summary.

Bob: If the technology is not ready, there are going to be instances where we have good ideas for technology that doesn t exist yet.

GV: there is a WAI research and development committee.

Bob: how do we say, this isn t ready for prime time, but we don t want to throw it out. So we don t pit against each other non-technical authors who will have trouble implementing, and people who will benefit from this tech.

LS: In this instance fits together nicely. More reasons to ask people to write clear, simple, literal text. Other reasons that it s good to do it now. Can use them to drive to future issues, since they share solutions. New user agents will have more capabilities to take advantage of this stuff for more users. Also, to Kevin s point. They may say this content isn t important for these people to be able to access it, because they couldn t (for example) understand the concept of taxes. You can t control your audience.

Gian: Agrees with Lisa.

JW: different hierarchical levels of conformance in a checkpoint. For 3.4, minimum could be some kind of metadata, illustrations and concept codes further down the hierarchy. As the tools become available, there s more reason to move down the hierarchy and do more. Guidelines could include all of the measures that are available. As user agents and authoring tools improve, there would be more policy reasons to move down.

LG: expanding vowels in Hebrew is an example of 1.1. PDF is like that too. We assume that text is always available, but that s not always true. My other comment is related to forward looking tech and until user agents. We ve said that until the user agent is ready you, the author need to compensate for the UA. With this tech, it s different. You don t need to do this yet, until there s a UA that can read it.

Bill: develop prototype tools. Government would put out an RFP for a tool, and people will try to implement it. This is the ultimate in the distinction between content and presentation. In essence, you are going to a most generic content in symbology, and the encoding in language is a form of presentation. One of the goals of markup has always been to go in that direction and conceptually separate out content from presentation. Concept codes is the ultimate of this.

WAC: more prototypes than the ones already being developed by WWAAC?

Bill: More is better.

Wendy: Lisa did you understand that concept codes are not available today? How does literal text get translated into symbols?

Lisa: mark up all non-literal text as such, perhaps using Ruby.

Wendy: Perhaps authors can work with services that already know how to produce specialized (e.g. symbol) content. Similar to RFB&D and other orgs that make specialized formats for the Blind. Maybe a Web Service? If I have to write everything as literal, that s not really realistic. I can t post a joke? Does the NY Times need to write to the level of USA today?

(General shaking of heads)

GV: checkpoint 1.4 make text unambiguous. Naturally ambiguous words marked semantically to allow disambiguation. When available, code work in conceptual root coding. BUT, what about jokes? Puns? These are often lost in translation (to sign or other languages), and people What about nuance? There is no word for some concepts in some languages. Concept codes could allow us to have a cognitive equivalent of a screen reader. That s huge. Why mark up if it s not used? So that when the content becomes legacy, it s useable by assistive tech.

CS: We should not make a requirement that all content must be at a 3rd grade reading level, and that puns aren t allowed.

Bill: Language is culturally contextual, particularly when we re talking about puns, jokes, etc, so part of doing the modeling is how do you take out the cultural context, or is it possible to do that. It may not be possible under certain circumstances. A joke may hang on cultural context or how the language sounds, so you loose the context when you change the mode. The mode from verbal to textual presentation of various kinds or a cultural context mode.

WAC: This has been useful, and I will try to summarize this again in the documents I have, and we ll see what happens in the list.

GV: If this comes across as new developments that will allow people with cognitively disabilities to get content, it shouldn t be controversial. Bob s idea for driving research. Need published work in respected journals. You could probably get serious funding for something like that with the right people involved.


JW: Things coming out of this discussion in relation to conformance. Within each checkpoint start stratifying conformance levels within the check point. Annotations in metadata, existing markup for semantics, symbol systems is another level of semantics. There is a continuum, and we could set down minima and different levels of conformance that could be reached. Policy setters could change the level of requirements and the technology progresses. We could examine the process for combining and reworking checkpoints and success criteria, to enable layers, without subsetting the checkpoints.

GV: to the group: This concept of having conformance not being throwing in or out whole checkpoints, but instead having different levels of conformance within checkpoints. When you get to a certain level, you can just say it s plus . You can just keep going and do more and more if you want to do a great site.

CS: isn t this what we ve been talking about for a month with testable line?

GV: this is conformance, which we were avoiding

CS: and the idea of plus. Need a way to claim conformance in metadata that s accessible to tools at a very granular level. Perhaps a requirement to do so.

Andi: specific example. 3.1, provide structure with content. Minimum would be it has as much markup as you think is appropriate . Is a .txt file ok? Not all documents require structure, for example poems or a letter.

JW: Possibility of deriving a readable linear reading order might be a minimum. Continuum from pure text to markup, to RDF/semantics, to the kinds of concept codes we were talking about in the last session . As you move down the levels, you will increase access to more people.

KB: I like the concept that everything we write down are things you have to follow, you can t ignore any of them. One thing I m concerned about from WCAG 1.0, is that there is not any flexibility for building your own policy rather than using the pre-defined ones (A, AA, AAA). I m not sure they ve been good for us in practice, but there s no way to tweak them before 2.0. What if conformance were not intrinsically linked into the guidelines? Perhaps we publish with it a complex compliance scheme, which could change over time. The only compliance that s in the guidelines is do all of them (at least minimally). We d use the complex compliance scheme to give people a tool for doing more. Analogous to XML base, and other related documents.

GV: similar to what we re doing in techniques.

Bill: There are efforts to have web sites assert what level of compliance they have to various standards. There s something at the University of Maryland that could perhaps be expanded.

GV: there s a bunch of other stuff in their markup. Like browser support, technologies, bandwidth use, etc.

Bill: yes, but that could be a good starting point.

GV: we haven t done the minimum browser stuff. But we ve got a conformance proposal that does the 2 things we wanted. In 5, as we talked about designing for robustness, you can have a minimum of robustness and levels of robustness.

JW: did a proposal. Minimum is that you provided enough for it to be machine readable. Next level is that you have tested with some level of tools. Next level is that it goes back to really old tech. Cynthia s idea was the web site must publish what the requirements and compatibility is. Could be used w/ or independently of what I m saying. Once people have reached beyond the minimum, content providers can make a fairly complex assertion about support.

Kynn: Do we need a multi-level 5.5 checkpoint?

CS: that s in my proposal. I like the idea of a minimum and complex. We could also provide the EARL for the minimum compliance, which could be linked to sites.

/* Andi Snow-Weaver returns to scribing */

Ben: Allowing author's to self-report which AT's they have tested with provides end users education on what features are available in AT. May find that there are more appropriate tools for them to use.

[Gregg summarizes for Wendy]

ACTION: Wendy to check W3C processes to ensure that we can include conformance requirements in "notes"

Gregg: Danger in Ben's suggestion is that it appears that people will only test with the major ATs so sites will appear as if they only work with those ATs even though they may actually work with others too.

ACTION: Editors will draft the conformance scheme so that the working group can comment. Due date: April 7, 2002

Wendy: "tested by" concept is not currently supported in EARL. Wendy needs to understand the requirement.

ACTION: Wendy to follow-up with Cynthia about possible issues/additions for EARL.

Gregg: If we allow a "tested by" claim, need criteria for that to avoid false claims. Don't want to require that authors mark site as "tested with". Many vendors fear that this is an "implied warranty"

Next F2F

[Wendy polls attendees for ability to participate in F-2-F meeting in Scotland July 11 - 12.]

yes - katie, ben, mark, cynthia, wendy

no - bob regan, loretta, matt, gian

maybe - andi, kynn, jenae, gregg (possible conflict teaching class), jason, lisa

Next teleconference meeting

Next meeting Thursday at the usual time. Gregg will not be attending.

PDF Techniques overview

Katie: PDF Techniques overview.

$Date: 2002/03/29 23:42:00 $ Andi Snow-Weaver, Cynthia Shelly, Wendy Chisholm