AGWG Teleconference -- 04 Jun 2024

<laura> Scribe: laura

Announcements

<scribe> Scribe: laura

ac: welcome.
... Any new members?

<Chuck> re-welcome Ben in your new role!

ben: I'm now with University of Oxford

<alastairc> https://lists.w3.org/Archives/Public/w3c-wai-gl/2024AprJun/0102.html

ac: Publication Sub-group is refreshing.
... If you are interested, have a look.

maryjo: getting very close to finishing.
... will be having ag review to be done by Tuesday the 11th. Then have CFC ending the 14th.

Conformance and structure continuation

ac: Will share my screen.
... continuing from last week.

<kirkwood> link to presentation?

<alastairc> https://docs.google.com/presentation/d/17VJvnm5UQW4WUzIoo9QNPVGfePgaZa8ifZWs-wtmv7E/edit#slide=id.p

Kevin: Conformance is Not Compliance.

<alastairc> (same as last week)

Kevin: Compliance relates to laws, regulations, and policy - not W3C and AG remit
... Conformance relates to standards - W3C and AG remit.
... Exclusions and reasonable efforts concepts are more relevant to Compliance

ac: went through structure last week.
... Guidelines are very high level
... Outcomes have an AND between them
... Required Methods have an OR between them
... Tests would have AND
... Assertions are under an Outcome, inside or next to the decision tree
... Best Practice Methods (Optional), also have an AND between them
... Some Methods may have both a Required and Best Practice level within them.
... User Needs are under Outcomes

<bruce_bailey> FWIW, U.S. 508 call its exclusions "General Exceptions" and the metric for "reasonable effort" is anything not an "undue burden".

<alastairc> https://docs.google.com/document/d/16Njhx88F2LhwcElu74Jgynx-cRmLJkQONt9hGBJ9Cm8/edit#heading=h.5dy11m3uvex6

Guidance Example: https://docs.google.com/document/d/16Njhx88F2LhwcElu74Jgynx-cRmLJkQONt9hGBJ9Cm8/edit#heading=h.wagpghmtgki

ac: we have categories instead of Principles and then Guidelines and Outcomes below that. After we have flushed out more outcomes, we should revisit this structure to see if we should collapse it or not.
... The purpose of this is to experiment with what the Guidelines Section of WCAG 3 might look like.
... (explains the Image Alternative Example)
... (explains the Keyboard Focus Location outcome example)

<Zakim> Rachael, you wanted to add a few clarifications

<Nayan> The document should make it clear that Conformance and Compliance are two different things and spell it out.

rm: we have 2 that were done by the group. The Chairs did the rest. They are not perfect. More work to be done.

<Nayan> Since the distinction is often conflated and confused

gregg: It is shaping up.

<Zakim> GreggVan, you wanted to say looks good except for the "required methods" part. Should be "Sufficient Method options" and allow for other and new methods. If we make it

gregg: If it is required it can't be an "or"

<kirkwood> +1 to sufficient method

gregg: methods should be thought of as sufficient methods.
... how should not be required.

<Nayan> Rachel I brought it up because Kevin spoke on it.

<alastairc> q/

gregg: best practices: would like them as a recommended at a higher level than the methods.

ac: not sure about sufficient methods
... test procedure is being lifted up.

<kirkwood> who determines importance?

ac: what if hiding decorative images were not as important as content images.
... that is one aspect of it.
... we have Methods then techniques. Could be one line.

<Zakim> Rachael, you wanted to explore Gregg's point about best practices

rm: think we have built in best practices at multiple levels.

<Zakim> GreggVan, you wanted to say that hiding decorative is not a proper requirement under "Alternates are available" it is actually an exception and a best practice.

rm: agree to bringing them up to outcome level.

gregg: We need to differentiate between outcomes and recommendations.
... hiding decorative is not a proper method.
... standards are fragile. They should say what the outcomes should be. But not how to do it.

<Zakim> alastairc, you wanted to comment on decorative images

ac: could have standardized end of the decision tree.
... useful to say you could meet this in a different way.

<Zakim> jeanne, you wanted to agree that outcomes should be normative and methods and tests are informative, but we don't have to have requirements and best practices. We could have an

ac: equivalent can be a null alt attribute.

jeanne: agree with the outcomes to being normative.
... assumption that we do conformance item by item. Could do it overall.
... could give more flexability. Like the direction we are going.

<Zakim> GreggVan, you wanted to say +1 to adding "or something else" ALSO look at the size of the techniques doc and the testing - and if we make WCAG3

gregg: like saying "Or something else". But then it is not a requirement.
... size of doc will be enormous if we put in all the support information in the document

<dan_bjorge> +1 to Gregg's concerns; I agree that I'm quite nervous about how much "technique" we're moving towards moving into the main document

<Zakim> alastairc, you wanted to ask whether outcomes alone can really be considered testable?

<Nayan> +1 on Gregg

ac: outcomes are higher level as current SCs.

<Chuck> +1 to wendy

ac: makes me a bit nervous. Seems like not enough there.

rm: could use who, what, when as wilco previously brought up.

<Zakim> GreggVan, you wanted to answer Alastair

gregg: Could do more elaboration as outcome level.
... could look at sufficent techniques.
... understanding the outcome is as important as having it named.

wendy: agree with not making techniques and methods normative.
... need to have a high level outcome. and make sure it is not gamable.

dan: trying to understand ac 's concern.

ac: wondering about the decision tee or methods where we have a common test procedure.

<Zakim> GreggVan, you wanted to endorse and ask a question of Wendy

gregg: yes. need to make it not gamable.
... what if every major browser provided great text alternatives?
... if it is better and if it automatic why not use it?
... author would only have to check it.

<Zakim> wendyreid, you wanted to answer

wendy: how do we define good enough? Context really matters.

<Zakim> Rachael, you wanted to state that one discussion topic we will need to have is when is platform provided solutions "good enough"

rm: this is a discussion that we have popped in and out of.
... we need to decided what is good enough. Can have a discussion on GitHub.

<Zakim> jeanne, you wanted to say that my concern is what motivation the browser would have to provide this service? All the experts I read dare being very cautious that the profit

<GreggVan> +1 to wendy we need to only adopt those approaches when they actually work

jeanne: Just want to be cautious.

<wendyreid> +1 Jeanne there's the market forces to consider too :(

jeanne: experts I am reading are talking about profit driven companies.

<GreggVan> +1 to disability not being a sufficient driver --- but getting out of having their staff do a lot of manual work is a big incentive.

<Zakim> Rachael, you wanted to respond

jeanne: Let's not rush into an assumptions that AI will improve a11y.

rm: good enough goes beyong AI.

ac: structure discussion on Github.

rm: Possible Overall Conformance Models.
... With Baseline Levels, Baseline + x%, Baseline + badges.

<alastairc> Accumulating points based on guidelines/outcomes https://docs.google.com/document/u/0/d/1nxKmm835paNJlZN6nJHyqtyjg93VTK44osbzWo3TGXU/edit

rm: Without Baseline: x% of guidelines/outcomes or Accumulating points based on guidelines/outcomes
... how important is it that we have a Baseline?
... Every outcome is important to someone so if we have a baseline, how do we prioritize and justify?

dan: recall conversation that without a baseline could create an inequitable conformance model.

wendy: In reference to meeting 25% of requirements but you could pick which ones.

<kirkwood> level of effort creates an ‘undue burden’. is what we are dancing around. depends on audience. baseline is impossible

wendy: Baseline is important for equity.
... doing to be difficult but we have to have one.

<Zakim> Chuck, you wanted to ask for scribe change

<dan_bjorge> scribe: dan_bjorge

<Zakim> alastairc, you wanted to provide my preferred goal

alastairc: John commented around "undue burden". I look at it from the perspective of standalone WCAG, where it provides a way to say "here is a reasonable amount of things you can do to make sure that something is accessible, without causing undue burden."
... By that I mean, applying a particular regulation makes it not worthwhile to do what you wanted in the first place. Thankfully that's rarely the case in digital technology. But if you have a baseline, that should be a reasonable thing for organizations to aim for. If you don't have a baseline... I don't know about that, it feels like a foreign

concept.

<kirkwood> 2.1 AA is a baseline

scribe: If there's a particular guideline you could give it "extra points", but then it'd enable folks to ignore other requirements if it's just meeting a point threshold. I lean towards having a baseline, plus mechanisms to encourage going beyond it.

<Zakim> jeanne, you wanted to propose a hybrid model that gives some protection

jeanne: 1. I think a viable direction is to look at a hybrid model, with an overall baseline but not specific requirements that make up the baseline.
... 2 reasons for this: 1, at workshops/talks, I've heard people think "there's a baseline, so I'm done once I meet it." Many organizations that want to "do accessibility" really only want to meet the baseline, so that's all they do. So I don't want to have that model encouraging that.

<ShawnT> Conference: Accessible Canada Accessible World: https://sites.events.concordia.ca/sites/accessconf/en/accessible-canada-accessible-world/home

<kirkwood> +1 to Jeanne there is no way to do equitable baseline

jeanne: 2, I recall a subgroup taking on a project of "what could be the baseline", and they concluded that there was no way to do an equitable baseline. There's always some accessibility requirement that some group needs, and if you make a baseline, you must be choosing some groups over another. Therefore they recommend no single baseline, but to

instead use a baseline that was overall-accessibility-of-a-product oriented.

scribe: Want to look at a hybrid model where we're looking at overall points or some kind of percentage and say "this much is required, you must do above and beyond to meet bronze". So has a little bit of both; some protection against skipping expensive requirements, but balancing doing more.
... In the proposal Rachael mentioned, I included some notion of a scoring means that attempts to address this. Of course it's not perfect but I encourage folks to take a look at it.

<kirkwood> +1 to Jeanne, talke a look at JF doc

<GreggVan> can you provide a link to IT

<alastairc> Accumulating points based on guidelines/outcomes https://docs.google.com/document/u/0/d/1nxKmm835paNJlZN6nJHyqtyjg93VTK44osbzWo3TGXU/edit

scribe: It's the "Accumulating points based on guidelines/outcomes" link in slide 9 of the Conformance Conversation slides

<laura> s/sufficient methods/sufficient methods/

wendyreid: I'm thinking about the idea of "one's role in the accessibility ecosystem" and that one's baseline might want to be different for different people (small business owner vs browser developer vs platform provider, etc)
... I wonder about conformance being based on who you are, with different responsibilities for different audiences.
... Maybe a business owner should be responsible for using accessible tools, where a platform provider might have more detailed requirements about how others can use their platforms to achieve accessibility

<Nayan> +1

<Zakim> GreggVan, you wanted to ask by baseline do you mean some subset of what we can require? or do you mean 'everything that we require' as being the baseline?

GreggVan: We're meant to talk about "what's accessible or not", not "what is the burden on different entities". Creating the burden isn't our job, it's the job of regulation based on our standards.
... We have things that are the responsibility of the tools vs the content creators and in the past we've used different sets of guidelines for those. You end up with problems if the content creators get confused seeing rules only applicable to user agent developers. If we put the latter in, in the extreme case, W3C leadership has shot down

user-agent-centric guidelines

<laura> s/flexibility /flexibility /

Jeanne: Clarifying stance on baseline: That in each guideline, we have a base level outcome that could be required, and then we'd say "you must meet some mininum in each guideline that meets the basic needs of each group". eg, we could require captioning. But then to say "overall, to conform, you must do enough things *beyond* the baseline too, and

those you can choose"

scribe: So there would be a distinction between "non-required (but testable)" and "assertions"

alastairc: So in a WCAG 2 context, that would map roughly to something like "you need to meet all A criteria plus some percentage of AA criteria"

<Zakim> alastairc, you wanted to suggest having view-level conformance, and then an aggregate/product level conformance based on product-level assertions. That could allow for differences

<laura> s/goood enought /good enough /

alastair: I suggest an option that considers the scope of conformance (views vs items, etc). Suppose we could say "this *view* passes at bronze level" that's primarily based on individual methods/tests, but to say a *product* conforms you need more of a focus on stuff like "style guides" that apply across a wide scope. That'd be a separate level of

conformance that *might* be required or not by regulations/etc, but could be considered separately from interface-level stuff that's more granular.

<jeanne> I also want to test how valid the proposal is and change the proposal as we have develop more evidence. I'm not attached to the details. I am attached to improving the equity.

scribe: If you have to consider all the interface-level stuff at every level, it doesn't scale.

<Nayan> +

<Nayan> Q

<kirkwood> concerned “baseline” conversation will get us nowhere- so whose needs do we eliminate/prioritize? … do we want to separate from regulatory requirements?

<laura> s/goes beyond AI. /goes beyond this AI discussion./

Rachael: Would like a straw poll on "should we have some form of a baseline" vs "no baseline"

<alastairc> poll: Should we have a set baseline? +1 / -1

<Ben_Tillyer> +1

<wendyreid> +1

<MJ> +1

<Francis_Storr> +1

<Chuck> +1

<bruce_bailey> +1 for some baseline requirements

<jaunita_george> +1

<Nayan> +1

<Makoto> +1

<GN015> +1

<Azlan> +1

GreggVan: Clarification: when the poll says "baseline", do we mean something that doesn't include everything possible?

<laura> +1

alastairc: yes

<maryjom> +1

<kirkwood> How does baseline compare to WCAG2.1.

<Nayan> agree

<jeanne> -1 because we cannot have equity for disabilities with the traditional baseline

<GreggVan> +1 definitely not % of things -- you pick what you don't want to do

Nayan: Suppose someone is missing audio description for their videos because they're embedding youtube, but the rest of the site is conforming. Would that pass this baseline (or accumulate "enough points")?

<kirkwood> you just said that by scoring sufficient points you could ignore other requirements. that is not baseline to me.

alastairc: No, that wouldn't pass a baseline, assuming audio description was part of the baseline. You could potentially still conform if we decided audio description was part of the "beyond"/"points-based" stuff.

Nayan: Then yes, I lean towards a baseline approach

<jaunita_george> +1 to john

kirkwood: Very concerned about this conversation. The idea that you could meet a baseline so that you can ignore certain requirements, and those requirements you ignore are the ones you need to satisfy a lawsuit or something, and it would not actually satisfy that. It breaks down to the idea of just trying to prioritize "these are not things we

should be doing at all, what things are required to make things accessible to different folks" - that's not us, and that's a bad trend to go down, this'll just collapse in itself I think. I don't think it's the right route.

GreggVan: +1 to John Kirkwood. If your whole page is accessible except this one part that you just can't implement, you should just say that. It should be up to the rule-makers whether that's acceptable or not.
... If I buy a business and their website isn't accessible, and I can't resolve it the next day, we should still label it inaccessible, and leave it up to the rule-makers whether there's a timeline to adhere or whatever

<Nayan> where does the "label" come in - conformance statement? accessibility statement?

GreggVan: If something is *required* to be accessible, that should be in the baseline. Beyond that if you want to make best efforts to meet best practices... well, most best practices are that way because they aren't testable. But if something is merely "hard", that shouldn't affect whether we say it's accessible or not - "too hard" should be a question

for regulators.

<Zakim> Chuck, you wanted to say that the difficulty is in the details

<Nayan> but the regulators are not capable in most part to make those cherry picked decisions

<Nayan> they are looking at WCAG

<jeanne> +1 Nayan

Chuck: I was seeing this vote as being "for" a baseline meaning "let's continue to explore this, being sensitive that not all baseline strategies are equal and a +1 doesn't necessarily mean I support one specific baseline"

<Rachael> +1 to the details and wanting to find a solution that is as inclusive as possible.

wendyreid: I think some people have heard "baseline" and not had a shared understanding of what we mean. "baseline" could mean "10% of the reqs" or "90% of the reqs". That's the hard part of this discussion, where do we draw the line. I caution folks against getting too caught up in how strict the bar is - the vote now is "should we have a bar at

all".

<Nayan> The bar needs to be "reasonable" to accomplish for a non-technical person

<laura> s/flexability /flexibility /

<Frankie> +1 to continuing to explore the potential of a more inclusive baseline, not support for one specific baseline at this time

<Zakim> alastairc, you wanted to say that whichever choice we make, we can't include everything.

juanita_george: Echoing John Kirkwood - anything we choose as our baseline *will* just become the standard. Organizations will simply hit that minimum and not go further. I worry about setting the bar too low and giving too much leeway to organizations. If you say "60% passing conforms", that's what orgs will do and no higher. If we set a bar, we

need to set it high.

<Nayan> there will be legal challenges to a bar that is very high.

<GreggVan> Can someone post the link to the proposal for baseline? I see a link to the non-baseline approach but not the baseline thanks

alastairc: I think because we're familiar with how WCAG 2 works and how WCAG AA became a de-facto baseline, I'm not sure how well I understand the proposal Jeanne made and how that could work differently. We'll have to come back to that.

WCAG 2.2 issues https://lists.w3.org/Archives/Public/public-wcag2-issues/2024May/0014.html

Rachael: I'd like to start an exercise where we propose what should be "required" and what shouldn't, to give us more of an idea of our shared assumptions for future followup on this topic

<Rachael> Please make the time to complete the exercise before next week

<mbgower> https://docs.google.com/presentation/d/14PG_NnH3hh0TcukLm9tb-xluKHzb79AE/edit?usp=sharing&ouid=118419493369958965106&rtpof=true&sd=true

mbgower: We have our usual set of proposed changes from TF to AGWG that got sent out in the agenda, but 1 item for discussion is particularly interesting (#3873 opened by Detlev), and I'd like to get a litmus test and get everyone aligned on the problem first.
... Crux of issue is "is there an assumption of consistent use of human language on a page?"

<kirkwood> what do you mean by ‘labels’

mbgower: Consider a page where one specific label on a form is in a language different from the rest of the page, or one specific heading ont he page is in a different language than the rest of the page. Ignoring accessibility for the moment, do we think such a page would pass QA?

Chuck: I have assumed that there is a consistent human language on a page. It could deviate from that, but there's a clear "parent language"

<alastairc> kirkwood for example, a form with "name" and "address" in a different language

mbgower: To me this is an underlying assumption

<Azlan> Would agree with the exception of direct quotes

dan_bjorge: Disagree, with several examples of pages with different languages of similar "primary-ness"

mbgower: Sure, but there's probably some rationale for using different languages

giacomo-petri: Also disagree, I think there are examples where multiple languages occurs

mbgower: Sure, but there's probably some rationale for using different languages

kirkwood: What do you mean by "labels" here?

mbgower: Form input labels

GreggVan: I think everything is in 1 of 2 categories, either it's on purpose and it's covered by "the exception", or it's a mistake/bug. We aren't writing regulations that say "there can't be a bug on the page", we're describing what it means to be accessible.

<GreggVan> Success Criterion 3.1.2 Language of Parts (Level AA) ljoakley2 The human language of each passage or phrase in the content can be programmatically determined except for proper names, technical terms, words of indeterminate language, and words or phrases that have become part of the vernacular of the immediately surrounding text.

mbgower: So is there an assumption of consistent human language for alternative text on a page? eg, consider audio in a video being in a page's language, but audio descriptions in a different language. Or consider an image on the page where alt text is in a different language from most of the page.

<kirkwood> so with example of labels thus means the translation of assistive technology fields. the audio description be in same language

mbgower: Presuming that these are not intentional, do we think these should pass QA?

<Zakim> GreggVan, you wanted to answer

GreggVan: Maybe it passes QA, maybe not. Not our job. I don't think it passes accessibility.

<DuffJohnson> Concur with Gregg. Content language must be addressed at the object-level - IMO, "the page" isn't significant to this consideration.

<alastairc> ack \join_subline

MJ: Agree with Gregg. Don't think it would pass QA either, I'd expect QA to test the audio description and to notice that it's in a different language. It should fail QA as they're doing accessibility testing.

kirkwood: I've seen this scenario before, and yes, the alternatives have to be in the language of the document. I've been in a courtroom around that one. It has to be in the language of the document.

giacomo-petri: I think there's a big difference in the "audio description" scenario vs the "alt text" scenario. One is clear to a non-AT user and one isn't.

<kirkwood> Urdu almost killed me ;)

mbgower: Scenario 1 has subscenarios - may or may not be visible to non-AT-users

<GreggVan> what about Success Criterion 3.1.2 Language of Parts (Level AA) The human language of each passage or phrase in the content can be programmatically determined except for proper names, technical terms, words of indeterminate language, and words or phrases that have become part of the vernacular of the immediately surrounding text.

mbgower: So, can we actually normatively fail stuff that mismatches language? Few criterion cover this

<MJ> Agree with Gregg about 3.1.2

Chuck: When you were asking "should QA fail it", I'd have actually said "QA should fail it but it's not an accessibility issue because I don't think it's a normative WCAG requirement". I'm a textualist; if the text doesn't make it a requirement but we don't like that, we need to fix the text, not invent requirements we wish were there.

GreggVan: What about 3.1.2 language of parts? Wouldn't it fail that?
... <inaudible>

<alastairc> Gregg - you can mark it up correctly, but it would still be useless to someone that doesn't understand it.

<GreggVan> also the word "alternative" it is not a proper alternative

mbgower: You can mark it up in the actual language of the alternative even if it's in the "wrong" language
... don't think that helps with this problem.
... Some normative options in the conformance section to try to address this. ...1: "Conforming alternative versions" are required to be in "the same human language" and if muiltiple language versions are available, then there must be a conforming alternative for each language offered ...2: Full page: conformance is for full pages only and can't exclude parts ...3: Accessibility-supported: requires that the way that technology is used be tested for interoperability with AT in "the human language(s) of the content"
... Suggest that stuff that refers to "programmatically determined" (including "text alternative", whose definition does), it's likely covered by. the "accessibility supported" wording

<Zakim> GreggVan, you wanted to say my main point was missed. 1) it is not an alternative if in a different langauge and b) if done by mistake - it is out of scope for us.

GreggVan: If it's in another language, it's not a "proper alternative"

<Ben_Tillyer> Gregg, all language of parts says is that "The human language of each passage or phrase in the content can be programmatically determined " Even if its in the WRONG language, its language can still be determined

GreggVan: We should define the term "alternative" to clarify that it needs to be in the same language and just put that in as a note/erratum/whatever. Anything else we do is just a whoopsie-doo. If you give an alternative in another language, it's not an alternative.
... You also keep saying "well if they inadvertently do it it's a mistake" - that's irrelevant for us/WCAG. WCAG just says "what's accessible", not "when are mistakes allowed".

mbgower: I think there's an underlying assumption that alternatives must be in a language matching the content

<GreggVan> yes that would be the thing to do and can be done in the UNDERSTANDING document

<Zakim> Chuck, you wanted to ask for return to the slide with the conformance explanations (5.x.x)

mbgower: We either have an ability to insert that assumption (if we believe it), but I want to be careful about intent. If there's an intent of language changing (as opposed to oversight), it might just be someone misunderstanding the language mismatch.

Chuck: What's key to me here is an understanding of the 5.2.x language. I think I'm landing where you're at that this text supports our interpretation.

mbgower: Task force suggested creating a stock statement to add to each relevant SC that cites the assumption of consistent human language and the supporting 5.2.x comformance language
... Will this lead to an immediate objection from anyone?

<Zakim> dan_bjorge, you wanted to clarify "add to each relevant SC" (understanding vs normative)

<Chuck> scribe+ Chuck

<alastairc> Comments: Place to comment: https://github.com/w3c/wcag/issues/3873

<Chuck> dan: Clarify, what you said the proposal is to add to each relevant SC. I thought we mentioned understanding docs.

<Rachael> Thank you for that clarification

<Chuck> mg: Yes, in understanding docs.

<bruce_bailey> The Working Group will undertake to add notes to understanding documents to state that text alternatives and equivalents should match the human language of the original content (normally the default human language for the page).

<GreggVan> +1

<Chuck> +1

<giacomo-petri> +1

<Rachael> +1

<bruce_bailey> +1

<alastairc> Poll: address in the understanding documents

<kirkwood> +1

<Ben_Tillyer> +1

<maryjom> 1

<GN015> +1

<MJ> +1

<ShawnT> +1

<scotto> +1

<Azlan> +1

<Francis_Storr> +1

<laura> +1

<Frankie> +1

<ljoakley2> quit

- DRAFT -

AGWG Teleconference

04 Jun 2024

Attendees

Contents

Announcements

Conformance and structure continuation

WCAG 2.2 issues https://lists.w3.org/Archives/Public/public-wcag2-issues/2024May/0014.html

Summary of Action Items

Summary of Resolutions

Scribe.perl diagnostic output