Meeting minutes
<alastairc> any scribes please?
WCAG 2 proposed changes https://lists.w3.org/Archives/Public/w3c-wai-gl/2026JanMar/0172.html
Rachael: proposed WCAG 2 changes
Patrick_H_Lauke: 1 week in on 2-week period
… for wider group to review proposed changes
… there were a few that are quite substantial
… so I highlighted those in particular
… please review disambiguation of path-based gestures versus dragging movement
… PR #4843
<bbailey> w3c/
Patrick_H_Lauke: a few other potential ones for errata
… amending certain definitions, e.g. contrast ratio
… one more week to go
… next week is a bit tricky as some people are on holiday
… unless there’s pushback, we’ll merge next week
… please review & comment now
Rachael: does anyone have announcements?
… or is anyone new?
Conformance subgroup update https://docs.google.com/presentation/d/17t8pDzWZzRZS4B2hU3TLUXT6nBS5ILy1hLtMK4kk4sY/edit?slide=id.p#slide=id.p
Rachael: next up is conformance subgroup update
alastairc: I’ll share screen
… I’ll note that subgroup participants have weighed in
… everyone has access to full document, and I’ve hidden some slides for this presentation
… if clarification questions come up, please ask as we go
… there is quite a lot to this
… will try not to take all 2 hours
… subgroup started in January
… we had conformance model from previous work at TPAC 2024
… agreed to put in draft
… didn’t start from scratch, but did try to tackle as many conformance-related topics as possible
… setting the scene
… WCAG doesn’t require anyone to conform or claim conformance
… but for it to be a standard, there must be a conformance model
Slideset: https://
alastairc: conformance means meeting 100% of requirements, but can be split up by level
… conformance claim must clearly define the scope and date of the claim
… that can be different from reporting
… 2 different things to keep in mind
… reporting builds on conformance
… VPAT, ACR, etc. are more than conformance
… we don’t necessarily define that
… people can use any approach
… we could provide examples though
… also, to differentiate
… compliance is out of our scope
… compliance = comply with law; conformace = conform with technical standard
… compliance could be more or less flexible than conformance
… depending on the region
Charles: acknowledging that compliance is out of scope for w3c, compliance around the world currently points to 2.x
… is WCAG 3 intended to supercede 2.x?
… or will 2.x always persist
alastairc: from a w3c perspective, 2.x will persist
… unless we declared otherwise
<AWK> WCAG 1.0 was superseded 18 May 2021
<kevin> Example of a superseded recommendation: https://
alastairc: we would want WCAG 3 to be the next standard that people reference
… that doesn’t necessarily mean a replacement
… and there would be some kind of timeframe for that
<Ben_Tillyer> From https://
alastairc: we often talk about “ruler” & “rule”
… “ruler” = a measure; “rule” = law or regulation
… we have unit evaluation for each provision
… page-level, etc.
… the scope of conformance is something that needs to be included
<Zakim> bbailey, you wanted to discuss clarification that 3 would not have dependancies on 2 ?
bbailey: WCAG 3 isn’t going to have any dependencies on WCAG 2, correct?
<AWK> God forbid
alastairc: no, I can’t see that
<Detlev> I can't see the meeting link in the info sent out by Alastair...
alastairc: noting a few key issues
… one common issue is that products rarely fully meet WCAG conformance
… that it becomes difficult for comparison
… if you’re working in procurement and assessing different products
… there is also conflation of conformance with accessibility
… we often talk about things being “accessible” when we mean “conforming with the standard”
… we can be our own worst enemies in misusing that language
… depending on type of topic, WCAG 2 is a set of true/false statements that apply to an interface
… some standards use a process to determine which requirements should apply
… which is an attractive approach
<Detlev> is there any way to get to the Zoom meeting link (maybe send to fischer@dias.de)
alastairc: because it helps the person who’s trying to meet the standard by letting them consider their context, risks, etc.
… would take them through a process
… so want to draw a distinction between those types of standards
… we might take a step toward that, but can be a very different process
… a time-consuming process
… we tried to break down previous topics
… conformance challenges document
… and others
… we’ve tried to work through these to determine which could go into “core” conformance
… and what should go elsewhere
… discussed a policy document
… there is a full documented linked in this slide deck
GreggVan: a slide says some standards use process, but these kinds are never required by law
… you go through and decide which requirements apply
… different evaluators will come up with different results for ISO27001
… I think talking about it is misleading, and we should avoid that
… and not look to ISO27001 as a model to imitate
AWK: I completely agree
… so that people can pick & choose
… but
<bbailey> https://
AWK: I do think that’s not a requirement for a standard of that sort
… it’s a combination of assertions, testable criteria
… and whether someone is making an assertion can be a testable criterion itself
… we may need to be creative
… because of the problems we’ve had in 2.x
GreggVan: I agree with AWK, not saying we can’t have process
… but that standard _structure_ is not good to look at
alastairc: I’ll speed through the next topics fairly quickly
… “accessibility supported” we’ve had a separate discussion
… will be part of normative text
… “authoring tools”
… not trying to recreate that in WCAG 3
… but where an interface allows a user to create content
… that should meet WCAG 3
… it should enable _that content_ to also meet WCAG 3
… where end users are authoring content for other end users
… need to scope carefully
… “equity evidence”
… a concept around us making sure that we are equitable around disabilities
… want to continue tagging provisions with functional needs categories
… but shouldn’t have any scoring mechanism as part of conformance
… a lot of difficulties and potential for gaming
… doesn’t mean that other people couldn’t take that on in *reporting*
<Charles> note: while a charter is not required, the ATAG CG is currently in the process of writing a draft charter
alastairc: “reporting”
… we would including minimum info for an optional conformance claim
Glenda: on the equity evidence
… does that mean that there’s movement away from any summary by functional need?
alastairc: that would intersect with scoring discussion
Glenda: from usability perspective, reading a VPAT that’s 50+ pages long and not having any summary
… that shows types of issues most impacting *this* disability type and *that* type
… is a disservice
… just need a summary
<Zakim> Rachael, you wanted to say an answer to Glenda
Rachael: if we continue to tag it, the reporting option you describe will be possible
… even if we don’t provide any sort of scoring approach
<Zakim> bbailey, you wanted to ask if we are still considering scoring outside of functional needs?
bbailey: scoring generally isn’t off the table, correct?
alastairc: please hang tight :-)
GreggVan: the problem is it gets complicated very quickly
… most provisions affect multiple disablities
… some more than others, sometimes
… to do that summary, one would need to have a complicated formula
… which provisions are more or less important for this disability, etc.
… zooming is irrelevant for a blind user, for example
… and then you’d need to weight each provision, etc.
… it gets complex
GreggVan: I’ve tried in the past, and failed
<Zakim> kevin, you wanted to react to GreggVan
<Zakim> kevin, you wanted to highlight need for clarification questions
kevin: bring us back to point of order
… clarifying questions only, please
… please save substantive questions for end
Jon_Avila: equity evidence may not be included for scoring, but does that mean we’re looking for other ways to include evidence?
alastairc: in terms of evidence, we should be looking at how we’re addressing different functional needs equitably
… but not presenting as part of conformance
<Charles> we should avoid determining what method is relevant for a functional need. a blind user may magnify a page to result in collapsing a global navigation.
alastairc: the bit that took the most discussion
… was setting scope
… we have conformance units (e.g., page, view, process)
… there isn’t a clear path to requiring particular pages/views/process across the variety of sites/products in scope of WCAG 3
… if we let people choose “this path is most important”
… it gets us into sampling side of things
… a lot of discussion was around if we go with that approach
… it’s not really talking about conformance of the whole
… component libraries
… the idea of claiming conformance for a component in a library
… doesn’t necessarily make sense
… because when it’s *used*, new failures can emerge
… different variables, attributes, contexts, etc.
… we’re not sure what conformance for a component library would actually say
… didn’t see a normative way to include that
… paths
<Charles> note: a component library could also be the entire product.
alastairc: e.g., processes, tasks, workflows
… people generally see as a way of prioritizing
… in practice, it *is* a good way of prioritizing
… the sales flow in an e-commerce site, for example
… but we get into that sampling problem
… difficult to apply across a variety of products and websites
… there may not be consistently definable paths
… how can you claim the site conforms when some paths fail
… difficult to pull into conformance
… 3rd party content was another topic
… things you might procure
… video player, comment service, news feeds, etc.
… that you pull into your product or site
… the crux of this problem is we’re trying to be the “ruler”
… the source of that isn’t important to the measure
… this will probably be a good topic for the policy document
… another topic was UAAG — user agent accessibility guidelines
… we do think we’re covering this one well
… so that as user agents do more, authors can do less
… bugs
… in conformance challenges
… in any size of product, there will be bugs
… in WCAG 2, any bug is a problem
… there are things that policy makers can integrate or do to make things reasonable
… but it’s not something we can build into conformance model
… large volumes of content was another topic
… might be possible to include some sort of assertion for this
… e.g., “we have a process for archiving content”
… something you’re showing as a process improvement
… and then user-generated content
… if you have end users who are adding content to site
… certain provisions for authoring methods
… we could have a “partial conformance” type claim for that content
… haven’t fully fleshed out
… but seems like a reasonable approach
… live content
… can be quite problematic
… multimedia especially
… need to get to a feasible set of requirements
… if something is impossible for live content, need an exception or scoping statement
… another topic was “not applicable to all content / art / games content”
… rarely used content, experimental content
… policy issue
… small business challenges
… could potentially have filters
… which show provisions which are applicable, assuming we can identify them
… mostly a policy thing
… questions or follow-ups?
stevekerr: you’d mentioned with 3rd party content
… broken into two
… things you’d procure was the first
… what was the second?
alastairc: user-generated content
… creating content on your system
GreggVan: 2 things
… we should actually be talking about 3 types of 3rd party content
<hdv> -1 to calling 'procured content' third party content… if you procure part of your site, you did that, it should remain your responsibility
GreggVan: 1) procured, 2) user-authored, 3) viewed on but is not part of site
<bbailey> +1 to @hdv comment
GreggVan: #3 is sticky
… if the site pays for people
… to put their videos on the site, for example
… if you rewarded them, then the value of the site is the stuff you’re rewarding people to put there
… #3 I recommend we only put into policy
… we talk about what is and isn’t accessible, but let policy decide
… my second point
… sampling
<bbailey> I am all for exploring distinctly different categories of “3rd-party content”.
GreggVan: very important
… has to do with paths also
… every standard uses sampling
… in its enforcement
… because when you make stuff
… you can’t test every single thing you can make
… some tests are destructive
… so sampling has nothing to do with conformance
… it has to do with testing
… it only works if samples are representative
… which means you have to randomly sample them
… and every time you test you have to take a *different* sample
<laura> +1 to Gregg. Procured content needs to conform to the standard because it operates as part of the user experience. Any exception could open the door to selectively outsourcing noncompliant content to evade compliance altogether.
GreggVan: you can also do structured sampling
Charles: for component libraries
… we need specificity of language
… there are libraries that expose that library publicly
… for review
… and there are libraries that expose the library publicly because that‘s the whole point
… differentiate between internal and public component libraries
AWK: speaking to comments about sampling
… make folks aware of what is the divide that we‘re trying to navigate
… effectively, no one is able to demonstrate that they fully conform with WCAG
… and this is currently required by law in some places
<hdv> note, in a few places sampling is the thing that the law says to do
AWK: while what GreggVan says make sense, if we left it entirely up to governments and orgs to identify what sampling they’re willing to accept for *compliance*
… we’ll end up with very different results around the world
… while 2.x AA has been the standard adopted in regulations around the world
… we have an opportunity for consistency
… so if we identify some mechanism for sampling, that may be worth our while
… is there some risk for gaming? yes
… but
<bbailey> +1 to Charles point about component libraries — not all are open source, and use of some component libraries are not transparent
AWK: with a combination of sampling and assertions, there’s opportunity for penalties
<laura> Sampling is about practicality in testing, not permission to exclude content from conformance. Every part of the site should meet the standard—sampling is simply the method we use to confirm that reality.
AWK: for people who assert things that are false
<Jennie_Delisi> +1 to Andrew.
<Zakim> kevin, you wanted to comment on sampling
AWK: so I do think sampling has value
<hdv> +1 to sampling has value
kevin: I think conformance cannot be claimed for a website that has not been tested
… if you’re claiming conformance on a small sample
… then there is content in there that may or may not meet standard
… from a compliance perspective, I totally agree that sampling is absolutely essential
… and also essential that we provide guidance
… on good approaches
… and I think WCAG-EM does that exceptionally well
<Zakim> alastairc, you wanted to comment on sampling where robust testing is very different from efficient improvement
kevin: and has been incorporated into some legislation
alastairc: chair hat off
<bbailey> +1 to AWK points
alastairc: the kind of sampling I’d do with a client when we’re systematically trying to improve the accessibility of their site
… in order to get everything to be improved
… but I do take the point that if you leave that to a site owner, they’ll pick all the same pages and forget about the rest
… just wanted to point out different approaches
GreggVan: would be useful to tag bullets in these slides to answer the question of “what goes where”
… some of these may not be belong in conformance, but are very valuable
… we cannot use the word “sample”
… unless we mean it to be a “sample”
… and a “sample” is defined as a subset that is intended to show what the whole is like
… it is representative
… you must take random samples
… it’s a question of probability
… we shouldn’t be redefining or building some different version of sampling in our conformance model
Jennie_Delisi: this gets to heart of major conversations among conformance subgroup
… if you think about the health of a car
<bbailey> +1 that i think when we talk about “sample” we mean “representative sample” — but “random sample” is a different thing
Jennie_Delisi: we’re using a dashboard to assess the health of the whole car
… depends on definition of terms
… it’s like saying there’s one thing wrong with the car means that the whole car won’t work
<shadi> +1 to Jennie
<bbailey> +1 to Jennie_Delisi
Jennie_Delisi: this is where we really need to get agreement from the group
… so that conformance can get a solid definition going forward
stevekerr: on user-generated content
<Charles> i perceive the car analogy as relative to non-interference. it depends on which 1 thing is broken.
stevekerr: if we can acknowledge that a lot of products with user-generated content can never reach full conformance
… dynamic interfaces, social media, etc.
… a conformance claim can be invalidated easily
… the claim could be obsolete in moments
… it seems like it would not be possible for many products to reach full conformance
… a lot of regulators are trying to put full conformance as their requirement
… that’s where my concern is
… if you have one of these products
… and do your best in good faith
<bbailey> +1 to stevekerr that there are tiers of “user-generated content”
stevekerr: it doesn’t really help any procurers when evaluating one product against another
… even if we fulfill ATAG methods, you can claim“partial conformance”, but that’s not enough
… if you were a procurer, you would have a bit more visibility to what the product is capable of
… instead of hiding in policy document
… would be good to integrate some of the ATAG requirements into WCAG 3
<Zakim> alastairc, you wanted to comment on which things aren't going into conformance and to also say "random" is required in the definition of sample
<Gez> Sorry, I'm unable to scribe
AC: Don't think that definition of sample means it must be random
… fairly standard practice to pick a good representative sample
… we should define the sampling type
Detlev: WCAG EM was updated and uses sampling of different types
GV: [discussing sampling types]
… contrasts testing a sample vs a "showcase"
… feels that randomness is essential to the sampling
… User Generated content came up again
… needs to be evaluated with and without it
<alastairc> I think some of this comes from the (unspoken) concept of "site conformance". As kevin said, if it's nottested you can't say it conforms, but you're still trying to provide *something*
GV: e.g. mail program web app with and without email messages from 3rd parties
<hdv> may I offer a link to WCAG-EM's sampling methodology https://
GV: also, people posting to a site need it to be possible to post accessibly
<Zakim> kevin, you wanted to mention ATAG work and non-conformant user generated content
<kevin> ATAG Community Group
Kevin: Couple of things
… attention to ATAG CG - publishing a draft there - take a look
… Re User Gen Content. Trouble with this is that purely generated content can create a barrier.
… think that UCG can be dealt with in policy document
<Zakim> bbailey, you wanted to discuss thinking out loud if we dare assume good faith users of wcag3 ?
BB: Thinking out loud
… we seem to be spending a lot of time around "bad actors". Can we assume good faith implementers?
<Zakim> alastairc, you wanted to comment on difference between end-result being accessible, and whether the tool allows for accessible content, it's a naming thing around "partial"
AC: On UGC
… draw distinction between end results being accessible and tool support
… maybe some naming "authoring tool conformance" or subset?
… if subset is met then the tool has done its job
<bbailey> Focusing on good faith actors means dropping regulators and litigation from our target audience.
GV: what does BB mean have WCAG 2 as backstop?
<bbailey> No, not requiring both!
GV: we wouldn't require both
… Kevin talked about ATAG. We shouldn't encroach on auth tool turf so we should incorporate all or none
<bbailey> I meant that until we deprecated WCAG 2.x the way we have with WCAG 1, WCAG 2.x for regulators and litigation remains a thing.
<hdv> +1, I know of multiple authoring tool vendors that use WCAG to measure how accessible their authoring tools are
GV: If we start saying authors are responsible for UGC beyond making it possible, where do we draw the line
<kevin> Responsibility of posted content is not our issue, that is something for regulators and policymakers to decide
AC: Thanks for comments.
… want to give more info about different proposals (there were 6, and now a combined 7th)
… (reviewing slide content)
… proposals brought up key discussion points
… sampling was one
… unresolved is "random/ non random" bit, suitability for regs
… Paths/workflows was another key point
… how to identify which ones is a challenge. We could include assertions to define a process for selecting sampling
… usability testing was another key point
… consistency of results an issue
… inconsistency makes it challenging for use in conformance
<Zakim> GN, you wanted to ask whether the usability tests are meant to be inclusive
GN: Are usability tests meant to be inclusive of users with different disabilities?
AC: Yes.
AC: Automated testing was another key point
… problems are that it tests what is easy to test not necessarily what is most important
… but poses as first-level of highly testable items
… Conformance score another key point
… a way to show progress
… score could be based on a number of factors
… also discussed number of levels.
… more levels increases chances of regulations varying in what is selected
SAZ: similar to Bruce - there is no evidence for negative phrasing - possibly if there are more levels that companies would be incentivized to do more in response
… countries all choose Level AA, but could choose Level A.
<kevin> +1 to Shadi's positive reading
GV: Need to not confuse accessibility with marketing
… Not convinced that web site owners are paying attention to competitors
<Ben_Tillyer> Completely disagree with GreggVan on this. Accessibility is becoming more and more of a competitive advantage, not only in niche firms, but FTSE/S+P firms.
GV: If there are multiple levels and one country requires level 9 then that becomes the level for an org that sells into countries with level 5/7/9
… concerned that too many levels may create chaos
GN: @@@
HDV: problem we as a regulator see is that orgs want to indicate progress when they are not YET conformant, so the bit before they reach what is required. This isn't a bit that we control in AGWG, it is decided by regulators (and as a regulator we're actually looking at how to make this better)
… I don't think we need levels of 'being conformant' for this, we need ways to show progress in what is 'not yet conformant' (and maybe regulators can do that rather than AGWG)
<Zakim> bbailey, you wanted to express appreciation for slide 30 conformance progression scoring
BB: On slide 30 - scoring. This is how people are using different eval tools
… Is 75% failing or good? Would be good to clarify
GV: We don't have any basis for saying that it is ok for PWD to only access one or two paths through a site
… ok if stairs and ramps because there is an accessible way.
… but that isn't how people are talking about using paths
… also concerned about user testing being useful for conformance
<Zakim> AWK, you wanted to talk about paths
<shadi> +1 to AWK
AWK: Not going to say that requiring paths is going to make it so that the whole site is accessible. Remembering someone saying "don't let perfect be the enemy of the good". If we do sampling / paths, based on log-data and priorities, these are important. If they include those or not, worry we're throwing away something useful/important.
… it is a trade-off, but it's a big challenge
GV: Paths is useful in "where to start document"
<GN015> +1 to Andrew
GV: if you have [large site] what to do first, or most easily, etc
… that is where I think paths should be
… people need to think about paths specific for PWD also
… don't think that any country could adopt a standard if it said that only some parts need to be accessible
<Zakim> alastairc, you wanted to comment on paths and the in-practice
<GN015> The conformance claim then has exactly these paths in scope, right?
AC: Interesting problem
… select a bunch of URLs without guidance -- not a stretch to say "pick your pages but make sure that it includes a/b/c"
… we do want WCAG 3 to be regulator-friendly, among other things
<Charles> +1 to perfect is the enemy of good. however, there still must be a minimum threshold for good. the WCAG 2.x concept of perfect is 100% conformance with all layers of guidance AND going beyond them.
AC: last key point - dynamic adjustment of weighting/
… more to dig into here
<stevekerr> It is not a requirement to list out the URLs or screens used in the VPAT. You can list the methods you used in the Evaluation Methods Used section, but it is not a requirement to list specific methods.
AC: conformance group will continue work
<Rachael> acribe+
AWK: We are going to be talking with an attorney tomorrow about some of the regulation impact within the conformance subgroup. Reminder to send meeting invite.
SAZ: how is external input documented?
AC: in first meeting of conformance subgroup we link to Q&A with regulator-adjacent ppl
SAZ: You were saying you go back to these people?
AC: We can. More input on sampling wo0uld be good
HDV: more documentation would be good, for the AB discussion
… not sure if confidential or not
Kevin: I have spoken to a number of regulators that we are connected with
… to provide background info
… not sure he can share who he spoke with
… if there is a strong desire to know who I can check with them
GV: Results?
Kevin: shared in meeting according to AC
AWK: I do think it is valuable for us to know what role or position is providing comments. If we haven't spoken with someone from a government, they may want to know. Its worth being as transparent as possible.
GV: could identify people without tying them to specific comments
GN: separate question. Is there still an onboarding meeting next Tuesday?
RM: 30 min before meeting
GV: I think that there is a rule in standards that requires identification.
SAZ: might be worth double-checking on this. I see full log of responses
… is ther a w3C policy on this?
GV: Need to be clear whether a comment is from a person who works at some org vs an official statement from that org
<shadi> +1 to consider that as individuals opinions from people in the field
HDV: often discussed on W3C AC meetings
Kevin: To be clear, I have not used the info in a way that suggests that they are speaking for their org.
Github discussions: https://
<hdv> and fwiw I am personally very glad these convos happend
<hdv> I see it as very helpful to the work we're doing
RM: github links above - get on them
<Ben_Tillyer> Thanks all