Meeting minutes
Retrospective
introductions round the room...
chuck: taking up retrospective. This will be our second after the first one at TPAC.
… the purpose is to reflect on the past to improve the future. What went well, what went poorly. This is a no fault process. Remember the code of conduct.
<Detlev> the only mic of people in the room that shows activity is Alastair's - I often use that to see who's speaking
<Chuck> https://
chuck: issue 1: regarding pronouns and corrections. There are skills and techniques available to make this a productive experience. Correcting in the moment can sometimes create an awkward situation. Chuck has provided a link to a guide with suggested behavior around mistakes.
<JenStrickland> I likely missed an opportunity to volunteer, since I've been sick for the past six months. Now you have at least one! :)
chuck: issue 2 the last time we met we discussed an onboarding buddy system. We now have a monthly onboarding session, we thank our volunteers for that. We are still looking for volunteers for the buddy system. If anyone is interested, please make yourself known.
chuck: issue 3 we have made progress on subgroups. We encourage groups to meet async as well.
chuck: let's start with the positives
greg: I thought breaking into groups and having continuity over time was helpful.
Dan: I remain happy to move to github and google docs.
Greg: seconds that.
<Detlev> why doesn't zoom show the current speaker?
Jennifer: I have appreciated the chair's efforts to tame chaos, and willingness to listen. My disability can sometimes make me reactive, I appreciate that you hear me and know that I mean well. I have appreciated that you listen and keep your cool.
Jenny: I want to say thank you for the ability to participate async.
<jspellman> +1 to it working well
Rachael: I like this format we are using of three phases of discussion
<dan_bjorge> +1
John k: I appreciate the incorporation of zoom and breakout rooms.
Chuck: its nice to hear all of these, we can leverage these tools to keep the momentum going.
… now moving to what went poorly. Reminder, this is blameless.
Steve F: We start a topic and move on before it is complete.
Steve: An example of this was "views". That feels unresloved.
GreggVan: We often go off to think about a problem but we need a summary doc to make sure that when we come back to the issue we know where we were, what points were made.
kirkwood: second that, we can also use the technology we use every day to improve this, for example AI to summarize conversations.
Dan: We are still doing manual scribing. This is frustrating. The scribe cannot focus on the meeting and scribing at the same time.
<Zakim> Chuck, you wanted to mention a positive from Coga work
Chuck: adding a positive. Documenting and recording the work we have done. My subgroup benefitted from the recorded work of the COGA group
<kirkwood> the scribe should approve ai minures. - to process
GregVan: I would like to propose that we have AI scribe the next few meetings and see how it goes. I've had cases where the scribe recorded the opposite of what I said.
Ken: we need to test this on hybrid meetings
<Zakim> hdv, you wanted to respond to use of AI tools suggestions
<kirkwood> +1
<kenneth> +1 hdv
<Zakim> Chuck, you wanted to discuss the ai generated scribing solution
hdv: I feel like accuracy of intentions are important, the work is very subtle. AI is not good at that. AIs are also not good ad transcribing non-native English speakers.
<jspellman> I found AI to be useful in finding prior WCAG3 research and proposals. :)
chuck: there are certain W3C policies where AI scribing may cause issues. There were some folks in W3C who have been exploring the use of AI.
<Jennie_Delisi> For future AI considerations: decisions should be considered re if AG meeting information is training the model, where information is stored, and whether the AI tooling will continue to show up in meetings (must have the ability to not have it come to all meetings).
Jennifer: as someone who leans heavily on the transcript, using the AI generated transcript may help folks understand the meeting better.
<Zakim> tink, you wanted to respond to AI topic
<Zakim> kevin, you wanted to highlight difference between transcript and minutes
Leonie: The AI tools may not know who said what. Also there is a need to make off-minutes comments, this might restrict the comfort to speak freely
Kevin: +1 to Leonie. We need to make a distinction between the transcript and recorded minutes.
Jennifer: Wendy has mentioned that there are efforts towards this in W3C
Chuck: when we were starting the help subgroup there were challenges with participation, that slowed us down.
JenStrickland: many members of have been in AGWG for a long time, there has been a gap in awareness in some types of disability. I raised an issue where two members with hearing and cognitive were having issues participating in sub groups and was told that we cannot make accommodations. We need more awareness about disability in our groups. We need
to learn from our mistakes.
Dan: I remain concerned about cases where we still have open research requirements. There has been talk about how we will fund and schedule research. But I'm concerned that we are not moving fast enough. For ex: Color contrast algos. We are getting into a place in the subgroup where we need this research to move forward.
Rachael: My understanding is that we can get interpreters for sign language for participation
Kevin: yes we have a contract with Sign language interpreters.
<Zakim> Rachael, you wanted to react to kevin
Rachael: This is not to negate that there was a failure of communication
<Zakim> alastairc, you wanted to comment on the research question
<Zakim> JenStrickland, you wanted to clarify, it wasn't about sign language interpreters - if that's referring to my previous comment.
Rachael: we are working on recruiting research
JenStrickland: The issue I was referencing had nothing to with an ASL issue. This is about clashing of cognitive overload. We can design the experience to improve this.
Chuck: We need to build up empathy towards intersectional disability.
JenStrickland: I think this exercise to develop empathy will benefit the standards.
Kevin: We are going to have a need for good reviews going forward. This has been a challenge in the past. Reviewing is hard, very detail oriented. I'd like to suggest we start thinking about how to do this well.
Jennie: Just thought of another positive, thanks to chairs for allowing us to participate remotely. Thanks for that.
<kenneth> big +1 to continuing to enable remote participation
<tiffanyburtin> +1 remote as well
<Zakim> alastairc, you wanted to comment on reviewing things
Chuck: Now we'll discuss what we can we improve.
alastair: we have a big group. We need to figure out how to ensure issues get the right number of reviewers, not too many, not too few. Perhaps people can volunteer for topical areas..
<jspellman> +1 to Alastair’s ideas of review
<dan_bjorge> +1 to alastair (the specific mechanism isn't very important, +1 to *some* mechanism to assign responsibility for reviews)
Chuck: one of things that appeals to me is the idea of building empathy. I wonder if there is an opportunity for those who are an expert in intersectional disability can present to the group to help us better understand and prevent future failures.
<jspellman> Adding to Alastair’s ideas, perhaps subgroups with a related interest could be assigned to review and devote a meeting or part of a meeting to reviewing and discussing.
Rachael: Has a question to Steve about the views transition. The chairs were realizing that there is huge ramp up to time to resume a conversation. Do you think the pathways approach will work?
Steve: I don't yet have enough info to say. Filing a PR is a way I try to focus discussion
<JenStrickland> Note: if we have any experts or people with lived experience provide a presentation to the group, research if there are ways to compensate them for the time and energy it will add to their plate to do so. They're already generally juggling a lot of workarounds to participate. Just a note to consider.
alastair: If things like views move out of the group conversation, people can keep working on them.
<Zakim> Chuck, you wanted to ask for a scribe change
Steve: What is important is that stuff gets done.
Chuck: Are there any other ways that might help us document our work? Beyond AI that is.
Jennie_Delisi: Just a reminder, sometimes some of the document formats cannot be used in some places of employment.
… For example, I can't access Google docs from my work computer
<JenStrickland> +1 to Jennie_Delisi
Jennie_Delisi: This contributes to the consistency with which I can contribute
dan_bjorge: On other options for improving scribing. Is there a world where there is budget to hire a human to do scribing?
alastairc: On the topic of reviews we do have processes to cover this but there is scope for everyone involved to consider what more could be done.
Chuck: Having a retrospective is a good opportunity to explore how we improve how we do this work.
… I like having 6 monthly retros. More frequently can tie up time.
… There may be ways we can discuss ways to improve without having to wait for a retro though.
… If there is anything specific, it would be great to hear about it
Chuck: Are there any other items from any of the 4Ps that we have been looking into?
<Zakim> alastairc, you wanted to comment on participation and how sub-groups make progress.
alastairc: Just on participation; lining up meetings across timezones can be tricky when forming subgroups.
… I don't know that there is a solution to this as it is timezone and knowledge dependant.
… Open to ideas!
dan_bjorge: There are tools to find coordinated timezones although I am not sure on accessibility of any of them.
… A separate suggestion is related to done criteria. I think we need to be intentional about defining early what the 'done' looks like early in the work.
… A common failure pattern is to decide on a number of actions to do and not be clear on what complete looks like.
<Zakim> jspellman, you wanted to say using AI for research on older work - continuity of subgroups
jspellman: Building on AI and Steve's point; I had success using AI to research past work done in the group.
… Being able to find early subgroup work and build on that was extremely helpful.
… I would be happy to write up this approach.
shadi: I find it helpful when the Chairs share the big picture now and again. It is easy to get lost in the strands of things.
Rachael: How often do you find this helpful?
shadi: At least quarterly. Doesn't need to be long or belaboured. Just the story so far.
GreggVan: Off of Shadi's, whenever there is a pivot there should be an update on the big picture.
… This might be important enough to send out as a specific note to the Group.
… Also, if you do it bimonthly it might help clarify when pivots are happening!
… Separately on institutional memories. We should look to formalize some of this material and share more broadly.
… Document and sharing any work that is done to delve into this is important.
… Particularly when decisions are made and being clear on the rationale
tiffanyburtin: Just wanted to being a gentle reminder that for certain disabilities speaking clearly and with volume will help the transcription.
<Chuck> ack
GreggVan: There are only two mics here. Would be good if there were many more mics here.
Kevin: W3C has gear. For meetings like this, where don't have tech folks, we don't get all the gear. It's hard to get it moved around the globe. For small meetings we need to come up with better ways.
Kevin: If you have pointers and ideas let us know.
shadi: On this topic, just to mention the ask for help, having a host for the meeting can be more beneficial for making accommodations available.
Rachael: This is a pivot. Because TPAC is later this year and because of potential cost benefits we are looking to have an in-person meeting at AccessU instead of CSUN.
<Zakim> tink, you wanted to comment on meeting locations
tink: Regarding meeting locations, the W3C is giving some consideration to not meeting in the US due to the current administrations attitude towards disability.
Chuck: Moving on to actions for now and later
… Just to emphases the Chairs are keen to not own all actions!
<jspellman> +1
Chuck: Jeanne will be looking to capture information on how she researched past item.
<Jennie_Delisi> jspellman - if you are looking for someone to help identify business requirements for that AI project, I would be open to a conversation.
Chuck: Another action would be to ask Wendy to update us on approach to AI as part of the scribing solutions
… Rachael to pick up an action to do a high level overview every two-three months.
… Not sure how we look at how to engaged with people who have intersectional disabilities. Chairs will pick this up for the moment
<Zakim> Rachael, you wanted to ask about measuring
Rachael: Regarding measuring there are two questions about this on retrospective: did we do a thing and did the thing work.
dan_bjorge: I would suggest that we assign dates to action items to help keep these moving.
kevin: Thinking about managing the actions, we lost the issue tracker, we have issues in the WCAG 3 repository. We need a method to track AGWG actions. Then we can look at them discretely and track the appropriately.
alastairc: There were potential options for better scheduling subgroups. If anyone in the group has ideas on this, please send them through.
… There was a question on using AI to summarize key documents.
… There is also a question on where we maintain design decisions and their history.
Jennie_Delisi: I hear a theme emerging that sounds like a less tech focused project management system. So an assignment could move between groups,
… a way to store information about topics, etc.
… Important to consider the tooling for this to ensure everyone can access and use it.
<Zakim> jspellman, you wanted to add to Jenni
jspellman: Just to add to Jennie's point, the issue may be more that we have properly linked all the different work.
… The hardest part when I was researching was restricting to the right set of assets.
<kirkwood> +1 to Jeanne
<Zakim> Rachael, you wanted to suggest using discussionsa
Rachael: When we have discussions in GitHub we are providing a summary. Linking in the Google docs would help to ensure the background was there.
… This could help cover core issues. Guidelines work should ideally end in a PR.
… This would keep everythign in Github.
hdv: To Jeanne, would be interesting to see the results of the work that you did
… Particularly around where changes in terminology occurred
<jspellman> Agreed, I was researching Task Completion, and I got so many false hits! :).
hdv: Also to note, this can all be done without AI as most material is public, there is existing knowledge in the group, and searching for material can be good
<Jennie_Delisi> * Novel idea - talking to humans!
<kenneth> +1 to using direct searches when available
<jspellman> Happy to share my different results as I refined the prompt
GreggVan: Whenever you tap some persons memory, even my "perfect" memory, make sure, with permission, it is posted to appropriate lists.
… Other people may be able to add to this information
<jspellman> Hidde, the big advantage to AI is that it read a LOT of documents for me! I had a memory of the project and could fairly quickly look at what I wanted
<hdv> Understandably, but don't forget you and Gregg have a lot of memory to rely on for factchecking
<hdv> (^ in response to jspellman)
<jspellman> @hdv, that is true, but being able to filter hundreds of documents down to a small number is surprisingly valuable. It took a job that I had been very discouraged about and made it doable. I didn’t rely on the summaries for anything other than whether it look relevant.
View Definition
alastairc: This section has two parts:
… 1. what has been discussed so far. Please do think about the problems and solutions
<JenStrickland> I volunteer to be the next scribe after the current one.
alastairc: 2. We have a bunch of different examples to test what we come up with
… Goal was to develop a definition that works with WCAG3 but also any other context such as ICT and Mobile
… WCAG2 defines this based on URLs but this has problems for eBooks and many other technologies
<jspellman> Link to the document being displayed, please?
alastairc: The sub-group came to the definition:
alastairc: "View: Conformance scope that includes all content visually and programmatically available at a given time, with the same purpose or for the same task."
… There are problems with this
… "Purpose" and "Task" are difficult to define. For example, adverts in a page.
… In practice many people are working around the page definition at the moment
… General agreement that this ia a conformance-scope thing.
… Unsure if we need a sub-view definition.
… Terminology difficulties with view (Drupal, iOS & others), we can call it a “conformance unit” for now.
… What about audio interfaces?
… An alternative view is to change to considering the view by reference to the context
… "View unit: “Conformance scope that includes all content visually and programmatically available without a change of context”
… Change of context: “Major change in the content of the view-unit that, if made without user awareness, can disorient users who are not able to perceive the entire view-unit simultaneously."
alastairc: That is the history, any comments on how we have got here?
GreggVan: It was interesting about when we talk about this previously we talked about a UI context
… Any time you think of 'view' then audio becomes differently
… UI context is less problematic to define than 'context' on it's own
maryjom: Another term that we explored was 'interaction context'
… There was considerable time exploring this terminology
alastairc: A key part of this is major change in the content of the unit that is being assessed
<Zakim> Chuck, you wanted to change scribes
<Zakim> tink, you wanted to ask about "conformance unit" outside of change of context
Tink: I'm curious about needing a conformance unit outside of a context scenario.
<Jennie_Delisi> * audio is quiet for Leonie and Dan
dan_bjorge Suggesting a test case the definition brings to mind. Pages that have a dynamic content below.
<Zakim> alastairc, you wanted to comment on why have a conformance unit
A Google doc where a user is typing into, does that become a different unit, every time a user changes something?
Above was dan_bjorge
alastairc: Regarding if we need a unit, … guidelines that say whether keyboard functionality works, then pack into page, functionality in context with keyboard but not be a button.
<Detlev> @Leonie a component in isolation probably would not be a conformance target
@tink: my question was different. there's no notion usually of a component in isolation triggering a change of context. where … can we use this idea of a conformance unit or just limit it to units in which a change of context is applicable?
Kevin: does this become another example we need to test against?
alastairc: if we are assessing for the component library, there will be lots of guidelines that don't rely on this. they would be essentially component level guidelines. Some components need the bigger view to asses.
<kirkwood> Therefore, a change of view is a change of context by definition?
<alastairc> kirkwood - yes, in this proposal
Rachael: coming at this with four different levels of conformance in wcag3, the most basic might be a button. then there's another level where it might scope to the view. that could be the entire page, or it could be a portion of it.
… remembering that context is helpful.
dan_bjorge: one thing I noticed about change of context is that it explicitly sets a change of focus. I don't think we want every change of focus to be a conformance unit.
Detlev: If I understood correctly, the v view would have the aim of capturing the unit of conformance, depending on the implementation and the context on the view/page
maryjom: I don't know if folks misheard, but it was not a change of context, it was a user interface context or an interaction context.
dan_bjorge: I hope it's okay to bring up now, another option I'd like folks to consider is closer to the current definition of page, surgically stripping out the part about being based on a URI over HDT. plus any other resources used int he rendering.
<Zakim> alastairc, you wanted to comment on claiming conformance and the different levels (product/task/view/component)
alastairc: could you paste that in, please?
<dan_bjorge> "a non-embedded resource plus any other resources that are used in the rendering or intended to be rendered together with it by a user agent. Example 1: An HTTP resource represented by a single URI"
alastairc: as it stands we have some guidelines that will start the scoping of the foundation level, some component level, and some product or task level… so that are the references the guidelines are making at the moment.
… in wcag2 we referred to a similar way with a series of UI contexts or whatever as a sample.
GreggVan: we have to be careful about component level… at a page level one has to evaluate all the components. if there's an inaccessible component, as long as it doesn't interfere or cause a seizure, it's okay.
… I'm not sure what we should be evaluating at a component level.
alastairc: at the moment we're dealing with that as we go along.
GreggVan: I'm not sure where we apply anything at a component level.
alastairc: for example in alt text it's each image.
@gree
: no, it's for each that's not decorative, or redundant with the test around it.
That was @gree
GreggVan, above.
hdv: usually with components in a design system, they are abstract versions but they depend on the individual usage.
<Zakim> Rachael, you wanted to speak to Gregg's point
Rachael: chair hat off, one of the goals for wcag3 is flexibility, handling at different scale.
… we centered conformance at the page level, and I wonder if we need to move away and potentially declare conformance at a scope of what can be tested.
@gree
GreggVan: Correct, at the scope of what I'm doing is the evaluation unit.
… if we get into scope, be real careful how we word it. Some people say our website is completely accessible up to X, but then you realize they meant up to Y and then the user can't check.
<Zakim> alastairc, you wanted to ask if that's our problem, and tangenting...
alastairc: this is a bit of a tangent, I'd like to come back to definition and the examples in those bins.
<Rachael> I've started a topic parking lot at the end of the slide deck
<Zakim> tink, you wanted to respond on scoping statements
tink: just a note on the idea of conformance statements, there should be some flexibility to state what the conformance unit was -- which is what I believe Rachael was saying.
GreggVan: so I think we're trying to say what is the evaluation unit.
<dan_bjorge> "a non-embedded resource plus any other resources that are used in the rendering or intended to be rendered together with it by a user agent. Example 1: An HTTP resource represented by a single URI"
alastairc: for each of these examples, how well does each work?
… first one has a bunch of tabs at the top with some content beneath, where the tabs change the content underneath.
… second is similar, but there's navigation at the top that moves between pages.
… third is both, nav at top, some content, set of tabs that change the underneath content.
… how well do those fit the terms.
shadi: the definition says unless the user or without awareness of the user or something, since the user actively initiates the change.
GreggVan: we're confusing unit of evaluation with that you shouldn't have change of context without the user doing something…
shadi: referring to something
GreggVan: I know that, I'm talking about the danger of using the word "context" instead of "ui-context" or something.
… where we're talking about a change of "view."
… if we go back to the 3 views, I think we were thinking of change of viewport.
… why are they different? for example, if you're blind your pages will all have a name and if it loads a different page, that name will change.
<kirkwood> within the context of ‘tabs’? would be a change, no?
<Zakim> Chuck, you wanted to ask which definition we are reviewing?
Chuck: which definition are we reviewing?
alastairc: we have two. Gregg's right, don't worry about the user awareness bit.
… major change in the content of the view unit, our ui-context.
… and whether resources are loaded from a particular place.
… the reason I provided this first one with tabs, it could be considered a change of context, since it is reloading the content.
<Zakim> kenneth, you wanted to ask how heavily this depends on implementation
Kenneth: devil's advocate here… I could see the first, second, or third be done through a full page reload or client-side requests.
… both of these things, regardless of what they visually look like would not be assessed as a change in context.
alastairc: we're trying to judge the definitions and how clearly they can be applied in all these different examples.
kirkwood: I would think tabs give context, if there are no tabs that would be change in context.
<Jon_Avila> It seems like we are talking about multiple definitions - what Gregg is saying is not was displayed on the slides related to context.
Jen: I think what Kenneth talked about, he's on to something. From what I'm hearing it doesn't sound like everyone understands. You are staying in what we call a "view", it all depends on how it stays together... html is content, css is style, js is behavior, we still need to think about these 3 separately, but together they create an
interface.
Jen: What you are trying to capture in this term, is the combination of those things coming together. That is something Rachael spoke to, trying to set scope to... Axe-core, when you scan the page, you can scan whole or part. Can we come up with what we are trying to do and then do a survey, maybe a card sort.
Jen: Because, what I'm hearing in this room, like definitions of "hot" may be the same, but one is Fahrenheit and one is Celsius. We need to go back to the archtypes.
Jen: Not certain we are making headway. Lots of content is in our heads.
GreggVan: change "context" to "ui-context"
hdv: when I see these example, I agree with Kenneth and JenStrickland, there could be different ways to split these into different units of conformance. For example, tabs may be fully visible in one screen or collapsed in another.
… this makes it complicated for our discussion. we could embrace that, different ways to split any kind of web ui today into difference units of conformance. how can we avoid that people might abuse this flexible system we have to define to match up to today's realities.
<Jon_Avila> If people can define their own units in their conformance claim then folks can choose what works.
<kirkwood> +1 to Gregg’s suggestion of “UI Context”
shadi: wondering if the change from one "view" to another… how much that impacts. what happens if I select a tab. if it's heavily templated, … is there a difference if it's tabs or navigation… if one has forms and the other has videos… does that impact what we call a "view"?
<Zakim> alastairc, you wanted to say we need to work a lot of this out before we go to the public.
shadi: I think it depends on what the content is.
alastairc: I think that would steer towards using something like a major change in content.
<Jon_Avila> I think we are talking about states of the user interface.
shadi: it's more if you go from one tab to another navigation. it's more what do I get when I select something as a user.
… as a user do I get a completely different behavior and interaction and it's a UI. I think I would call it a different "view", I don't see a difference between the three.
alastairc: talking about putting a survey out … I don't think we're there, yet. I worry we'll get all the answers under the sun, and we need something concrete instead to present.
<shadi> +1 to refining before asking for broader feedback from the public
alastairc: then to shadi's point the difference between 1 and 2 from a user's POV could be non-existent. I would rather go with a definition that isn't based on the technology.
… that's its a user experience pov instead of tech.
Rachael: looking at the alternative selection… where there may be a programmatic change that doesn't show up visually… I'm struggling with how we scope that. If visual and programmatic change both are required to define a "view."
Jen: Wejen we talk visual, that makes me "icky", we are trying to do this for everyone. Has everyone built a web page? In the past we had people who were more content oriented.
Jen: Are trying to pick the whole plate or the portion of the plate? Portions of a meal segregated on the plate.
Jen: is that what you are trying to scope within the term?
Alastair: It is the UI context you are perceiving. By the end we will want programatic content. In the end programatic follows.
Kevin: We are not limiting this to just a site perspective. All modalities of perception are in the scope of this concept.
<tiffanyburtin> +1 Jennie
Jennie_Delisi: if we look at the 3rd example, kirkwood, @rain, and I did a lot of work on behalf of COGA. We played a lot with these types of scenarios. if a person magnifies -- either through a tool or magnification -- some of the changes we're talking about could be outside the concept of that individual's view. we need to be cautious what would
qualify under that major change, given someone's view could be more limited than another's.
alastairc: if we were using a major change in the content, the "ui-context" then I think the first two would be equivalent and the third is arguable, but the tabs probably wouldn't create new… s
… referring to an example showing the Slack interface.
… showing potential groups, options, channels, and the messages.
shadi: I don't agree with the previous characterization, it depends on the content, what's in the tab.
… again to me, something like "add channel" is different functionality than switching channels.
… more the type of content than the amount.
<Jon_Avila> Some dialogs could a different UI context while others could be part of the same UI context.
shadi: if the type of content is significantly different than also the interaction of what I'm doing… the purpose is different… that weighs more than how much changes.
GreggVan: this example is the same as the last example, tabs with tabs, etc.
… if it's a change of context or a change of "ui-context"?
[scribe can't follow what's going on over there]
Would alastairc or GreggVan please consolidate your discussion into a summary in IRC?
<Zakim> JenStrickland, you wanted to ask shadi what the templating has to do with it?
<Detlev> AFK - have to attend to something
Jen: To ask Shadi about templating. If it's "templating". I don't know how that relates to "view/view-context".
Shadi: I should not have used "templating", as it is a technique. When you click a tab or go to a different page, it still shows the same page that does not need a mental model. I meant in terms of consistency that they all look and feel exactly the same.
Shadi: Individual channels in slack will have very different content, but the functionality is the same.
Shadi: When I click add channels, even though it appears as another tab, but I go to something else. "Template" is an incorrect word.
Jen: Where that fits with view unit... are we looking for a way, do we need this term to equally be able to describe 1 or 2 or more... as well as the collection?
Alastair: We are looking to agree on this definition.
Jen: I think that some are trying to define a small portion... maybe we can get some clarity.
Rachael: I'm going to post some definitions to help me clarify
Rachael: I suggest we talk about content layer, thing that user perceives and programmatic layer, the thing that test tools and AT perceives
<Zakim> alastairc, you wanted to ask Shadi whether he's thinking of Major change of context, or content? and to also run through the other examples quickly.
<JenStrickland> +1 to Rachael
Rachael: change in content layer doesn't always match programmatic layer, so I think we can't use it to scope definition
alastairc: re shadi's point… how you describe things, wonder if you were reading it as a change of context and not change of content?
<JenStrickland> I heard shadi refer to both change of content and change of ui-context
alastairc: re other examples… let's look at the others first, so that we don't miss things.
dan_bjorge: to respond to Rachael's point… when programmatic changes might not match content changes… I think we can just say 'a change in either counts' , not sure why it would cause problems
dan_bjorge: are we trying to determine what is the boundary at which someone would have to test, to test one thing,or are we trying to determine the boundary of what counts as one thing for the purposes of testing? They sound similar but are not the same, eg the channels in a Slack UI, are they to count as different things?
dan_bjorge: are we trying to say it's possible for someone to make a conformance claim to say 'my view channel has been tested'… does our definition emcompass every channel in this (Slack example) list?
<kevin> +1 to Dan's question for clarity on what we are trying to define
dan_bjorge: would be a big advantage to look at pages as most people most of the time have pages and it is the most consistent thing we have
dan_bjorge: the URL works for most cases
<Rachael> +1 to having a definition that allows for a URL but also provides alternatives.
alastairc: what we're aiming to do… we need this kind of definition for (a) keyboard requirements, in the context you're in, what would an alternative be, (b) conformance to say this, this and this has been tested and conforms
<Detlev> back
GreggVan: +1 to Rachael's separation to content layer and programmatic layer, that is very helpful
GreggVan: also +1 to Dan's question whether it is really important…I think in some cases it is, some it is not. Both visual presentation and the DOM are derived from the same thing
GreggVan: in patterns like tabs, in some cases the DOM is changed, in some cases it is not. So, it depends, it's hard for us to make rules
GreggVan: Rachael, do you have a suggestion what we could use instead?
GreggVan: for pages, we've always said pages should have names. If we use something other than URLs, how can you give that a name?
GreggVan: would we give different names every time there is a 'change'
GreggVan: ?
<Zakim> alastairc, you wanted to run through examples
GreggVan: I love the idea of UI contexts, but still can't figure out how to make it work
alastairc: [shows more examples in the slide deck]
alastairc: here's a panel with threaded comments, that can be a panel or can be popped out such that it is a new window
alastairc: and then there's this thing on iOS that can slide up from the bottom of the screen
alastairc: sometimes you can see content behind it, sometimes it is full screen
alastairc: then dialogues … in some cases the screen behind is blanked out and you cannot interact with the content behind it, some cases you cannot (modal vs non modal)
alastairc: then there are 'mega menus', that are sometimes 'semi modal', they kind of blank out the background but you can click the background
alastairc: there are also 'non mega menus', that users just hover over
alastairc: there are also side panels, like the 'themes' panel on Google Slides, that one is related to the content as you change what it looks like
alastairc: but there's also a maps panel
alastairc: what are the attributes we're looking at?
alastairc: I think, chair hat of, we can use the page definition and kind of fudge it to include everything you could 'pull up' as long as it is not completely separate content
<Zakim> kevin, you wanted to comment on unique identification vs communicating change
kevin: the URL is a useful way to identify the thing you're testing, but we're actually moving away from it, as there's no unique resource identifier for any of the examples alastairc just joined
kevin: do we need a unique way to identify these things?
<JenStrickland> Although it could have a unique URI, if it's to an #anchor on the HTML rendered document.
<Zakim> Chuck, you wanted to suggest how we procede
Chuck: we'll continue this after lunch, which will happen after current queue, which is now closed
GreggVan: you said URIs can't apply to parts, we do have targets
kevin: yes but won't define a boundary
<JenStrickland> The #anchor typically applies to an HTML element that is a container within the "page."
GreggVan: parts of what we think of as a web page have a URL that could be given a name
<Zakim> kenneth, you wanted to unique identifiers
kenneth: looking at the examples, like any time you're changing a tab in Slack, you're changing a URL and have a unique ID; there are existing patterns e.g. history.pushState and the popstate event, which also enable back button functionality
<Zakim> hdv, you wanted to discuss identifiers
<Jon_Avila> Examples might be a screenshot of an app.
hdv: It is very common for people who do conformance assessments of doing an assessment on something that does not have a URI. They can explain what was looked at.
<jspellman> +1 Hidde that is a common practice
hdv: It's not a beneficial as a uri, but it is very common.
<kevin> +1 totally agree hdv
hdv: I know that from personal experience, in spa that don't have uri's, some of this may be a theoretical problem. I don't feel that they are the big distinguishing feature.
<Zakim> alastairc, you wanted to add examples where page=URL doesn't work: ePub, PDF, kiosk, SPAs,
alastairc: URL doesn't work for epub and pdf either, one URL represents multiple pages, there are certain guidelines that didn't work (there's a long email thread about it years ago). Also kiosks and SPAs that 'hide' URLs
alastairc: it's kind of a 'I know when I see it' kind of thing
<kenneth> One of the things I was trying to argue, which is probably tangential, is that SPAs that hide URLs are inherently inaccessible, but that's a bigger issue than just whatever "views" are called :)
<Detlev> I call it a day, thanks!
<alastairc> Re-starting
Alastair: Stepping back from the conversations. What problem are we trying to solve?
Reading slide 3 at https://
… problem 1: scoping a requirement
… problem 2: scoping a conformance claim
… problem 3: Other definitions rely on "views"
alastairc: We have other terms that we are using or intending to use.
… product - website, app, and other big holistic things.
… Task which we will be discussing after views.
… Set of views / page, UI Compnent
Kevin: We need a way to uniquely identify a thing. For a webpage it's a URI.
kevin: I don't think we need to require it
Kevin: Its about being very specific about stepping away for a bit.
alastairc: The problematic side. I know it when I see it. When I was looking through the examples. Most everyone here has scoped a website and tested it. What factors contribute to your decision about what to test/what you are testing?
<Jon_avila> Content that has a different purpose, email inbox, email message, email settings, etc.
alastairc: if a static website, that's fine. If its a single page application or other complex state, how do you define it?
… We have a difference between how people name things.
… Sometimes we vary what we test based on who owns the content (3rd party content, which development team works on it)
… So what other ways do you scope an audit?
<jspellman> I name it whatever the client named it. Often it is page name or a description of the app screen. I also name parts of the page that are common: header, footer, mega menu, side bar, etc.
Shadi: I'm wondering if for better to use purpose. Purpose often maps to type of content as well.
… What is the UI trying to accomplish. One is a form and the other is a video. These are different objectives. What is the primary purpose of the page? It would be a different description.
Hdv: How is the report going to be useful for the team I'm providing it. That can also inform how I scope things.
… maybe I know that listing problems with a certain part will be helpful.
… you also want to make sure you have a complete set of things.
joryc: I often try to test as little as possible while providing as much coverage as possible.
… What is the discrete material that I am testing? What is repeated?
<Zakim> alastairc, you wanted to ask whether purpose and types of content should feature into the definition
alastairc: On Shadi's point, purpose and types of content. I think I see that as part of a methodology for choosing what to test. I think that is a different thing than a unit of conformance.
… that is different than whether the context includes a video or a table. To me, that's kind of separate.
Shadi: With keyboard, it goes back to what else is within the page. If there is a form within a tab, there is a different keyboard context and behavior. Just looking at keyboard. With a video, the keyboard behavior is quite different.
alastairc: The examples where you have a set of tabs. You are saying it matters if a table is in one and a video is in another.
… that is different than saying this is a unit where keyboard must be defined.
q_
alastairc: What is the method of capturing things?
Rachael: In website navigation that doesn't work well, sometimes I've given an alternative to go down a level. If you can't get down to a lower menu, but you can hit enter, can we approach this instead of talking about... can we approach this from a task flow perspective?
Rachael: Where is the boundary point of acceptable alternatives? To think about how we want to scope things.
Alastair: Any ideas?
… [reading slide 5] Ownership, purpose, optimizing...What contributes to your perception of scoping a type of thing?
Dan: For the purposes of what is relevant to do a test. For a webpage it is the whole page. Pages can be made up of sub resources such as frames. The conformance boundary must include all of those things. There is a layer that must be at least that wide.
alastairc: I wonder if we could maybe get a sense of whether people thing, after going through examples, whether something should be the same or a different thing?
… so on that slack example with different columns (see screen share if needed), if I selected the yellow icon to switch from one slack instance to another, would it be a completely different UI context or not?
GreggVan: I've been staring at this, and I'm not sure any of those levels are different. If this was an email program, if each item is a post from a different person, then it's kind of like a TV screen. When it changes shows, it's a change of content within the same context.
… If this is a toolbar at the top, then it's changing pages. I'm wrestling, it's sort of like did the layout and control change? If it did, then it's a context change. If the controls and layout stay the same, then it's a change of context. Different controls and different layouts would be a different context.
<Zakim> Chuck, you wanted to say I think it's part of the same unit.
GreggVan: change of UI-Context would involve a change of layout.
<maryjom> +1 to Gregg
Chuck: Agree. when you are changing the content, the nav bar would stay the same.
<Zakim> hdv, you wanted to agree with Gregg
<Zakim> jspellman, you wanted to say that it can change per guideline for what it is appropriate for, and not have to be a conformance issue.
hdv: I agree. If different controls, then this would be a change in context.
Jeanne: I think we could do this by individual guideline. Some care about component, some care about page, some care about task level. However we want to do this. Looking at it from a conformance level, does whatever this is conform-- I think we could let the site owner or tester determine what that is.
… Then we don't have to be extremely fine grained. Risks painting into a corner.
ack
<Zakim> Rachael, you wanted to say I think this is good direction but how to avoid circular def
Rachael: This is an interesting direction where you are changing controls and layouts, and those would be the boundaries of a view. We define some of the guidance that may have circular definitions. I'm still struggling with a single page app. If you are showing the controls or now, how does that play with this approach.
Shadi: Side note, it would need a weasel word like substantial. When you open a component and get options. That isn't significant.
<jspellman> Even if it is a SPA, the client has identified what all the parts and pieces are as part of the development process.
Shadi: I'm hearing three types of conformance discussions. One is me as a vendor. Whatever I can control as the vendor.
… Second: As an authoring tool, what content is being created and what is in it?
… Third: As a tester. When we talk about conformance, there are additional procedures needed.
<Zakim> alastairc, you wanted to say that the requirements need it.
<Zakim> GreggVan, you wanted to say CHANGE OF CONTENT = changes information but not layout and controls (like changing channels on a TV or different movies showing up in a player with layout and controls the same) CHANGE OF UI-CONTEXT = changes in the layout and controls . (the question is when we move from one to the other) and to say CHANGE OF CONTENT = changes information but not layout and controls (like changing channels on a TV or different movies showing up in a player with layout and controls the same) CHANGE OF UI-CONTEXT = changes in the layout and controls . (the question is when we move from one to the other) MAYBE "inline controls like links are considered content" MAYBE layout changes including controls in layout
alastairc: Jeanne was saying about approaching it as requirement by requirement basis. I think we need a uniform definition that we can use. I think for people making conformance claim, I think we want a definition. I wouldn't want it to be too flexible.
<jspellman> Reagent, make minutes
<Zakim> alastairc, you wanted to comment on other examples, potential definition
Gregg: [see above] The larger page could still have the definition but the requirements could also have additional information. Example: No part of the page can cause a seizure. It doesn't work to say if one part of the page causes a seizure, another part does not.
<Chuck> +1 same ui context
alastairc: Check: On this example, are we agreeing that on this example, any change that changed the content but not the components. Then these would be the same?
Rachael: If we treat it that way and I have a post that has flashing content, but the other doesn't, how would we treat that?
Rachael: If someone posts flashing content, then would this fail?
Dan: Would you have to assess every post?
Shadi: We have the idea of non-interference which would have to apply. You could say this unit complies barring non-interference.
GreggVan: If there are movies or web pages where the content failed, then it would be out of scope because the provided didn't create it. But if the program made all outside content flash, then it would fail.
Dan: I agree but think it's a bit orthogonal to the question. Let's take the user out of the question. If your company has a set of help web pages that are all the same except for the content, do we want to consider all the pages as a single unit or consider each page separately.
Gregg: Same situation as static web pages built off a database. The page passes but you might have to evaluate each page that could be generated.
… The page passes but the database content fails.
… must separate the ui-context from the content within the ui-context.
… now it looks like we have requirements on ui-context and requirements on content.
<Zakim> alastairc, you wanted to comment on components vs content
Gregg: We have had this problem but never addressed it. The guy who creates the page has no idea what's in the page. It's generated by other people.
<Zakim> kevin, you wanted to ask does it matter
alastairc: It does help us to have this one concept for ui component and one for content. We could use an assertion that you've tested this.
kevin: Does it matter? If we go back to what we are trying to define. We are trying to scope a requirement. We could say, my scope is this entire page except for the content in the database. You could also say this is a conformance claim for the entire page which would include both. We are saying you define the scope of the conformance claim.
… We need to have the definition for use with certain requirements. reflow only applies at the page/view level.
GreggVan: Seems to be we have UI contexts and information that is not a UI context but is displayed with a UI Context. Movies for example. I consider controls that flow with text (ex: text links) to be content.
… we have a new item as well. AI. A company may create a site where the content is generated by the user on demand. It's like a live performance. You can't view it in advance.
<Zakim> alastairc, you wanted to add a proposal (and it might mean we don't need set-of web pages?)
GreggVan: you can say after the fact that it was accessible but can't do it in advance. AI would need to have the accessibility rules as items not to violate. Maybe that's how we handle that.
Alastair: User Interface Context (UI-context) - The layout and components available (visibly or programmatically) at one time, including dynamic components that appear without making the previous layout & components unavailable.
… may need a weasel word like "substantial"
Ken: Is the "ncluding dynamic components that appear without making the previous layout & components unavailable" doing too much work? What were you thinking?
alastairc: Let's go through some of the examples. For slack. It is essentially the same as long as the same layouts and components are available.
The modal dialog would be a new context because it makes previous items unavailable.
alastairc: We may not need "set of webpages" because that is a set of web pages.
Gregg: @@@
alastairc: It effectively replaces set-of-web-pages, as even opening a new web-page doesn't change the UI-context
Gregg: The answer is yes and no. within a set of pages, you want the controls to be labeled consistently. This would mean that within the set of UI contexts, you would want controls labeled consistently.
… it would be nice if it would be the in the same location within the same instance.
alastairc: it both grows it and shrinks it.
GreggVan: It allows us to shrink it within web apps.
Dan: The word "available" is giving me pause. If I have separate tabs open, does the scope now include all the pages?
GreggVan: By the author?
… discussion that provided by the author may work. Must tie back to the author in some way.
Dan: Making it independent of the user agent doesn't make sense to me.
… End users view web content through user agents.
alastairc: Trying to make this more widely. A native app doesn't have a user agent.
Dan: When some parts of the page become available or not.
kevin: I wanted to ask if there are fundamental flaws with this?
Dan: I am still thinking about this.
Kevin: Is there something of concern in this that is fundamentally wrong.
Dan: This is very different in terms of tools that we build, between a "page" and a "page state." When you are running tests, you are likely to run a given page in a number of different states. This is attempting to define a page state.
… it is problematic for requirements that are only happening over time.
alastairc: I'm not fond of that because I was trying to think of time.
Proposal: "The layout and components available (visibly or programmatically) at one time, including dynamic components that appear without making the previous layout & components unavailable."
Rachael: I like this definition, because of the page component brought up earlier. You are opening or closing a diaolog, when I complete the content of a form, I get the submit. If it's not removing things, only adding things, that's all within scope. It solves some of the problems.
GreggVan: What I meant was that controls that flow as text don't disqualify a block of content from being content rather than a new ui context
… in other words, if we have a ui context with dfiferent content in it, that content could have controls that flow inline without necessarily creating a new ui context
… eg, if you have emails, and the email program is a ui context, the emails themselves aren't a ui context, even if the email content might contain links
… to answer "do we still need 'change of context' or 'change of content'": we may still need that idea, though I'm not sure we should use the same naming. eg, we'll still need to be able to make a requirement that discourages changing content unexpectedly
<Zakim> Rachael, you wanted to say "available within a single instance of the user agent"
Rachael: How about "available within a single instance of the user agent", so we can avoid edge cases with multiple tabs/windows?
GreggVan: I don't think so - we need to make the responsibility "what an author causes to happen", if the user opens the same page in 3 tabs at once that's not on the author. This is similar to why we exempt user generated content generally.
dan_bjorge: Though I think this new proposed definition has some issues to work though, I think it's more promising than other alternatives so far
GreggVan: "Available" is the problematic word - I can see 6 different laptop screens right now and they're all "available" to me. Something more like "author-provided", perhaps
Update: "An interface with layout and components (visibly or programmatically) provided by the product, including dynamic components that appear without removing the previous layout & components."
Rachael: We'll plan to continue this discussion through 4pm. 10 minute break.
Rachael: We're resuming, welcome back
… We are on slide 21
agenda/
How to incorporate Tasks/Processes in conformance
Rachael: When we did the silver research, it was a large effort for what WCAG 3 would need differently from WCAG 2.
… A lot was based on needing more flexibility.
… A goal for WCAG 3 is defining "task flows" so they can be used for conformance, and how they can provide flexibility for conformance evaluations and better reflect needs of people with disabilities. We want to do so without creating new loopholes.
… Our current definition for this idea is currently using "task flows" as the term. Current definition is "Task flows are a series views that support a specified user activity. A task flow may include a subset of items in a view or a group of views. Only the part of the views that support the user activity are included in a test of the task flow."
… This is very much like views where we'll come back to this and have more discussion about the definition, but let's continue with that as the starting point for now
… <goes through pizza example from slide 24-25>
… So let's talk about how we define "tasks". Does the proposed definition work for everyone?
GreggVan: When I'm thinking of conformance, it's confusing because we have things like "page" or "ui context", and if a page is part of a task flow that sounds reasonable enough at first, but if we talk about a task flow only including parts of a page, and task flows being a unit of conformance, I worry it creates room for parts of a page outside
of "task flows" being incorrectly excluded from evaluation.
<Jon_avila> Screen reader or keyboard users may need to navigate past parts of the page to reach other parts of the page.
GreggVan: Say an ad or a privacy notice isn't part of a shopping "task flow". But of course people with disabilities should be able to know about those too. In general, if it's not decorative, it's not optional.
GreggVan: So I worry about "only things that are needed for the task" being in there.
shadi: I hear and agree with Gregg's concerns. But I wonder if the content should fit into tasks. If it doesn't fit into one task, it will fit into another. I think we tried prioritizing on severity and didn't manage it because something can always be important to somebody.
<Zakim> alastairc, you wanted to comment on step-back, problem it's solving, additional or instead of page/view/ui-context
shadi: I see this as trying to prioritize by context and the journey of what I'm trying to accomplish and having no barriers to what I'm trying to accomplish. I think it would help a lot to be able to prioritize according to that and not having everything be treated as equally important/severe
alastairc: I'm trying to remember if task flow conformance was intended to be "in addition to" or "instead of" or "a prioritization mechanism for" conformance based on pages/UI contexts
<Zakim> GreggVan, you wanted to comment on "prioity vs accessibility" and to comment on "prioity vs accessibility" (ref Shadi's comment)
Rachael: One driver was just that certain requirements need to be evaluated in terms of processes instead of pages. It was another way of looking at conformance for the purposes of getting work done, but the biggest driver was just that some requirements needed to be based on that level
GreggVan: Discussion seems to be slipping into "what do we need" vs "how do we prioritize what's important". I think the latter is more a regulatory concern, not a definition of what is accessible or not. Comments about priority are good but belong outside of our scope (in regulatory scope, for example)
<kirkwood> concerned with multiple flows to complete a task? that is why ;flows’ is plural? … ‘task completion’ is standard cognitive language.
<Zakim> jspellman, you wanted to say that most of the proposals over the years was a cross cut approach in addition to View conformance.
jspellman: My memory of various task flow proposals over the years is that there's been a cross-cut approach with page-level conformance/requirements.
… but there have been many discussions where everything has been a task. eg, "reading" could be a task. The advantage that "task" gives is that it lets you do multi-page evaluations. In COGA we often bring up examples like "if you can't log into the ecommerce site, you can't use the ecommerce site".
… It depends how you divide it, but I think it's a reasonable approach for us to explore, could have many advantages.
… I remember first seeing this in WCAG EM in 2014 with a very early pass at what we're now calling task flows. We've been thinking about this for a long time and I think we should continue to explore this definition.
<Zakim> Rachael, you wanted to say that we don't regulate but we build the ruler and it needs the ticks to measure on it
shadi: I agree that moving to a task model might be a slippery slope, but I do think we should move away from defining "is something accessible or not". In reality it's not a binary, it's a gradient. I encourage us to think about impact levels. Not necessarily priority, but it's a very real situation that different severities of problem exist. Not
being able to log into a site is a fundamentally more severe problem than one optional widget not being usable. I don't think we should treat all issues the same.
Rachael: to Gregg's point, I agree we aren't regulators and shouldn't try to be, but we are the ones defining the tick marks on the ruler regulators use. (scribe note: huge +1)
<Zakim> GreggVan, you wanted to say 1) Plus to task-flows -- for the same reason we have had "processes" But to include - not to exclude. 2) Accessibility is not binary. True but conformance needs to be or it is not objective.
Rachael: I think it's important that we figure out how regulators *could* assign priority/severity and how it'd be use
GreggVan: The idea of a task flow has always been there in process and that's fine. I'm just worried about the idea of using it for conformance in place of evaluating pages because I'm worried parts of pages might be excluded in practice. It also makes it much more subjective rather than objective about what's important/part of a task flow.
… to shadi's point that accessibility is not binary: that's true in terms of accessibility, but in terms of us defining a minimum accessibility standard, there does need to be a binary; either you conform or you don't. You're right that that's at odds to what real accessibility means, but our job is to draw a line.
Rachael: Are people happy with the straw poll above?
GreggVan: Are you talking about an alternate conformance (instead of pages) or a supplemental?
<Jon_avila> Most sites won't fully conform - if we are only concerned with full conformance then that makes it hard for folks to measure progress toward conformance or encourage them to improve conformance.
Rachael: Supplemental
+.5
<Zakim> Rachael, you wanted to react to GreggVan
shadi: In terms of conformance being binary, I disagree. Discrete, sure, but we already have 3 levels in wcag 2. We could define as many as we like.
… They don't even necessarily have to form a ladder like they do today in WCAG 2; it might be orthogonal tracks.
<Zakim> alastairc, you wanted to comment on how to prioritise or measure severity, by requirement, by instance, by task, sans-context
alastairc: I'm sure everyone here that's done testing uses a system of prioritizing issues. Usually this happens by requirement, or by manual adjustment based on individual instances of issues. But "by task" is one of the easier methods to help prioritize. Task gives you an ability above "by page" to say what things are most important for a user to
be able to do.
… You might not have to do the whole view every time you see on as you go along in a task, but it might help prioritize them as you go along.
kevin: We're saying "can an individual complete this task", but what we're doing with conformance is "does this content conform to these requirements". I don't know if that second one could allow us to state anything about the first one. On that basis I don't know that using tasks as a framework for claiming conformance could be done without saying
"every view in that task must conform"
… otherwise you need to get into "what applies to which users", which is a content-focused thing
<kirkwood> recommendation: change wording to define in singular. ‘task flow’ defining plural version confuses definition
Rachael: For next steps: the current approach to conformance is that we will have some basic foundational requirements level of conformance that is somewhat comparable to 2.2 AA. There will additionally be some additional higher level that incorporates assertions, points, scoring, etc in some TBA configuration.
ACTION: Go back and review all requirements that tie in process, and see if they have a direct path through.
Rachael: Current approach suggests a scope that uses a combination on views and processes (see slide 27). We'll need to go back through requirements and double check that this works for everything.
… Historically, we had a scoring proposal from 2020 that had one idea of task flow based conformance. We've moved on from much of that proposal, but it has a task-based sampling mechanism I think would be worthwhile coming back to as a means of making sure conformance is done well and that they aren't just picking a few pages that score well and
call it good
<kevin> +1 as a background to conformance sampling approach
<Zakim> alastairc, you wanted to ask is it best as something to get before conformance
LenB: We have digital releases every 2.5-3 weeks for multiple products. We kick all the tires, click all the links, etc, and our end to end testing verifies that the new feature flows work and such, but for stuff like Gregg was talking about - footer links that don't change with new features, for example - we need to schedule testing of those on a
recurring basis
alastairc: I've been looking at tasks as a way of helping our clients prioritize the key things they should be fixing. I think there's 2 directions here: that's a useful way of prioritizing for folks that are so far from conformance that they really can't do everything at once, but I don't think that's so important to include in conformance itself.
For conformance itself I think it's more interesting to think about it as a sampling mechanism
<Zakim> Rachael, you wanted to react to alastairc
maryjom: going back to task flow: you might encounter other things when you
<Zakim> kevin, you wanted to ask if this is an update to the Evaluation Methodology
maryjom: 're going through a task flow. ads, etc. Even if they aren't part of the flow you at minimum need to make sure they don't block you from going through the flow
kevin: thinking about LenB's comments, I'm wondering if this is really just an update to the evaluation methodology. Like LenB mentioned, it wasn't really something that was defined by wcag, it more came from informative material about what/how to evaluate.
<Zakim> Chuck, you wanted to comment on encountering things along the flow
Chuck: Answering the poll earlier, I gave a +.5 because I am for the idea of task flow conformance but with supplemental conformance requirements for stuff like "ads on the page" near the task flow
kevin: I wonder if you could really make something like task flow something part of the conformance model. I think it's challenging to define the boundary of what's included in it
kevin: I'm not saying "no" but I think it's challenging to make it normative
shadi: I agree with it being challenging. WCAG EM was trying to backfill the gaps in WCAG (stuff like *gasp* defining web sites). I'm encouraging looking at it, not necessarily bringing it in, but I don't want us to push it off to informative
<Zakim> alastairc, you wanted to comment on my previous point - either it is supplemental support info, or it is a way of choosing pages.
kevin: I think there's 2 parts to that: making sure that the ideas are captured so we don't leave them out again and understanding what should be normative vs informative. I think it's an important consideration in how we support people using wcag 3. I'm asking how you define the boundary between these ideas and what is actually needed for
conformance
alastairc: Would we be doing away with page view based conformance? If not I don't see why it matters, it could just be a subset of page based conformance
dan_bjorge/GreggVan: because tasks can span multiple pages
<jspellman> That’s a different task. All of the content could be a task or another task
<kirkwood> a blocker to task completion would render a site inaccessible, no?
alastairc: I still think the main useful thing here is that it's a mechanism for identifying and prioitizing pages to analyze
<Zakim> GreggVan, you wanted to say "we do not require conformance of websites -- so a website is whatever you want it to be because it is only used in stating the scope of the conformance -- and that can be anything (as long as you don't scope something that is the subset of a process)
Rachael: Think it would be valuable to get sampling idea into documents
GreggVan: If we can define "task flow" centrally and concretely, then it's objective and it could be something, but I worry it could cut off parts of the page outside the task flows.
<Chuck> +1 I'm leaning towards Alastair's approach
GreggVan: we don't ever require conformance of websites, but of pages and processes. The only place a website shows up is in the scope of my conformance claim. There's no need for us to define website at all because it's not a measure we use for anything
shadi: At the time some people did feel the need to make a conformance claim on the level of a website --
GreggVan: -- They can. They can say their conformance applies to their site. They can already do that.
<jspellman> +1 WCAG-EM. It rocked at the time.
shadi: WCAG EM offered guardrails for explaining to folks how to claim conformance for a whole website. In terms of wcag 2 definitions, the "process" definition was very transactional. I'm questioning, alastairc, why you keep coming back to web page - even the requirements now that rely on looking at a web page don't necessarily have to continue to
exist.
<jspellman> WCAG-EM demonstrates how much the industry wanted something more than WCAG2 conformance. Sticking with WCAG2 Conformance is still not going to address this need.
alastairc: Keyboard navigability is an example that I think should be at a page level not a task level
shadi: Navigating through a task flow via a keyboard, even if it involves moving through an ad/etc, could be part of a task flow
<Makoto> Japan adpted WCAG 2 and established random sampling method to make conformance claim as a whole website before WCAG-EM was published.
dan_bjorge: Something that feels confusing to me in this discussion, is from slide 27, conformance of UI context and processes and conformance claims being able to use a combination of both. I don't see a problem with that. Some requirements work well as a view and some as a task flow. It makes sense to have both and allow both. I think this already has a reasonable paragraph.
… I think what already is written here.
<jspellman> +1 Dan
Rachael: We aren't suggesting we throw it out.
<Jon_avila> I don't think we are changing everything to change to task flows over units.
Rachael: I think we need the definition.
Dan: Can we talk about it?
Rachael: One point of confusion is, we have process defined in WCAG 2. We have ... that covers some of what was mentioned. A process may go out of views (authentication, help pages). A process can stay in a view, or it could potentially be bigger than the product. I think that needs more definition.
<kirkwood> can we define ‘task flow’ rather than the plural ‘task flows’
Rachael: We have process from WCAG2, and the need to have a small step by step based thing, let's go back to definition of task flow, and I think we'll take this offline, refine, and bring back to the group.
<Zakim> GreggVan, you wanted to say two bits outside of task flow that are important 1) non-interference (keyboard trap/acces, flash , sound etc.) and 2) access to the 'non-flow' content - what is the *objective* test for content not having anything to do with the taskflow for *any* user (Help, ads, context? etc
GreggVan: So I think I'm hearing you can define conformance either by flow or by UI context, and for a process to be accessible, and the UI contexts in it must be accessible.
… I think there are 2 things to think about: non-interference, if you can't get to the flow because you get stuck on the page someplace else, seizure risk on the page, etc. Other aspect is content outside of the flow. If your task flow is a subset of the UI context and there's information outside the flow, what is the objective test for the fact
that those things outside the flow are never going to be used/useful/required by any user? If there's help outside, ads outside, anything, those are things someone might need. If you're excluding them, I don't think everyone will agree about what's required and if you try to do that, you won't be able to create a clear line of what's in and out of
flow.
… So I recommend staying with the idea of UI context and processes within them
<Zakim> alastairc, you wanted to ask which you'd include in a conformance claim, would you pick ui-contexts and included all tasks, or vice-versa?
alastairc: This discussion makes me wonder if I've been misunderstanding what "process" means. "completing an activity" is super broad and might spider into lots of other pages.
<Jon_avila> I don't agree that you have to checkout in order to use an e-commerce site.
GreggVan: The activity of "wandering around" might, but a more specific activity would probably be more constrained.
<Jon_avila> A person could be comparing products or something that doesn't require checkout.
kirkwood: this is an important part of cognitive and task completion. My only recommendation would really be that we use the terminology "task flow" instead of "task flows" where we talk about multiple flows to complete the same task.
<Zakim> Rachael, you wanted to go back to strawpoll
Rachael: We won't have time to finish this today. We'll take an action to come back to this. We have made good progress on exploring the problem space today.
… Next step for chairs will be to consolidate these thoughts into a meeting/GitHub discussion and get it back to you all.
Thank you all.