AGWG extended meetings pt3 -- 25 Mar 2020

<alastairc> meeting: AGWG extended meetings day 2

<alastairc> scribe: bruce_bailey

Break at 10 past the next hour

12: 10 EST

AC: Reminder use present+ in IRC to mark attendance
... adds to minutes

From yesterday, see: https://lists.w3.org/Archives/Public/w3c-wai-gl/2020JanMar/0311.html

AC: we do not keep score, but can be helpful
... sometimes used to remind people that they were at a meeting

WCAG 2.2 Error Correction (was Confirm before submission): https://www.w3.org/2002/09/wbs/35422/wcag22-confirm-before-submission/

https://www.w3.org/2002/09/wbs/35422/target-spacing/results

<stevelee> where are we meeting - can;t see on the aenda page?

Touch target spacing

https://docs.google.com/document/d/1sszSUKB8t3VuRzxHtOjLfQZjNYCw-xr_EbuMwW7WiGc/edit#heading=h.mntlv4jvrc29

<alastairc> https://docs.google.com/document/d/1sszSUKB8t3VuRzxHtOjLfQZjNYCw-xr_EbuMwW7WiGc/edit#heading=h.6u4xclqrczua

AC: pls see email from yesterday

https://lists.w3.org/Archives/Public/w3c-wai-gl/2020JanMar/0311.html

<alastairc> "Resize: A mechanism is available to change the size of all targets so that the width and height is at least 44 CSS pixels; "

scribe: adjusted a couple things, added resize
... trade off / conundrum as increased size increases scrolling
... also using CS pixels, precludes use of zoom and hard to translate between CS pixels and physical size
... conflict between large target size and ability to visually scan page all-at-once
... even before public feedback, we know it is not perfect, but we have a bit of a wall
... so do we go to cfc with this conflict, or does anyone have better ideas?

Can we have +1 to have as-is into working draft?

<LisaSeemanKest> +1

<kirkwood> +1

Rachael: i don't think adding a mechanism should be too hard
... don't have a code sample yet

<Rachael> +1

<JakeAbma> +1

<alastairc> + 0.5

<Chuck> +0, no objections, a bit concerned

<LisaSeemanKest> +1 but want to add "mechanisim as discussed yesterday

Rachael: would like to go forward with asking for comments, seems similar to widgets for font size

<LisaSeemanKest> thanks

AC: pls -1 if you want to hold off
... better attendance than yesterday!

<GN015> +1 , with mechanism

AC: not seeing any -1 votes
... I will have to think about note for FCPWD
... Current draft has a note, but it needs updating to reflect current wording of SC
... for now, will ask for feedback on scenerios where this SC would be hard to meet

RESOLUTION: Accept Touch Target Spacing pending technique creation

AC: need technique before goes to CFC

WCAG 2.2 Error Correction (was Confirm before submission): https://www.w3.org/2002/09/wbs/35422/wcag22-confirm-before-submission/

https://www.w3.org/2002/09/wbs/35422/wcag22-confirm-before-submission/results

<alastairc> Doc: https://docs.google.com/document/d/1pcg6ixAfuwlo6jb2tkZBGTDhF0fAiO49h21E6HCbQ6I/edit#

For processes which require the user to submit information, all the following are true:

[AC continues to read from doc]

AC: Navigate between steps key concern
... from survey, we seem to have worked through most issues
... Wilco had concern about SC being too strict

Wilco: Yes, survey answer describes concern well

AC: Web chat should not be a multi-step process

<alastairc> Some web forms consist of a sequence of ordered and distinct steps. When u

Steve Lee: Web forms are not chat

AC: We have been using processes to be series of steps
... took steps out

SL: had huge problem defining "steps"
... rewrite avoids issues

AC: For Wilco, should we add "for process with multiple steps"?

Wilco: Yes, would like to have that in this SC

AC: Agreed, and concept will apply to other SC

<alastairc> “A step can be a page or other mechanism of showing certain information and inputs at a time. The current step is the information and inputs that are available without navigating. For example, if a form uses an accordion mechanism to show one section of a form at a time, that is a step.”

AC: This was for "redundant entry"

<kirkwood> +1

<kirkwood> seems close to me

AC: a bit off the cuff, but is this useful and close?

Wilco: Would like to see step as normative defined term

AC: agreed, prose adapting for definition.

John Kirkwood: agrees

AC: since we have multiple SC that could use this, kind of plain language, but term will be re-used
... trying to avoid conudrum of single-page apps versus websites

John Kirkwood: One key aspect is that information that goes away via accordian style control, and not being able to go back

scribe: proposed phrasing is useful for situations with information in steps

<Zakim> Rachael, you wanted to say 2.4.5 also references steps: More than one way is available to locate a Web page within a set of Web pages except where the Web Page is the result of, or

<scribe> ... new language is helpful

<ChrisLoiselle> Perhaps adding a glossary term for step , or clarifying a note underneath current glossary term of "process"

<ChrisLoiselle> process series of user actions where each action is required in order to complete an activity Successful use of a series of Web pages on a shopping site requires users to view alternative products, prices and offers, select products, submit an order, provide shipping information and provide payment information. An account registration page requ

Racheal: i am only seeing one SC that uses the term, but agree that it works there (for 2.4.5)

AC: If we say for "multi-step process" and define step ... that gets us most of the way there
... for people coming to this fresh. If you have not filled in survey, please comment now!

Steve Lee: Okay

AC: yes, we have information and controls as bit of catch all
... should be UI and controls?
... key concept is what you can see or get from screen reader without navigating

Stefan: i have proposed less confusing title
... error correction is quite vague and not tied to modest processes as written

AC: Error Correction seems wider than what we are addressing?

Stefan: See comment in Google doc

AC: We need short name, two or three words

<alastairc> proposed: "Error correction in multi-step processes"

<kirkwood> process error connection ?

<alastairc> Multistep process error correction

Stefen: Multi-step error correction

Steve Lee: comes from error correction and prevention in multi-step process

JK: multi-step error correction is the right name

<Rachael> Correction Across Steps, Multi-step correction

JK: process error correction?

<LisaSeemanKest> or for proccesses

AC: We have EC L/F/D

<LisaSeemanKest> +1

<ChrisLoiselle> Error Prevention: In a step or process ?

AC: so how how about EC / processeses

<alastairc> Suggesting: "Error correction (processes)"

<kirkwood> multi-step error correction

<alastairc> https://www.w3.org/TR/WCAG21/#error-identification

<Rachael> Correctable Process

AC: error suggestion

<kirkwood> error correction (processes)

Steve: happy with error correction (processes)

<Rachael> I'm ok with Error Correction (Processes) too

AC: Correctable process sounds like your process is wrong and I can correct it

<stevelee> +1

<alastairc> New: Error Correction (Processes)

<GN015> +1

<kirkwood> +1

<pwentz> +1

<Jennie> +1

<JakeAbma> +1

<Chuck> +`

<LisaSeemanKest> +1

AC: +1 or not on Error Correction (Processes)

<Chuck> +1

<Stefan> +1

<Rachael> +1

Stefan: can leave with it

<Chuck> +00000001

AC: no objections, sounds like a winner

<Nicaise> +1

AC: we have SC, understanding document, technique needs a bit more work
... Understanding will need tweak to account for new glossary term

Steve: Add term?

AC: yes
... open to editorial updated to SC and new definition

RESOLUTION: Accept Error Correction as Amended (pending editorial updates)

WCAG 2.2 Fixed reference points: https://www.w3.org/2002/09/wbs/35422/page-break-nav/

bruce clarifies that resolution covers new term

<alastairc> https://www.w3.org/2002/09/wbs/35422/page-break-nav/results

<alastairc> Document: https://docs.google.com/document/d/12Zn0_TGcqrM-L_wb0PIFHM4AnHJ64wPsucZyRGlf2Fg/edit#

AC: This is essentially the EPUB requirment

<alastairc> When a web page or set of web pages is an electronic publication with page break locators, a mechanism is available to navigate to each locator.

AC: current text
... this is a navigation requirement

<pwentz> link from Zakim requires Administrator rights - you have to remove the " and it works.

AC: as Jon notes, this is limited in scope
... not a requirement to have page breaks
... but is a requirement that if you have them, that you can navigate through them
... in terms of comment, start with Wilco

"Electronic publication" this definition can be misread. Does this "in the form of ..." only apply to documents, or does it apply to articles? If so, this seems to say "Content presented as (...) a single article (...) in the form of a (...) newspaper article". This is fairly odd. I'm not sure defining electronic publications based on the type of content is correct. Why not define this in terms of having a predefined printable layout? Or maybe even [CUT]

<alastairc> Definition is: Content presented as a collection of related articles, a single article, document in the form of a book, textbook, magazine article, journal, scholarly journal, or newspaper article in digital format.

AC: struggling to think of an alternative...
... maybe any technology that has page break locations?
... EPUB based on HTML and CSS, and HTML5 can have defined page breaks
... we went back-and-forth but didn't want to be too universal because it ends up not being so useful

Wilco: Why would this be limited to EPUB?

AC: If defined on any page, implies requirement for table of contents on typical HTML documents

Wilco: Does this help with short EPUB documents? What is point for two page EPUB doc?

AC: We had previous conversations where this came from, EPUB folks liked it.

Lisa: COGA groups identified similar issues with videos, can loose focus/thread/ability to track subject in A/V.

<Shri> Also, providing a mechanism to move to a locater is often done by the document reader software. How content creaters can coply?

Lisa: could also apply to VR, need way to easily return to index

AC: in original formulation, we tried to apply more generally, overlapped with other issues of content and user agents
... we focus on problem trying to solve, understanding describes welll
... Context is print disabilities where people working from EPUB version cannot track to hard copy
... so issue is not going back to TOC or earlier section
... need is to address where teacher says go to page X and EPUB user is lost

Lisa: Had similar problem with books on audio tape

AC: Right, book on tape needs to have mention of print page and how to navigate back to them

<LisaSeemanKest> sorry i missunderstood

<Stefan> http://kb.daisy.org/publishing/docs/html/dpub-aria/doc-pagebreak.html

Lisa: okay, seem context in SC

Stefan: inherit semantics from separator role, and what is this page break useful for?
... screen reader users report feature is very useful. So what needs to be clarified?

Stefen: Not having a specific target point as an anchor will not encourage screen reader feature.
... There is no reason why UA like Chrome or Edge could not offer access to these, when UA used to open PDF.

AC: What we heard from David is that that feature is not well supported. And even if Screen Reader does it, does not help low vision user.

Stefen: This is similar to text space spacing issue, we are getting into features that should be in UA, so this gets to Silver requirements
... this is UA feature for navigation.
... I am interested to know what is left open?

AC: Primarily, it is the issues people have in survey. Wilco had a few.
... Some have been address, Mike G and Rachael seem addressed.

Chuck Adam: WRT Wilco concern for small documents, it could be up to author, regardless of document size.

scribe: if print pages exist, make them available. It is not much overhead.

AC: Mechanisms we have are pretty limited. We do not have UA in our scope. Silver has a little bit.
... reason SC says "a mechanism" allows for UA to provide feature, so no work for author.

Stefan: This is my concern, that SC implies authors need to add keyboard-oriented page navigation'

AC: Agree it would be better for UA to address, but we cannot assume feature is present.

Stefan: So can authors just rely on screen reader features? Do not think so! This has been a problem for authors to understand what they need to do.

AC: Does this requirement put burden on authors? Does it benefit end-users?
... So we do have question about feasibility, but this is why we want to go out for wider review.
... If we put a requirement on authors that is better addressed by UA, then authors put pressure on UA developers.
... Coming back to this SC, is it a benefit for end-users?

Stefan: Good reason to have in EPUB, and many machines make use of features.

AC: Agreed, this comes from EPUB group. This is their 1st priority ask for WCAG 2.2.
... Chuck asked about multiple print version.
... Response from EPUB group was to scope to formats that have print page number identified.
... Wilco in survey: "page break locators": I don't think we can use the word "page" here, it is really easy to confuse this with the definition of "web page"
... I don't think it will be confusing in context, but can people think of improvement?

Wilco: Word "page" is overloaded, but I don't have a suggestion.

AC: make page break one word?

Wilco: yes, pagebreak is better than page break

[AC updating doc]

AC: in current form, have option of not putting in "pagebreak locators"
... best option for not applying too widely. Don't have good definition for "long document".

Chuck A: Agree we should try to define long and short. Let author decide if they need page break locators or not.

Wilco: short documents still have page numbering, so not a strong argument IMHO

Chuck: Organization will choose templates and rather or not documents should have these print page break
... 1. the organization is part of the author decision
... 2. the physical page break will be there or not

Wilco: seems like too much of a heuristic

AC: This does not address to automated pagination
... depends if author made explicit choice
... This came from digital publishing, we tried to apply it more widely, and could not make that work

CA: Automated association of page breaks is not covered?

AC: Correct, only if author specifies.

Wilco: What about print style sheets?

AC: Print style sheets can use role of page break, but not by default.
... would have to apply CSS to role of page break. Probably not a good idea.
... not reliable. The way you print out might work quite differently from what web page author imagined.
... We have slightly tweaked meaning in doc. Any other Qs or comments?

Wilco: What about my comment about electronic documents being defined by their content?

AC: Need proposal.
... just any technology that has page break locators is too broad, since that would include HTML

Wilco: Should this cover all PDFs?

AC: Perhaps, but not in practices. PDF does not have semantic page break location. It is all automatic.
... This issue is only for reflow-able documents.

Wilco: As written, included visual markers for page number. That is a page break locator.
... Seems to apply to all PDFs?
... Any document with a page number at the bottom fits this definition!

AC: Agreed. PDFs have page number navigation built into the client.

Wilco: But page numbers in PDFs often do not match print page numbers.
... Page 1 is title page, but labeled as roman "i"
... Page 3 (from PDF UA) might be print page 1 (as displayed visually).

AC: Mismatch not covered by this SC.

Wilco: We are not right people to have this nuanced discussion!

AC: Missing some of this aspect on this call.
... SC not really tuned to PDFs
... at time

Lisa: Think this can be covered by 1.3.1
... things that look like a page number should be a page number

AC: no, this is a different issue, since no visual
... EPUB could have page breaks in completely different location
... EPUB markup will have bit of invisible code noting location of physical print page breaks

<ChrisLoiselle> Scribe: ChrisLoiselle

<alastairc> Break: back at 30 minutes past

<bruce_bailey> [break until half past]

<laura> Thanks.

I'm back.

Chuck: Page break location, content author and user agent , will this cause a problem?

Alastairc: Epub may not use page. I.e. kindle , decreasing page size...changes page numbering on device.

Chuck and Alastairc: Epub has contributed , but is not on call today to confirm.

<Zakim> Wilco, you wanted to bring back the conversation about "page"

Wilco: Using page break is a great solution, however definition is using page . If looking at WCAG on where we are using "page" we should not overload the use of word page.

location of page in relation to a set

<Zakim> bruce_bailey, you wanted to say EPUB doc is a set

Wilco: Epub has one url.

BruceB: typically Epub is ZIP file and then opens up to a set of web pages.

Alastairc: There could be situations where Epub is one url and many pages associated with it.

Wilco: A PDF is a single web page, even though it has multiple print pages within it.

Epub is one page , per definition for last 12 years.

<bruce_bailey> i agree that ZIP file is at single URL, but i still think an EPUB is still a "set of pages"

<bruce_bailey> we already know that there are some issues with our definition for "web page"

Wilco: Nothing to exclude a set of web pages in current text. If we are trying to narrow down to Epub, the definitions we have don't align fully.

Alastairc: Any ideas?

LisaS: it not intuitive to documents , i.e. how the file is perceived , a document vs. a web page. It should be viewed as a piece of content.

<david-macdonald> I'm available now... I saw that Page locators was being edited

LisaS: approach should that we don't narrow the definition. Or talk to content with multiple pages.

DavidMacDonald: joins the call.

Alastairc: Are the definitions pointing to what we want to be interpreted by the definitions ?

Wilco: Talks to type of content, a book , etc. Epub and Epub like technology mapping?

DavidMacDonald: Websites to be included currently. Epub was not just to Epub, but limited to cover that type of publication.

Definition based on content, but not limited just to Epub. Webpages presented as books would want to be covered by this SC.

Not random webpage missing a heading.

I.e. a single article, very long html page with a lot of links on it. journal, magazine, scholarly etc.

Perhaps removing "a single article" from definition?

<Jennie> Article will need to be define because one product that creates EPUBs uses articles very specifically for organizing content.

Alastairc: Page break locators , page being a loaded word in WCAG

Stefan suggested using pagebreak as one word to make it a bit clearer?

DavidMacdonald: Agrees.

Page numbers is what we are trying to get at with the SC. Example: Ebook open. What is page 65 in their context of use? I.e. Low vision, magnifying and page numbers change.

AlastairC: PDF would correspond with Epub, would we put navigation in PDF?

DavidMacDonald: Page titles would need to map and be able to be located. Page should match what is visible presented.

<alastairc> "role parity to host language"

Stefan: pagebreak may not be needed as a separate term. ARIA and glossary terminology , parity to host language.

DavidMacDonald: role = doc pagebreak for example.

If you have a page break, you need a way to get to it. Visually, put a page number. Or programmatically, there is a way to get there.

Alastairc: To Wilco, does that make sense?

Wilco: On mute, per zoom, but not currently on call.

LisaS: Why not just content?

DavidMacDonald: article is a journalistic term, i.e. magazine article presented digitally.

Alastairc: Perhaps adding journal article to clarify the word article.

Jennie: to describe the word article in the understanding SC text. I.e. indesign, talks to article in publishing content, so would need clarification on vocabulary

a student in k-12 , provided in epub, the requirement would want to talk to that.

journal article vocabulary term defined vs. creation software terminology.

DavidMacDonald: we can talk to the fact we are not talking about creation software

Epub may be a one url, but if an HTML doc, then a set of web pages. Thus both are in there.

Alastairc: I think we covered Wilco's comments.

DavidMacDonald: We are creating a new term, pagebreak, and then will define it.

Alastairc: Any other comments as last call on this SC?

<alastairc> https://www.w3.org/2002/09/wbs/35422/page-break-nav/results

DavidMacDonald: I think keeping at "A" level is consistent as it is pretty specific. I think we are ok with the results / comments.

Epub did write up a technique.

<Chuck> +1

<laura> +1

Alastairc: Could we have plus +1's for including in draft?

<Jennie> +1

<Rachael> 0

<david-macdonald> +1

<JakeAbma> +1

<alastairc> +1

<kirkwood> +1

RESOLUTION: Include accessible page location in 2.2. draft

<Stefan> +1

Jennie: wrapping up findable help and visual indicators?

Alastairc: also redundant entry as well, if we can.

Redundant entry

Alastairc: CfC didn't pass. Definition of step in a process.

<alastairc> Doc

<alastairc> https://docs.google.com/document/d/18MCz5XDsMmglcAe2j-HzQbpADpw_HtdVjDZ3EHX4-xk/edit#

request to update redundant entry to align with the clarification of what a step is.

<alastairc> "For steps in a process,"

does anyone object?

<kirkwood> no objection

Findable help

<alastairc> Doc: https://docs.google.com/document/d/1fX4Iw169OGUny5RTd70S8qAneYy5e0hr7zupE21gPBM/

Alastairc: We were agreeing on SC , Jennie has that been updated?

Jennie: I have updated many items that were discussed. Response are in document, i.e. march 23rd items. programmatic sequence , consistent reading order etc. to address Wilco's concern on consistency

Visual reading order was primarily reason for proposal of SC. Same relative order , in mobile, may adjust in viewport size

Alastairc: Same relative order for same breakpoint size , same visual order across breakpoints and pages would negate media queries.

Jennie: We could get public comment and I'm ok with that to have people identify the term.

Alastairc: perhaps talking to this in understanding document ?

Jennie: chatbot is one piece left to resolve. Hours of operation / FAQ portion are two major outstanding items.

Alastairc: If you provided a phone number , vs. hours of operation...

Chuck: Without hours of operation, expectation would be that it would be staffed 24 hours a day

Rachael: Can we put that into understanding documentation?

Jennie and Rachael: Examples vs. qualifiers , I agree . Looking to make sure everyone is on same page with sub bullets topic

<Zakim> JustineP, you wanted to discuss a request for essential exception

Justine: can an essential exception be presented?

<Chuck> a timed test

Justine: how to find help in a secure platform

<alastairc> Except for archival unsupported content which is clearly labeled as such, or where finding help would invalidate the activity.

<LisaSeemanKest> +1

<Chuck> +1

<Rachael> +1

<GN015> -1

<JustineP> +1

<Jennie> +1

<kirkwood> +1

<JakeAbma> +1

Alastairc: Can we include this in draft for 2.2 ?

<Nicaise> +1

I need to drop off for another meeting , can someone scribe remaining couple of minutes? Sorry!

Thank you!

<Rachael> scribe: Rachael

GN015: There needs to be online help

<GN015> compare ISO 9241-171 Sektion 11.

Alastair: We are out of time for today. It would help if you would write an email to the list. Outline your concerns. We will consider that approved pending. We have been through this a number of iterations and we can then check on your email later.
... we got through a lot today. The next meeting will be on Silver conformance. Rachael will be chairing. Talk to you all then.

<scribe> meeting: AG Extended Meeting Pt 4

<alastairc> scribe:alastairc

<scribe> scribe: alastairc

Rachael: Still looking for scribes, please do volunteer...
... back on the Silver topic, on conformance and scoring
... been lots of work in the TF, good to walk AG group through these options.
... Will ask each person who's worked on it to walk it through. Then clarifications on that item. Then next proposal. Then open conversation about pros/cons.

Current Silver Proposal (Jeanne)

<jeanne> https://docs.google.com/document/d/1LfzTd_8WgTi0IUOOjUCRfRQ7e7__FRcnZow4w7zLlkY/

Jeanne: We've done a couple of versions of this example. It came out of the first survey. Needed an example to show all the moving parts.
... Note there was an earlier, more detailed version, that evaluated a real website against *all* criteria.
... but the feedback was that it's too complicated (for this purpose), and we re-did for the 3 new guidelines.
... we did save it if the questions come up.
... There are links for the spreadsheet & WCAG-EM for more details.
... this goes through all the steps, from scope, to scoring, to total score, minimums. (E.g. talking about min per category.)
... we used the W3C before/after demo, the 'before' version as that has mistakes.
... Because the demo is part of larger website(s), we state the scope is sub-set of W3.org, this demo.
... does not need to be declared as URL.

Next you would take representative sample of the website. For the demo, which is 4 pages, not very helpful. Looked at the WAI website for this purpose.

Jeanne: [reading from doc, step 1 onwards]
... this walks through the scoping, following WCAG-EM. In step 3 you create representative sample. [list in doc]. Step 3b, you randomly select some, based on assumption of 120 pages.
... then 3c you include processes, but we didn't have any for that site.
... The next thing is new, which is taking a sample from a non-website. WCAG-EM is very webpage oriented, so we needed to make some adjustements. This is not cleared with the editors of that, would be interested in discussing an update.
... We took the 'for you' section of NYT iOS app.
... this expresses a scope, but not with URLs.
... [reads from doc under New York Times section]
... Step 2D is about web technologies, which needs more work in adapting it.
... The sample is setup the same way, based on the screens IDed.
... Create an account & login were added, to make 19 screens.

<Rachael> awk GN015

Gundula: Would like to know what common means in this context?

Jeanne: Mostly template pages, where there is a unique style to it, a typical page.

<Wilco> Here's the definition of common page: https://www.w3.org/TR/WCAG-EM/#common

Jeanne: Looked at the scoring, we went back to the before/after demo. Needed a page we could be critical of.
... we've moved away from this in the last week, but just before F2F we would take the 4 things we were looking at around headings: Semantics, visual hierarchy, succinct, strict hierarchy.
... [reads scores]
... took the sum of those, and took average to get 75% for this guideline.
... then we scored Clear Language. You might remember the rubrik for that?

<sajkaj> Shouldn't that be 98.75%

Jeanne: have a lot of items we want to evaluate by. We took the actions of the rubrik, how we tested it, then the score.

<sajkaj> i.e. (100 + 100 + 100 +95)/4 = 98.75

sajkak - there's a 4% as the first value

Jeanne: Because we have 3 guidelines, we'd normalise the score. Every guideline is worth the same. Each is critically important to someone, so treated equally.

<jeanne> https://docs.google.com/document/d/1JctjFhXQrsQJtKMH1mjm12v9ubg_3gVuuDkVapszGtY/

Jeanne: here is an older version with the minimums
... for a real website we took the summary scores and evaluated as silver criteria.
... the table above shows the scoring.

when we come back down to the minimums, we plugged in the numbers for each category (functional).

scribe: the rest are plugged in, the point was you could go to the bottom and see a score per category.

<Zakim> Wilco, you wanted to ask about the level of effort such an audit would take

Jeanne: [reads the levels] Gold > 97%, silver 90%-96.9%

Wilco: Counting all of these seems like a lot of work, partly the addition, partly the counting words aspect for language.
... there's a level of effort that should be looked at.

Jeanne: We don't want it to get much more work either, we have a lot of manual testing already, but would like to be clearer so people know what passes/not passes.

<Zakim> sajkaj, you wanted to check NY Times heading score math

Janina: Is the maths right on the NYT?

Jeanne: There's a first figure of 4% which pulls it down.

Peter: Clarifying question to open it up later... nothing in this analysis looks at the nature of the failure and how much it would trouble the user. E.g. a page with two headings, small amount of text, and a heading wrong would be a 50% failure. But it doesn't establish the harm/impact.

Jeanne: We still have a number of sections that we want to work on, and that is a section.

Rachael: Any other clarifications?

Two Currency Approach (Bruce)

<bruce_bailey> Brief description: Multiple Currencies (no exchange rate)

<bruce_bailey> Postulation: Any point-based rating scheme will need two or more categories.

Bruce: I'll copy from my notes to IRC, can't share doc at the moment.

<bruce_bailey> One set of points tracks towards minimum requirements.

<bruce_bailey> Points beyond minimum can be banked towards “achievements” or other recognition of improved accessibility that reflect best practices.

<bruce_bailey> Simplest approach would be two scoring categories: base/core and achievements / best practices.

<bruce_bailey> A more nuanced approach might have separate point totals in 7-10 FPC categories.

Bruce: not a fan of that, but possible.

<bruce_bailey> Multiple currencies address some obvious problems with single point system.

<bruce_bailey> How to award points for nice-to-have features while making sure basics are not skipped?

Bruce: have to be able to score base requirements before those above and beyond minimals.

<bruce_bailey> How to recognize an author who scores better than 100% in any one GL?

Bruce: with 2 currencies you can do that.

<bruce_bailey> Multiple currencies will also facilitate scoring transition from 2.x to W3CAG.

<bruce_bailey> Email to list:

Bruce: translating to 2.0 to 3.0 would only be the points, none of the second currency.

<bruce_bailey> http://lists.w3.org/Archives/Public/public-silver/2019Jun/0045.html

Bruce: Observation that we need these two sets of points. One base, one for recognising beyond the basics.

Chuck: Comment, an example discussed of going 'above & beyond', was the usability testing, to get the bonus points.
... are there other examples?

Bruce: E.g. headings, there are structural aspect, but maybe you score more for strict correct structure. Best practices you get points for.

<CharlesHall> another example is starting an Accessibility Champion program in your org that includes people with disabilities that contribute to the product in scope

alastairc: That helped, having points *within* each guideline rather than per guideline.

<JF> @CharlesHall that does not scale, and would penlize orgs too small to justify a "program"

Bruce: I don't have more details, it's an amendment to the other proposals.

Rachael: Have you thought about how it would play out with Gov for conformance?

Bruce: E.g. under ADA/508, there could be something that federal websites should meet certain levels.
... with transition stage, I would guess that 2.0 criteria would only be points, and the others will be achievements that go beyond that.

Peter: You don't see any exchange between the currencies? Or could it be: You didn't hit the WCAG points, but you did so much more so you pass.

Bruce: I don't think so, but it would have to be a fundamental feature of 3.0, that you can't translate achievements into points. You might just advocate that instead of meeting it.

Alternative Approach (John F)

<JF> http://john.foliot.ca/demos/HeadingsTestStart.html

JF: I shared this a little yesterday. Taking the headings requirement, I looked at the current docs *and* some current WCAG SCs.
... I looked at the 9 functional requirements in the EN. (Section 508 has 7, but some are grouped.)
... the cognition one might need breaking out.
... the first is a 'perfect' page, with proper structure, styling,
... basic but gives you the idea.
... distilled down to: Headings defined, navigate blocks, programmatic structure, visual structure.
... The second example uses H1s throughout. However, still styled to look different.
... then compared to the 7 visual requirements. Made it into a visualisation, with green yeses, yellow partials, red nos.
... when it lacks programmatic structure it impacts a few groups.
... The next example breaks in that some divs use role=heading, but no level. Incomplete.
... then the categories show more 'red', breaks for more people more often.
... example 3 more harmful.
... the example 4 was great for screenreader users, but the styling is gone so no visual clues.
... point of exercise is that a) using the functional requirements as a filter/tool for granular measure is useful.
... not figured out scores, but when you look at the impacts you can see that some are more harmful than others.
... For scoring, I think this needs to be brought in somehow, someway.
... I've only used this on headings, which is conveying two bits of information (heading & level) visually / programmatically.
... for others they might be pass/fail like language of doc.

Chuck: This helps ID the impact on a user group, did you envision a composite score?

JF: original proposal was to use these func requirements as a multiplier. In most cases it would be 1, but if something is very critical then we could multiply to weight more.
... problem is that the maths gets complex.
... I'd like to see, but not sure we'd get there.

Peter: I didn't see a boiling it down into an end number, would that be surfaced by disability? Would that be useful?
... e.g. for hiring scenario?

JF: Not thought of it like that, but looking at test 2, it's a partial pass. Want to give credit, but also didn't measure what happens when the hierarchy is broken. In some cases it might not matter than much, so awarding that something even when not perfect.
... OTOH you could inject more confusion by getting headings wrong. Want to weight by disability types, but including composites. Figuring out #a score based on this seems useful.
... wanted to get feedback about whether it goes down the right path.

<Zakim> bruce_bailey, you wanted to ask if JF can say anything about FICO model

Bruce: What happened with the FICO style scoring?

JF: I mentioned that before, it is a method of calculating credit worthyness, goes from 550-800. within that you also have fair/poor/ better/best.

<bruce_bailey> @JF is FICO type scoring compatible with this more advance model?

JF: most people who have one, have no idea of how it is calculated, it is behind the scenes. But you know there are certain things that go towards it.
... most people don't know how it works. For WCAG 3.0, that's where we want to go to, have a simple score, but that the calculation would be very granular.
... some of the WCAG 2.0 SCs are more important than others.
... as we go through requirements, might want to think about weighting

Bruce: Hearing that it is compatible with simple scoring end point.

JF: Would think it's a gradual increase

<Rachael> alastair: When Jeanne and John referred to chairs (aka me) being too complex to present, what I meant was that going through all WCAG 2.2 as part of the example was to complex. Focusing on the new ones was sufficient.

<Rachael> ...There are two importance levels, even within headings, there are different scenarios that make points more or less important based on context. This adds granularity but we aren't accounting for task based importance.

JF: The impact was one thing, the scoring was real. Idea would be that this is one form of currency, other types of test would be another currency.

Jake: Would like to tell just a little bit about how I judge a website with 4 steps. I use WCAG-EM, and base some human judgement like John has done, and give a severity based on context. There's a lot of concerns about a complex approach with first example.
... then complex example from John. I use a sub-set of pages, then pass/fail like WCAG. Then build on top of that one. Next part is low/medium/high severity level. Subscract high severity levels. Third step is delivering the top 20. Only take high severity/critical ones. Human judgement thing, but all testing is.
... always explain. 4th part is task-completion. E.g. fill in a form, might be hard to get there, main mains difficult etc. separate thing on it's own.
... I need human judgement info. Why is it is fail, is it not?
... that's more like your background info for every auditor.
... all info we've provided today is like the background info to inform your top 20 / task completion.

<Chuck> yes, sounds like a new proposal

Jake: Could build this on top of what was proposed...

<Zakim> JF, you wanted to respond to jake

JF: Caught onto the impact level, our process also has levels of severity. We have to admit that some things are worse than others.

Jake: difference between automated and human judgement.

<Zakim> jeanne, you wanted to ask Jake if his measures severity by disability, or by importance on the page?

Jeanne: When you assign a severity, is it by importance to a disability group, or the importance to the task?

Jake: that's where the key lies, can't always be judged. I'd like to see it judged by me with my knowledge, and explain why/how/who.
... would like to put that into a framework, so the points system is simple/easy, then have a judgement call as to impact.

<Zakim> bruce_bailey, you wanted to discuss admin question b4 break

<Rachael> break for 10 minutes

<CharlesHall> regrets, I will not be able to stay on the call after the break. so dropping now.

<PeterKorn> It's beer-o-clock many places.

Adjectival Ratings (Bruce)

<Chuck> scribe: Chuck

<bruce_bailey> Brief description: Adjective Ratings

Bruce: ..just as Wilco saying "this is a lot of work". Which is kind of what my impression was last week when we did the dry run with the tally.

<bruce_bailey> Problem 1: Manual tally of FICO style rating (up to 1000 points) is not humanly possible

<bruce_bailey> Problem 2: Many GL (e.g. Plain Language, Visual Contrast) do not lend themselves to point assignment based on counting.

Bruce: I see 4 big issues with where we are going with these numbering systems. 1) Manual labor... fico style, that's a lot to do by hand. Based on Jake's description, it's a lot more than many of us do.
... many of the guidelines, plane language and visual contrast, don't lend themselves easily to a tally or a count.

<bruce_bailey> Problem 3: Automated testing only can catch ~40% errors.

Bruce: Also, there is no good reason to believe that automation will solve this. Automation can only get x%.

<bruce_bailey> Problem 4. While automated tests are good for tracking incremental improvements, scores from different (brands of) automated tests not directly comparable.

Bruce: I heard Jeanne say that she hoped tooling would get better.
... Automated tests are very good for tracking incremental improvement. In 508 we encourage automated testing. It lets you measure progress over time, but you can't compare 95% from one product to another.
... The metrics you get are consistent for a specific site, but not very comparable across sites.

<bruce_bailey> How scoring is handled? Three are three steps. (Skip Step 2 for single-page apps.)

<jeanne> Jeanne: I didn't mean that I wanted the tooling to get better. I meant that I expected that tooling companies would adapt their products to WCAG3, the same way they adapted their products from WCAG1 to WCAG2.

Bruce: I propose changing to an adjectival based system.
... Rachael outlined something for me.

<bruce_bailey> Step 1: For each unit (e.g., webpage) in the scope of conformance (e.g., website being evaluated), assign adjectival ratings for each GL.

<bruce_bailey> Strawman adjectives: Outstanding, Very Good, Acceptable, Unacceptable, Very Poor

Bruce: Step 1, go through for each unit (an example would be a web page), assign an adjectival description for each web site. Outstanding, very good, acceptable, very poor.
... For each guideline you score how they did.

<bruce_bailey> Step 2: Derive a representative rating for each GL in the scope of conformance. This representative rating may be mode, lowest rating of any page, mean (by assigning point values), or some other rubric.

Bruce: Step 2 is for overall site.

<david-macdonald_> scribe: david-macdonald_

<Chuck> Bruce: It's an overall using same adjectives.

This bit is a bit arbitrary because you have to decide how you can aggregate the scores from several pages to one

If you have nine varied goods and one acceptable then your rate is very good that's one option you have to decide what you're going to do ahead of time.

<Chuck> Bruce: It's a judgement. Could be the lowest of 10 pages, or maybe you decide to do the highest. We decide ahead of time.

The other thing that happens is you get numbers designed to each of the values for example an outstanding is a for a very good is a three acceptable is a two,

<Chuck> handing off to Dave, thanks.

<bruce_bailey> Step 3: Derive an overall adjectival rating for the scope of conformance, using a predetermined rubric. For example:

The point is you really need to decide what you can do overall before you get to that step

<bruce_bailey> Outstanding: At least two GL rated as Outstanding. No more than one GL rated as Acceptable, and all other GL rated as Very Good or Outstanding.

<bruce_bailey> Very Good: Half or more GL rated as Very Good or Outstanding. No GL rated as Unacceptable or Very Poor.

<bruce_bailey> Acceptable: All GL rated as Acceptable or better.

<bruce_bailey> Unacceptable: One or more GL rated as Unacceptable or Very Poor.

<bruce_bailey> Very Poor: Two or more GL rated as Very Poor.

<bruce_bailey> https://docs.google.com/spreadsheets/d/1G0KLv1Nfvy5QWN7t9jPxyE6UEcTHE5A8tKYiDOhuZRY/edit#gid=1833982643

The important point is there is no overlap between any of these

<bruce_bailey> Semantic headings are used throughout to organize content.

Bruce: going through text of this link https://docs.google.com/spreadsheets/d/1G0KLv1Nfvy5QWN7t9jPxyE6UEcTHE5A8tKYiDOhuZRY/edit#gid=1833982643

One important thing to note is that the outstanding has some things that go beyond the WCAG to best practices

<AWK> Seems odd that Very Poor is worse than unacceptable. Just a thought.

<bruce_bailey> Step 1, rating a single GL on a single page, somewhat subjective.

Bruce: continues to go through the link contents

<JF> +1 AWK

<bruce_bailey> Step 2 (rating of GL for whole site) is somewhat arbitrary, and Step 3 even more arbitrary than Step 2.

<bruce_bailey> Thresholds for Steps 2 and 3 will need to be revisited.

<bruce_bailey> Finish clear language write-up.

<bruce_bailey> Draft adjectival rating description for alt text and keyboard accessibility.

<bruce_bailey> Experiment to see if we can get inter-rating reliability on five GL for a few model websites.

<bruce_bailey> Disadvantages: Not granular. Can’t rank one Very Good website against another Very Good website.

One of the problems is that it's not granular

This whole rating system comes out of contract and grant reviews, annual performance appraisals and generally one tries to have pretty large gaps between good and very good

Because you don't want somebody filing a lawsuit or complaint because they missed getting very good versus a good rating by 1/10 of a point

If someone thinks be a close so not sure may not work in real life

<Zakim> Wilco, you wanted to bring up the WCAG-EM predecessor

Wilco: I must say I like systems like that it seems straightforward and viable and it also reminds me of the predecessor to the we can make evaluation methodology

Which was WCAG evaluation methodology mark for the Dutch quality Mark

You counted up issues throughout the sample and you went past a certain threshold you would fail the success criterion so there were some level of tolerance built into it

The problem with more granular types gifts and an over stated focus on specific numbers which are not very accurate

<Zakim> alastairc, you wanted to ask if it would be by guideline, or method, or TBC?

Alastair:

When we started to go through it I thought it was going to be per method rather than per guideline because it kinda struck me as something that will be suitable is that and then he would use those sorts of methods to score the guidelines

What you make it a lot more complicated than Bruce intended

Wondering if Bruce considered the granularity level of that in taking this approach could lead you towards slightly more granular guidelines

For instance for headings you might separate the semantics from the visual side or something like that

<PeterKorn> +1 to that idea Alastair - adapting SCs to this model.

Bruce: not doing it by method was deliberate but I'm not sure I'd be in favour of breaking down the headings any further.

One of the examples that I became aware of with John's assessment was there were holes in what I have covered with what heaved presented. But I think sticking with the guidelines is the most fruitful path

Alastair: did you think some of the guidelines scope could fit into the framework

Bruce: we want assessment to follow the guidelines not the other way around

<Zakim> JF, you wanted to say nobody suggested that scoring calculations would be a manual process

John: I've never seen that last example in the wild where was an extreme outlier so if we define the post we can focus on the middle we can see the boundaries. It was intended to be outside the boundaries

But I would be interested to see your scoring mechanism applied to the four or five examples that I put forward where I'm asserting that the difference some example some of them are better than others

I tried to give a range of the types of things that we might see in a wild so that we could try to test some of the scoring mechanisms against

I found it a bit interesting. He said it was can be complicated to this calculation then you added more things where there was a manual process.

<alastairc> My guess is that the adjective approach would average-out the odd examples JF created, because it combines the requirements into one guideline / score.

I don't think anybody thinks that arriving at the final score will be manual but I think in the final analysis where you have to have some kind of tooling that will allow calculation_automatically

Because forget to get away from a binary Passfield model into a more weighted model or score that recognizes 65 but 85 is

Distilling it down to percentile will be automated and so will be that difficult to get more granular.

<alastairc> presumably 'math' based scores could be one part of the adjective category?

The other, and I wanted to make about scoring is that one of the things we've not talked about this decay rate. If I run a report tomorrow I have a high level of confidence that the score will be accurate. But how ocular will be in six months

<alastairc> e.g. 20% non-contrasting text could fit into 'acceptable' (or something)

Are we going to have a decay rate that says over time your score that you've advertised is going to degrade if you don't continue to repeated testing and repeated evaluations to ensure that you not slipping

<PeterKorn> as is mine

Chuck: Bruce you'd mentioned at the start that it's not numeric but it's adjective based

What is value in staying adjective when there is a corresponding number

BRUCE: it comes from my experience with dealing with awarding contracts and annual reviews. The goal is to stay with the adjective as much as possible any time you move away from the adjective

<alastairc> Choosing the adjective is not numeric, but then applying 0-4 per guideline would then give you an overall number?

It's not wrong but it's weakness

On the other hand it's in need, how do you define between average and very good. So we need some way to distill multiple adjectives to a single adjective. In the best case you do that through three tally

He saw with my step three where I had two outstanding's and one very good as a whole it was outstanding

That's the best case scenario

You might not be able to do that all the time so the other example I have is where you have multiple webpages and half of them are very good and half of them are excellent

What rating do you end up with overall across the website? That's where you typically need to resort from a number and a mathematical mean.

CHUCK: so you could have good and 1/2

BRUCE: no you would return to the adjectives because you can't trust the mathematical meeting, you have to read to sit to an adjective reading

Peter: I really like this have you thought about with her might not be an adjective reading like three flashes?

Bruce: some will be pass fail.

But I still think you need an adjective at the end

Peter: I think about captioned quality which is been a big issue in the deaf community. I think where we are with headings and I can easily see with our single-A AA AAA

We could build it into say look. Great headings everywhere but maybe we don't have heading to and heading three properly and would map to this going properly to the nested levels would map to an excellent.

John my concern with percentage is that it doesn't leave room for more than 100% where's we can always add another adjective like outstanding or magical or something like that

<sajkaj> RE caption quality, Judy was noting on an earlier unrelated telecon today that the most common ASR error is getting negatives wrong, e.g. do vs don't; is vs. isn't; etdc.

<Chuck> dm: I actually ... Bruce I like it, because I do a lot of evals, my fear when looking at some of your early models.. I do a LOT of evaluations, 60% is marking down "missing a role"...

<Chuck> dm: I do lots of testing and evaluations. I'm concerned about the complexity of the testing, and testing passing things makes me nervous. If I have to check all the headings, get a % of bad headings...

<Chuck> dm: It's gong to 3x the time to do the evaluation... I like this in that they look at this, they assign some adjective, it's honest....

<Chuck> dm: I never say "it's 63.5% if you add another 3 headings its 66%". I like this direction.

<Zakim> jeanne, you wanted to say that I really like the adjectival proposal. I want to see it tested with real websites, because we learn a lot from real testing. We need help to do

<Chuck> dm: What's the time when we end?

Open discussion of scoring

Rachel: there was a lot of conversation around the difficult and the labour required in the testing challenges that didn't really account for things that cause trouble to users

We talked about the tension between manual and automated testings along with decay rate now working to have an open conversation about all of this

<Zakim> JF, you wanted to discuss percentages

John: I want to respond to a comment that Peter made Peter Peter in the world of math there's no such thing as a number greater than 100%. I know we want to think of it something is greater than 100%.

We can say that something is 110% but were really deluding ourselves.

Perfect would be 100% and I don't think we'll ever see a webpage that is 100% ever mined a website. Because perfect is an unachievable goal

So I believe that anything we get could be expressed as a percentage just by taking whatever number you have dividing it by itself and multiply my 100

I'm less concerned about that that I am concerned about this idea that David's concerned that it will take longer

I think one of the goals were trying to do here is to get beyond this chasing after score and moving towards a developing towards a score

I recognize that early on and in our life example now were evaluating content so that it can can be remediated that seems to be the second part of the conversation

But I think all of us are driving towards the day when that will be about remediation but how much is realistic to achieve realizing that will never get the perfect

Peter: before I get into the new topic that I want to get into and by way of observing that I have a mathematics degree in Berkeley. The issue that I have with a perfect was illustrated well in the early silver conversations

Where one idea was if you did everything in WCA G but didn't do some of the new things that were introduced in 2.1 or 2.2 you might have a score of hundred.

But if you got more things I got introduced later the old score would go down you just have more to score higher

And I think that that's just one thing about adjectives. We want to have room for folks to go above and beyond and how do you recognize that if you capture score at perfect for what we thought was good enough at one moment in time

But I'd really like to get folks reaction to this question how do we recognize and weave into this the importance of one thing or another on one page or another or one part of the UI or another

Traditional software testing you look at what are the key workflows key tasks that you try to do what your app may offer any number of tweaking settings how do I adjust the current thing that I'm doing in Microsoft Word adjust the white space above and below the footnote separator line

And if that wasn't completely accessible would we see that Microsoft office is not accessible because this one obscure future isn't fully accessible

Or do we say look there are more and less important parts of an opera site and we will use that in our scoring to come up with something that realistically measures what most users with most disabilities are trying to do most of the time

<JF> Bingo! Maybe it's 96% accessible peter. :-)

And I'm really curious about what folks think about weaving that into silver conformance

John: I think that again comes back to Bruce's idea of multiple currencies resonates with me. Because with peers talking about with me is that some these requirements. Their subjectivity to it

This almost can be subjectivity involved with I've heard Jake say it, David alludes to it what happens when we actually do things to enhance user experience that's can be harder to measure

<bruce_bailey> I have posted my notes to the list:

<bruce_bailey> https://lists.w3.org/Archives/Public/w3c-wai-gl/2020JanMar/0315.html

So I think having a category of measuring those sorts of subjective evaluation criteria as opposed to the more pure play mechanical criteria would be useful

One of the things I talk about in social media is that in that scenario there are some things that you as a content creator have control over and there are some things you can't control

Your posting a graphic file on Facebook you can certainly add the all text and you're certainly responsible for colour contrast to some of the usability once it is posted in the Facebook platform is completely outside of the control of the author because you can't control the backend of their system

So already thinking about categories of stuff that is really granular versus something and has a level subjectivity seems to be to natural and complementary springboards which are also unique and different at the same time

Jason: so think one thing of asking the question is to say what kind of policies can the proposed conformance scheme support

<PeterKorn> Excellent idea Jason!

It seems to me there different policies that one might want to establish for certain different kinds of content in different contexts so in some cases

Everything in the application might say is critically important, such as a voting system needs to satisfy highly on highly in every dimension

On the other hand there are large's websites with lots of information some parts are not so critical

I think it depends on. We need to be able to establish what kinds of policies are is to formulate with the proposed conformance scheme

Is it easy to formulate a policy for example that would ensure that something which is completely inaccessible along one of the functional conformance statement dimensions

Is not going to satisfy the policy

You may have to set a high threshold for the scoring so that you ensure

So I think the question for me what kind of policies are rational to create and what kinds do we want to rule out absolutely and how do we set out our scoring system to make sure that it's easy to specify the kind of policies that are reasonable in different contexts and for different organizations to create

And not taking into account users constricts your ability to get an accurate result.

<alastairc> Noting we have requirements: https://w3c.github.io/silver/requirements/#requirements

Alastair: in my head sort of comparing between approaches from Jean sort of initial full list conformance model approach sort of builds from the 2.X with variations within the guideline sticking the guidelines as equal

And I think one of my bigger concerns there is that it's an additive approach in that you kind of build up a score and you have to certain meet a certain level switch sounds reasonable

But it can make it very difficult for simple sites to actually either meet that her score will were as Bruce's adjectives makes better use of the shoulds and appropriate's, which gets around the problem of having to build up a score or building staying at the top and knocking down a score

I also have a question which is about the interactions of Bruce's two proposals. If you have an adjective approach how would the two currencies fit into them because seems well suited to the first currency but not so well to the

If you going back and looking at task-based things

Bruce: my two proposals are totally incompatible with each other which is why asked to have them separated time in the presentation. I think there so both addressing the same sort of issue the bigger meta-issue is that we want silver to have a way to reward going beyond the floor the bare minimum

You can do that by having two currencies or you can have that by adjective readings

Rachel: I see synergy across what I heard today one of them was what I heard in Jake's proposal and I would like for Jake to write that up so we have something to reference to it but the concept of steps with different approaches to ratings

In the concept of being able to select which guidelines of portions or guidelines seems like a way to narrow down the challenges and different steps could have different ways of reading and could end up in a final score

There pieces of each of these and have a place and it might be stepping back and thinking about the process of testing itself that could pull them together in a way that would work

John: I like to get back to the K rating idea but the reason I'm in Q because I'm concerned as I'm listening to this we talk about points I think were still very focused on web content as we know it today talking a little bit about the applicable about single page apps

One of our goals was also to be scalable enough to deal with any type of emergent technologies

There going to be those types of technologies since platforms where many of the rules and requirements we have today may not be applicable or need to be thought about in a different way

Augmented reality or virtual reality today. The pool of things that need to be done to prove accumulator scores can be different then one of assessing an e-commerce site to order pizza

Tasks are different functions or different experiences are different yet at some point we also want to have at least in my mind the ability to do a comparative analysis say that this website to order pizza is 85%

And so is this virtual reality where 85 captures the experience rather than the content that was being evaluated

Again this idea of having a scoring Metrix that is really based on the seven functional requirements. We already have this in some ways being mandated to us in 508 and in the EN 301549

<Zakim> Chuck, you wanted to say that Bruce's two proposals are not really incompatible

Chuck: Bruce I disagree with you that they're incompatible because I think if you were to apply the adjectives to the first scoring system your alternate scoring system is already representative as an adjective system

It's not points already and it's just choosing different things to rate with those new adjectives. So I think they pretty well together

I think you could tweet part one of your proposal and part two of your proposal and put them together.

<alastairc> Yea, that's what I meant where adjectives replaces points, and then a separate thing for achievements above that.

Bruce: I think one of my motivations for the adjective ratings is to avoid the close counting of each error as Jake described in his and as David described in his.

If you get a whole bunch of alt images, you know if there's a problem with generating alternate text you have to keep doing everyone and evaluating each one you know that it's crappy

Peter: I just wanted to add in this verse the adjective approach were using that internally at Amazon and find it to be incredibly powerful both as far as when you are evolving and maturing and accessibility effort

To say that our goal this quarter or this year this adjective and next year it is this next adjective.

And you can also set goals for specific user groups. It's a powerful educational tool. It's easy to explain to you because you can boil down excellent just passing or different adjectives much better than a percentage

I'd like to give three cheers to adjectives, you can add an adjective above great like exceptional or amazing what you can't do if you're using a maximum number

<Zakim> alastairc, you wanted to outline a similar approach to Jake's

Alastair: you mentioned Jake's semi proposal I have netted along a similar line where we done two audits for one client at the same time. One was the external facing site and one wasn't internal GIS application

<bruce_bailey> i think if you need a labor-intensive way to score and then assign an adjectival rating from the score -- it kills the main advantage to adjectival rating

Which was not responsive which had no accessibility and it was a car crash. And they came within two WCAG issues of being equal in our report.

Based on that we took a similar approach to Jake's if you look at the kind of tasks can do that websites and barriers that are the issues we found from the BCG point of view of those barriers impact the task and then we gave a score out of four groups

Flexible input flexible output screen readers and metadata and understandability and cognitive, which doesn't matter too much but we can see that there's roughly an expert percentage chance that somebody with a disability would encounter a barrier trying to achieve a task

It's high level you can get much more granular but we did find that very necessary in that top level reporting to say that even though the public website at the occasional heading issues where is the internal site that had all of the heading issues

On the adjective one. I think anyone who does reports would agree it's good way of doing conformance but I would anticipate anyone doing professional reports would need to add the kind of instances

Jason: just a brief comment that I think it would be helpful for people working on silver related conformance proposals to think about how they would apply to natural language dialogue applications such as dialogue only applications that might use speech recognition

Or that might provide text input for those who cannot give spoken input and in those types of instances

You don't have the same level of granularity that's being discussed in these conversations you might have capabilities that you wish to reward in a conformance scheme

For example say your speech recognizers better that helps better recognize different speech disorders then the next guy you'd want more points on that

If the system is able to adapt automatically or through configuration so that addicts better able to support people with certain cognitive needs such as additional reminders or facilities to help clarify their utterances

<Rachael> [please] close the queue

This is adaptive in the automated or machine learning or a manual configuration is personalized will in that sense and that's another thing you'd like to reward as well

<Zakim> JF, you wanted to disagree about "number of alt values"

So how's the system going to be able to handle the kind of case worked potential and when we get something that's distant from these different technologies that don't apply the same as applications or webpages

John: disagree with Bruce alt text 5/200 are lacking alternate text and we can't two axes. In our new conformance scheme I want to reward you for the 97.5 the received. We want to account for the_calculation

<alastairc> JF - that would be within the guideline/adjective, like the headings example

To say you're all text is pretty good how do you quantify pretty good. Out of 200 images only missed five

I'm all for the adjectives we need to be able to get down to the tactile difference between good and very good so were all singing from the same songbook

<Zakim> jeanne, you wanted to say that I really like the adjectival proposal. I want to see it tested with real websites, because we learn a lot from real testing. We need help to do

<bruce_bailey> i think JF example would be acceptable at best

Jeanne: this is been a valuable discussion I like the adjective proposal but I do want to see tested with real websites because we've learned a lot from real testing but we need help to do that

<bruce_bailey> i am happy to use adjectival rating to a couple web sites

Particularly we need help from people who have experience with testing of real websites. Please please please talk to us and volunteered to help us try out these proposals is the devil is in the details thank you

<Rachael> https://www.w3.org/WAI/GL/wiki/Scribe_List

- DRAFT -

AGWG extended meetings pt3

25 Mar 2020

Attendees

Contents

WCAG 2.2 Error Correction (was Confirm before submission): https://www.w3.org/2002/09/wbs/35422/wcag22-confirm-before-submission/

Touch target spacing

WCAG 2.2 Error Correction (was Confirm before submission): https://www.w3.org/2002/09/wbs/35422/wcag22-confirm-before-submission/

WCAG 2.2 Fixed reference points: https://www.w3.org/2002/09/wbs/35422/page-break-nav/

Redundant entry

Findable help

Current Silver Proposal (Jeanne)

Two Currency Approach (Bruce)

Alternative Approach (John F)

Adjectival Ratings (Bruce)

Open discussion of scoring

Summary of Action Items

Summary of Resolutions

Scribe.perl diagnostic output