W3C

– DRAFT –
(MEETING TITLE)

22 September 2022

Attendees

Present
AvneeshSingh, CharlesL, ChrisOliverOttawa, Gautier, GeorgeK, gpellegrino, Madeleine, mgarrish, MURATA, Naomi, tzviya
Regrets
-
Chair
AvneeshSingh
Scribe
Madeleine

Meeting minutes

Update from meeting with ARIA WG, at TPAC for proper implementation of DPUB ARIA roles in Apple's accessibility API, AXAPI.

<AvneeshSingh> https://www.w3.org/2022/09/16-aria-minutes

Avneesh: update from DPUB ARIA

Charles: Met with James Craig to show him examples from Taylor and Francis
… showed how publishers are showing info
… cleared up confusion on long post about priorities
… Big meeting on Friday morning where Avneesh, George and Gregorio (phone) spoke
… Group still has concerns about the roles and would like a smaller sub-set; developers think some aren't needed
… top priority ones seem likely to be in the API but not the descriptions like chapter, glossary etc.

<Bill_Kasdorf> could somebody put the Zoom link in the chat? The one in my calendar is apparently out of date and I can't access an .ics file.

Charles: concerns about links to help screen-readers users skip quickly but our doc doesn't say what must be spoken
… users need ability to change settings

https://us02web.zoom.us/j/89231066800?pwd=ZytmVmZJZmVnV3JvTjV5ZEZ3RHpBdz09

Avneesh: the meeting felt more positive than previous meetings and James now understands more issues
… but he is concerned about 30-something roles. He thinks maybe 17 should be exposed to users
… He also suggested the very highest priority roles should be in main aria spec as well as DPUB

Tzviya: after working on this for several years, what does this mean? what does this mean for roles considered less important?

Avneesh: if doc-chapter becomes chapter, doc-chapter will still exist
… for roles deemed lower priority, we have made a good start getting many roles supported. Some tools support all roles, so if Apple moves incrementally at least they are moving

Tzviya: what about AT support for APIs

Avneesh: list of support: Google Chrome with TalkBack supports almost all DPUB roles. Google Playbooks uses Chrome so it supports

<Zakim> tzviya, you wanted to ask what this means for other roles

Matt: did you get a sense if this effects DPUB-AAM mapping to get the update out that is hung up
… wanted in parallel with EPUB
… can we continue to work on that update?

Charles: no other hangups raised so seems good, but if there is a specific issue we should check on let us know

Matt: hoping we can move forward with the ARIA module to CR. Maybe need to take up separately

Gregorio: Seems Apple will update the mapping internally. Upgrade to main ARIA roles is great for us and should get better implementation

Charles: table with AX API with AX sub-roles for landmark region or nil -- James agreed that nil could be changed to biblio-entry or whatever
… others should be able to map as well

Avneesh: objective is to update the AX API in the mapping but Apple needs to work internally first

George: Testing Android and Chrome, data was filled in for description. Is sub-class also filled in?

Charles: We didn't want to touch the description field. James agreed. There are localization issues.

George: TIES call should take a look at this in detail

Tzviya: With roles and subclass/superclass, there is a bigger issue with ARIA superclassing. It doesn't work the way most of the ARIA group thought it should work
… inheritance structure is not clean

Matt: it is a theoretical model and doesn't work with all nodes. Tread with caution. Things may not inherit all properties.

Avneesh: things like list parent?

Matt: not the AAM mapping

Avneesh: We focus on AAM mapping, not main spec

Update from meeting with APA at TPAC.

Charles: The ADAPT is a new personalization module that is close to going to CR but the TAG has major concerns
… we thought we answered everything so we though it was resolved, but actually it was closed because of lack of activity
… so issues still need resolution
… they love the mapping of symbols to phrases using the W3C registry
… Can tag things in web documents with numbers and they can be converted to Bliss symbols automatically
… Other areas where they had concerns is priority levels like critical to page vs medium priority -- reduce complexity on page for users who are distracted
… more work needed in that area
… were three modules, now four, but bring Symbols module to CR first and then go back to others

George: Synthetic speech: APA wants more features that were supported in the proposal from the Pronounciation task force
… interlinear speech also wanted, where there are two things and you need to pronounce both or select which one
… they want to enhance spec and want only one option while the current spec picks between two
… not sure they mean one attribute only or one option

Makoto: The current doc has two approaches. One uses only one attribute but it has rich structures
… other approach is multi-attribute approach
… EPUB has been doing multi-attribute approach.
… Now they have three but I don't know what the addition is
… but they said they will pick one
… I personally prefer a unified solution supported by browsers, online tests, and textbooks

Matt: They said they'd discussed with WHAT WG and had good response, but that they had some internal issues within their group
… if they have buy-in, that's a sign of movement in the near future

Avneesh: May take years to get implemented even with browser buy-in
… Wendy has action item to talk to publication community about interlinear text
… APA working group said send change log and we will decide if review is needed

Guidance document for accessibility summary

George: Accessibility summary has been updated

<GeorgeK> https://htmlpreview.github.io/?https://github.com/w3c/publ-a11y/blob/main/drafts/schema-a11y-summary/index.html

George: I updated several things.
… Example section: added one thing and limited to 500 characters

<Bill_Kasdorf> Is that correct?

George: removed priority column from features table. It didn't seem helpful
… added text to Hazards based on discussion. What it is, where located, is it essential content. Added example
… questions about features: 26 rows. Some are strange for EPUB
… transformability, large print, high contrast. To me, not going to be used. Do we want to remove them?

Madeleine: is this doc for publishing community only? If not, maybe note that some rows are not as useful for books but are useful on the broader web

Naomi: Some of these might not be relevant for all books, but could be for specific book formats.
… Or does it go up a level to the parent description, not the summary
… the metadata display document -- User Experience Guide

Charles: some might not make sense -- if we have nothing to say in the suggested language column, why are we including it
… could have a collapsed detail section about less popular items

Naomi: or just note it is not exhaustive

George: will add a link to the list of all features
… then can remove ones that are built into EPUB from this document
… concerned about children's book issue that Naomi raised, though

Tzviya: I disagree with others -- publishers struggle with summaries and the ones they have aren't helpful

<Bill_Kasdorf> +1 to Tzviya

Tzviya: having the guidance for what to say for each value would be really helpful for publishers to automate useful summaries

Avneesh: we have adjusted thinking on Summary

Matt: Schema.org was not thinking of EPUBs. It was more general. Summary not meant to be repetition of what is in the other metadata.
… you could make it a duplication, but we are trying to avoid that now in this guidance document

<Bill_Kasdorf> Tzviya's strategy would be an option; publishers that could write a good summary could keep doing that.

Matt: for EPUB where we are requiring this to be present, we know more what to expect in the metadata, so we can tailor the summary to what isn't conveyed in the metadata
… correction "requiring other metadata to be present"

Charles: some rows still seem to recommend duplication. But clarification might be needed for captions (video, not fig captions)

George: Will work on this, moving information into the summary to match table
… I will remove some rows, make nice crisp text, and explain what we mean such as "all math is in MathML" so you can bank on it

Avneesh: Chemistry can also be represented in MathML so let's add that example

George: is anyone has comments on the new Hazards text, please let me know
… Makoto, I need help with Ruby. If we use Japanese characters, what would the character limit be? Can you help with an example summary in Japanese

Makoto: Character limit not huge problem. Adaptation is not supported by reading systems. We need to promote Ruby rendering based on user preferences
… if this book is expected to support both thoroughly annotated, partially annotated, and no Ruby, reading systems may not support this variety.
… users may be disappointed by their experience based on info in the summary

George: this is true for math also, if the RS doesn't support the features in the book

Makoto: Systems can generate human readable info from the machine-readable metadata.
… this could be automatically adjusted based on what the RS actually supports

George: I thought vendors would display metadata directly and not change it based on their implementation
… if they don't support mathml, they shouldn't hide metadata that says the book has MathML. Same for Ruby

Naomi: For features we know aren't supported, publishers want readers to know the feature is there, but if the platform doesn't support it
… it is misleading to say it is supported
… leaving it in machine-readable metadata rather than summary lets vendor trim the metadata to what matches the RS tools
… but the vendor shouldn't mess with the text of Accessibility Summary

Avneesh: this topic needs longer discussion

Naomi: when we do accessibility testing and something is a bit strange, like acronyms or other odd reading results, should we mention that
… or is it common knowledge that acronyms will read poorly
… there are ways to force a fix, but if we haven't should be mention it?

George: too much detail! don't put that in summary

Avneesh: Move this discussion to the Summary task force next week

Charles: Might have ascii text for simple equations and mathml for complex
… also need to update acknowledgments of participants

Further Feedback for MARC crosswalk.

George: Charles and I will get together and work on this

<AvneeshSingh> https://w3c.github.io/publ-a11y/drafts/a11y-crosswalk-MARC/index.html

Chris Oliver: Update on MARC: delays due to back to school season
… interesting to hear the publisher input from the beginning of the production change since libraries are at the end
… when feature is in the EPUB but the reader can't offer it, is it still in the EPUB file?

Naomi: Some publishers may be making custom EPUB for different vendors, but most make one EPUB for all systems

Chris: So users can't use the feature in their tool but it is in the file
… when an EPUB is being prepared, and you get the manuscript from the author, is it is the publisher who decides on a11y features?

Naomi: different at different publishers. At Penguin/Random House, we have no contact with authors or editors, just get the file and move it to EPUB
… but in some cases authors and editors are getting more involved
… smaller publishers will have fewer people involved and may have more conversations about this

Chris: So do extended descriptions go back to authors for writing?

Naomi: in fiction publishing this comes up less. But it varies, can be author or epub team or others. Completely different across industry though

Charles: some publishers have conversion vendors and may pay that vendor to do descriptions.
… if going for conformance they may be paying for it then

Charles: for how we display metadata, may want to add that if a vendor doesn't support a feature, they could note that features exist in book but are not exposed in this tool

Gautier: Image description is a major problem for publishers facing European A11y Act. French publishers association has a document about it.
… varies by publisher, but because of EAA and back publication list needs, it is more important now
… some conversion can be made by conversion team and some in house.
… Big issue in Europe for publishing industry

Avneesh: Further discussion in the future. Thanks all.

Minutes manually created (not a transcript), formatted by scribe.perl version 192 (Tue Jun 28 16:55:30 2022 UTC).

Diagnostics

Succeeded: s/API/APA

Maybe present: Avneesh, Charles, Chris, George, Gregorio, Makoto, Matt