EPUB 3 Working Group A11y Telco — Minutes

Date: 2021-08-12

See also the Agenda and the IRC Log


Present: Avneesh Singh, George Kerscher, Matthew Chan, Ben Schroeter, Vijaya Gwori Perumal, Murata Makoto, Will Awad, Charles LaPierre, Gregorio Pellegrino



Chair: Avneesh Singh

Scribe(s): Matthew Chan


Vijaya Gwori Perumal: 16+ years in digital publishing. Hope to support this group with my technical expertise.

1. Pull request for Media Overlays section: all meaningful text should also have audio.

See github pull request #1759.

See github issue #1738.

Avneesh Singh: the background is that in MO there was no specific requirement that you should have audio for every meaningful text element
… for a11y this is very important because MO provides experiences like screen reader
… the PR above addresses this need

George Kerscher: whenever I play MO I start it and let it go. I expect that everything that should be presented to me is presented. I’d never think to go back and look for content that is skipped. So this change seems perfect.
… if there is a portion of a book that is presented as MO, then that portion should be self-contained

Avneesh Singh: these requirements are informative so as not to break backwards compatibility
… we are informing publishers that this is our expectation
… in the future we may make this normative (i.e. future version of spec)

Proposed resolution: approve issue #1738 (Avneesh Singh)

George Kerscher: +1

Will Awad: +1

Matthew Chan: +1

Murata Makoto: +1

Vijaya Gwori Perumal: +1

Resolution #1: approve issue #1738

2. Start review of EPUB Accessibility specs and techniques.

Avneesh Singh: Specs: https://w3c.github.io/epub-specs/epub33/a11y/index.html

Avneesh Singh: in Nov we go to CR stage. The EPUB a11y spec is complete. Now is time to do a thorough review.

Avneesh Singh: Techniques: https://w3c.github.io/epub-specs/epub33/a11y-tech/index.html

Avneesh Singh: Issue tracker: https://github.com/w3c/epub-specs/issues/

Avneesh Singh: please go through the spec and techniques documents. If you find anything, please file issues in tracker
… proposed deadline for issues of 8th of September?

Murata Makoto: you said that we are going to create CR, but is that for epub a11y 1.1, or for techniques as well?

Avneesh Singh: CR is only for epub a11y 1.1
… no objections to proposed deadline?

Murata Makoto: No objections from me.

Avneesh Singh: okay, so we’ll set that deadline at Sep 8th

3. long standing topic of 100% a11y requirements

Avneesh Singh: Historically WCAG is pass/fail for A or AA level, etc.
… this zero tolerance policy has been challenging
… this will be addressed in WCAG 3.0, but that is far in the future
… is there anything we can do to help this situation, or is it better to leave to WCAG?

George Kerscher: I think its important for us to make it clear that when a publisher is claiming epub a11y to level A or AA, that if there are occasional errors that it does not violate the conformance claim
… not sure where we should do this, but i think we should do it
… worried that publishers could be sued for a single error in, for example, a 750 page textbook

Charles LaPierre: i’m wondering about how we’re doing the conformsTo examples. Could we do something there?
… could we add “strives to “ before the conformsTo text?
… OR we could add something to the a11y metadata to the same effect

George Kerscher: I’ve heard people start to say “largely conforms to…”

Ben Schroeter: publishers should be given leeway as to what they are saying they conform to
… we were thinking of “significantly conforming to…”
… the “striving” language could be interpreted to mean that the publisher will conform in the future

Avneesh Singh: so are we thinking of adding more combinations of possible text for conformsTo?

Ben Schroeter: we need a way to describe how a publication relates to WCAG, because that’s the information the schools want to know so that they don’t get sued
… the reference needs to be to WCAG rather than the epub a11y spec

Avneesh Singh: where legislation requires WCAG AA, the spec would be helping, wouldn’t it?

Ben Schroeter: we want to be transparent and accurate in the conformsTo statement, but also to be responsive to market concerns
… maybe by allowing conformance not in a binary way

George Kerscher: in the real world there are many companies and certification programs where companies advertise their conformance to a standard (e.g. hotel cleanliness, etc.) but there’s always little things that happen that cause imperfection
… i have a hard time imagining someone could be sued if a book wasn’t perfectly conforming to an advertised standard
… we like to see companies have an a11y page on the website, and a way to contact the publisher if there are issues
… if there were a significant violation found in a book, i would think the person would just inform the publisher using this method rather than go for a law suit

Avneesh Singh: to summarize, we are mainly striving to provide publisher a way to make an accurate claim. Legal issues are not our arena.
… I think we can ask mgarrish to create some strings to address this desire

George Kerscher: is there a requirement to use these exact example strings?

Avneesh Singh: no, but because the goal is to have these strings easily parsible, that is why we’re giving examples

Charles LaPierre: actually, spec says the value for conformsTo “MUST exactly match” the example strings

Avneesh Singh: we should review whether it makes sense to have a MUST here

George Kerscher: I like the exact text strings, but I think we should clarify that we mean significant conformance to whatever level is being referred to

Charles LaPierre: I also support keeping the MUST. Don’t want this field to be the wild west.
… but we could have qualifications to what these statements mean

Avneesh Singh: but would allowing additional qualifiers make it difficult for anyone trying to parse the field?

Charles LaPierre: not sure that it would. There could be multiple strings in the conformsTo statement. One will be the one that matches exactly. And there could be additional qualifying statements.
… as long as one of the required strings is present, it should be fine
… so the qualifying thing could either be a refines element, or just separate from the MUST statement with a space

Gregorio Pellegrino: what do we mean by “significantly”?

Avneesh Singh: so do we mean that we will also provide text strings for the qualifying statements?

Charles LaPierre: we could provide those after discussion

George Kerscher: is there an opportunity in the accessibility summary to resolve this issue?
… i believe that in the spec itself we should be making it clear that inadvertent errors do not invalidate claims
… to gpellegrino’s point, if the publisher leaves out MathML etc. this would be a clear violation. Spelling errors, etc. would fall into the category of inadvertent error
… we’re trying to make it so that little tiny mistakes don’t fall afoul of claims

Avneesh Singh: we should try to define “significant” somewhere
… but this is good enough for mgarrish to start drafting
… but there is lots for us to discuss in the issue tracker
… okay, the next call will be 2 weeks from now. We will now continue our bi-weekly calls.

George Kerscher: to meet our september deadline, I think we should continue regular meeting schedule

Avneesh Singh: thank you everyone for being here! See you all on the issue tracker.

4. Resolutions