EPUB 3 Working Group FXL A11y Task Force Telco — Minutes

Date: 2021-04-13

See also the Agenda and the IRC Log


Present: Matthew Chan, Deborah Kaplan, Wendy Reid, Gregorio Pellegrino, Laura Brady, Ken Jones, Shinya Takami (高見真也), Charles LaPierre, George Kerscher, Brady Duga, Rachel Osolen, Will Awad



Chair: Wendy Reid

Scribe(s): Gregorio Pellegrino, Matthew Chan


Wendy Reid: welcome everyone to FXL TF
… last week we decided there were 3 main areas that we wanted to work on
… updating Daisy KB to reflect more current FXL production and provide better examples of what is done in the real world
… write best practices around building accessible FXL
… and experimentation around modern CSS (e.g. grid and flexbox) in epub that leads to better a11y
… we need volunteers and a plan for tackling those

Ken Jones: i’m happy to be involved in the first two

Gregorio Pellegrino: i would like to make some experiments with FXL, would be happy to help with modern CSS in FXL
… including modern HTML and ARIA etc.

Wendy Reid: the business group is also collecting some use-cases around mixed use of fixed and reflow content in epub
… i’ve asked for updates on this so as not to overlap work too much

Laura Brady: I am happy to help as well.

Will Awad: one of my team is joining today, veejay, and we’re happy to join

George Kerscher: so i can envision techniques to make these accessible, except for low vision
… i don’t see how it can be done unless there are alternative version available
… if enlarging means having to pan left and right, that’s not accessible

Charles LaPierre: on that note, i’m happy to help test it
… CSS grid and flex might be able to help
… i.e. having it reflow so that the content can be scrolled vertically
… also as part of our global certification, I can make sure that our best practices here will get sent out to all our certified publishers and conversion vendors
… also able to help with Daisy stuff

Brady Duga: one thing we could add is where the text in these books is, and how it is laid out
… we can then zoom the text only in these areas (e.g. the bubbles in a comic book)
… not sure if anyone would actually take advantage of this though
… might work better for some genres (i.e. cookbook) than other (i.e. illustrated children’s books)
… so its not necessarily a technical limitation as much as a buy-in limitation

George Kerscher: if you could go into a mode where you’re walking the text, tab from one text object to the next, while having those enlarged, then each one could be enlarged, spoken, and brailled at the same time

Brady Duga: that can solve the problem
… you get issues where you have a line of text that already stretches the length of the screen, so you’re not sure how to reflow it
… but yes, that’s what we do for our comic book reader

Wendy Reid: regions navigation was a good idea that never got implemented

Brady Duga: except for our bubbles implementation

Wendy Reid: some RSes have done it on the RS side without collaboration with authors/publishers

Charles LaPierre: if there were alt-text/ARIA labels, couldn’t you use those instead of source text?

Brady Duga: comic books are often pure images, so there’s nothing to stick an ARIA label on
… you could create objects to attach those labels to, in theory
… one publisher created a font that is a replica of their hand lettering, an idea

George Kerscher: the EUAA will not allow FXL that are not accessible to be sold
… the legal hammer is probably going to have an impact on what gets done
… we have opportunities here to come up with technical solutions that people will embrace

Wendy Reid: timing-wise we’re in a good place
… i think in terms of the tone of our best practices, its worth reminding people that a lot of things don’t really need to be FXL
… we can provide ways to do things in FXL but also some alternatives (e.g. flexbox even in reflow content)

Ken Jones: i’m a fan of the approach of having more than one output (e.g. an FXL version marked with the appropriate Schema warnings), and then have a maximally accessible 2nd version

Gregorio Pellegrino: 10 years ago i did some experiments with FXL using only responsible CSS (so not flexbox etc.), and it worked on some RS but not on others
… that could be a quite simple way that we can test

George Kerscher: 25 years ago people would put up a separate website for the blind
… but it was a total failure
… and lead to clear recommendations that that was not the right approach
… having a version for people with disabilities probably would not be accepted
… though it would work
… and now with better metadata it would be discoverable
… and under EUAA you couldn’t sell the inaccessible version unless an accessible version existed

Wendy Reid: this kind of falls into the alternative version idea, and there are already organizations that do this sort of work

George Kerscher: back then my idea was that multiple renditions would be used to package up fixed and reflow versions of the same content in one package

Ken Jones: i’d be happy for the more accessible versions to be in the same package
… if the alt-text etc. gets worked out, i can see how we could generate a more textual version
… i wonder if it makes more sense to think of the versions as textual vs visual
… seems to tie more into the Schema accessModes

Wendy Reid: there’s a reason we’ve removed multiple rendition from the core spec
… we need these recommendations to reflect how publishers, RSes, do books today
… asking for a publisher to create multiple versions of a title aren’t going to work
… that’s why we have conversion houses
… any help we provide should steer people towards making a single better version
… with push behind born accessible, it seems like publishers are more interested in providing a single better version

George Kerscher: so we do not want to rely on alternate version producers to provide accessible versions of these products
… we should be figuring out the best solutions and practices that will allow AT to get at the underlying information
… to me that means that the image representation of text is pulled out by alt-text, ARIA, whatever, and that those strings can be reflowed and reordered

Wendy Reid: we can’t change the spec, so what we recommend will be best practice, code samples, example book

George Kerscher: and we have to identify two mainstream implementations that can render that content

Charles LaPierre: the other thing i was going to mention is that once we have some options available we could create a test book for epubtest.org for testing on a number of different RSes

Brady Duga: since we’re not changing the spec, two implementations aren’t actually required
… no normative statements

Wendy Reid: we’re also not having anything to do with testing HTML or CSS, so yes

Ken Jones: sometimes when you present an FXL to RS that isn’t able to display that content, it gives you a stream of modified content
… maybe that’s what we should be thinking about
… a different way to view the same epub

Wendy Reid: a RS that is doing that isn’t really following the spec, which requires that FXL work
… but its better than breaking completely since it is, at least, a kind of fallback

Gregorio Pellegrino: agreed re. the need for implementations, but since FXL support is inconsistent, I think we should at least map whether our recommendations work across RSes

Wendy Reid: if we encounter situations in our examples where our recommendations don’t work, might be work leaving a note for the RS implementors

George Kerscher: we see in education that there are organizations like Bookshare and Access Text Network, that if there is a commercial version available that is DRMed, then those versions can be made available DRM-free
… if we could identify RSes that are better for accessibility than others that would be useful
… people could be directed to those RSes
… and also give other RS developers incentives to do in those directions

Wendy Reid: maybe we could work with epubtest once we have those best practice examples
… if our recommendations meet EUAA, then yes, this would give RSes and content producers incentives if they want to continue operating in the EU

George Kerscher: for epubtest we’ve decided to come out with new titles to test experimental features
… i’m in the process of creating a 90-day plan for epubtest
… should I include an experimental FXL book in that plan?

Wendy Reid: yes, i think so
… and i can help

Gregorio Pellegrino: the test books on epub test are mainly for testing if features work, not much content in them
… but yes, we could have an FXL one

Wendy Reid: for the Daisy KB update and the best practices, we have volunteers
… the volunteers should be put in touch with mgarrish
… volunteers please review what is currently in the FXL section of the KB

Wendy Reid: http://kb.daisy.org/publishing/docs/fxl/index.html

Wendy Reid: think about where we can expand on that, and come up with a plan to address those gaps
… then share that with the rest of the group please
… for best practices we have the doc that Ken shared
… we can convert that into framework
… i’ve talked to mgarrish and ivan about setting up a folder on github for that purpose
… agreed?
… and we’d want these back by next week?

Will Awad: we’ll do our best

George Kerscher: Marisa uses github to manage the books in the epubtest repo
… you can just put your test FXL in that same repo

Ken Jones: would it be a good idea to get some content early on
… to compare against what already exists
… if so, where would that content come from?
… do we need samples from publishers?

Wendy Reid: we’d need to find some, or make so, yes

Brady Duga: one of the problems is that there are so many esoteric use-cases of FXL
… there isn’t really a representative FXL book
… sometimes its all pictures, sometimes its SVGs, sometimes there are images and text overlaid
… we may need more than one sample book

Wendy Reid: we’ll need at least one sample of each representative archetype
… kids book, comic, recipe, coffee table, etc.

Will Awad: might be able to get a sample FXL, but it wouldn’t be in English. OK?

Wendy Reid: yes, that would be great
… don’t want to fall into a trap of making our recommendations anglo-centric

Gregorio Pellegrino: do we have statistics on the usage of FXL?
… are they more used by children, etc.?
… to help us narrow down the 5 or so use-cases to start with

Ken Jones: does the content need to be totally public domain?
… publishers probably want to volunteer, but might not want to share their content forever…

Wendy Reid: it doesn’t have to be public domain, but it’d have to be postable on a github…
… if we could use public domain content then we don’t have to worry about any of that

Brady Duga: i have access to content that my parent company owns, and we’re under strict licensing, so we can’t contribute it
… and the types of FXL content most commonly used will depend on territory
… e.g. in JP it would be manga

Wendy Reid: for us it would be kids books, comics, textbooks, in that order

George Kerscher: one approach would be for publisher to give permissions for use of a small portion of a copywritten work

Brady Duga: we could ask publishers for silly demo books
… i.e. we give them content that they run through their tool chain

Wendy Reid: good idea

Ken Jones: we want the examples to be professional content
… maybe we could get extracts?
… i’ve written a children’s book, and might be able to clear rights to the illustrations…

George Kerscher: this is a good opportunity to have the CG working with the business group, we can ask the business group for help here

Wendy Reid: yes, I can ask
… maybe ask for the same amount of content that a publisher would agree to post as a sample
… is everyone clear on what we need from them for next week?

Ken Jones: so I’ll look through the KB?

Wendy Reid: yes, i’ll send summary email along with minutes after we end
… thank you, we’ll see you again soon