EPUB 3 Working Group FXL A11y Task Force Telco — Minutes
Date: 2021-04-06
See also the Agenda and the IRC Log
Attendees
Present: Wendy Reid, Ken Jones, Gregorio Pellegrino, Laura Brady, Matthew Chan, Shinya Takami (高見真也), Matt Garrish, Laurent Le Meur
Regrets:
Guests: Dave Cramer
Chair: Wendy Reid
Scribe(s): Matthew Chan
Content:
Wendy Reid: there’s no agenda today because TF meetings are less formal
… i’m hoping for us to review the wiki that was posted some weeks ago
… and then decide on a path
… our TF’s role is not to change the spec, but to produce a WG note that we can use to help people produce better FXL content, as well as to inform possible spec changes in epub 4, etc.
… the note should be addressed to RSes and content creators
Gregorio Pellegrino: about the wiki page
… i think we should split the document based on the different ways you can create an FXL
… as far as i know you can have rasterized images in the spine (which are fully non-accessible)
… vs putting SVG in the spine (where there are different ways of adding a11y)
… vs having fixed layout HTML in the spine
Wendy Reid: you can’t put images directly in the spine currently
… but we see a lot of books where the pages are just single images
… one of the challenges here is that there are so many types of FXL content
… e.g. this slideshow style is very common with comic books and books that are very design heavy
… or textbooks, or MO style
Gregorio Pellegrino: http://kb.daisy.org/publishing/docs/fxl/img-spine.html
Dave Cramer: about the accessibility of the RS themselves
… if there was an FXL epub that was image-based, but the images had alt-text, do most RSes expose that alt-text?
… if we did an FXL book with beautiful accessible HTML text layer, how much of that would get surfaced in most RSes?
Shinya Takami (高見真也): as you may know, in JP there are lots of FXL ebooks (e.g. for manga)
… this is a topic we are exploring currently
Wendy Reid: i haven’t tested FXL in most RSes, but alt-text gets surfaced in reflow books
Laurent Le Meur: if alt-text is good, screen readers will pick it up (e.g. NVDA)
… where there is a mix of images and text, alt-text may not be discrete enough to be picked up
Wendy Reid: https://github.com/w3c/epub-specs/wiki/Fixed-Layout-Accessibility
Wendy Reid: i don’t think there will be a single practice for FXL since there are so many types of content that are being turned into FXL
… there will probably be several solutions based on the type of content
Dave Cramer: do we know if there have been efforts towards a11y for web comics? Maybe something we can learn from the HTML world there
Wendy Reid: the only comic i’ve seen is the thing from Human (the comic-a11y example project)
Laurent Le Meur: the only thing i’ve seen is from DeMarque(?) I will send a link to those interested
… they are looking for better solutions because at the moment they must have a heavy editorial phase to get it working
Ken Jones: in my experience i don’t think alt-text is read out in regular readers for FXL, its skipped over
… another approach is putting live HTML text in the page, but hiding behind an image of the page which might retain the more complex layout of the print page
Laura Brady: we make a lot of FXL at our publisher, and we’ve seen that alt-text gets read by AT but not necessarily in RS
Ken Jones: might this be good default behaviour? Not everyone will want alt-text read out
… or at least make it an option
… there’s also a question of whether alt-text should be read out if the RS is in TTS mode
… perhaps in TTS mode default behaviour would be to read out alt-text?
Gregorio Pellegrino: in automatic TTS tech you can set it up to read out alt-text, but it requires a little bit of code
… but those tools are not used by blind people, but by people with dyslexia
… so there are two target markets
… focusing on different technologies
… and usually those TTS solutions don’t speak out the semantics of the page (headings, etc.)
Laurent Le Meur: Note sample, accessible comics: https://m.nota.dk/embedded/43210?guest=true
Wendy Reid: https://comica11y.humaan.com/
Ken Jones: i’ve posted a couple things in the past about picture books with double page spreads
… i.e. how to alt-text it when the same image applies in both HTML documents
… also a lot of illustrated books aimed at older children have montages of multiple images overlapping and blended with one another
… how should we describe those?
Wendy Reid: also, webtoons don’t do alt-text either (and they do images of words, unfortunately)
Dave Cramer: just to check my understanding, given the different types of FXL, some of the US publishers do FXL that is still very text heavy, but my amateur understanding of a11y is that the foundation is still having semantic HTML
… but FXL is often weird CSS
… though there is nothing preventing the underlying HTML from being semantic
… so the other big issue is how are such FXLs produced?
… often with some upstream tool
… i’ve seen raw export from some of those programs, and it is a little hard to parse for a human
… so how much progress have we made in getting passable HTML out of INDD, for example?
Ken Jones: is this group going to recommend the best way to work with well known tools?
… most of what we can do in reflow we can also do in FXL, it is just that it is currently much more time and work intensive
Gregorio Pellegrino: we know that the FXL export from INDD is similar to PDF
… words are individually positioned, tables are text on the page with the bottom of the table as an image
… so we should not focus on these
… also, the use of spans isn’t often picked up on by AT
… so well structured HTML that makes heavy use of spans still isn’t great
Wendy Reid: this document is going to come down to best practices. We shouldn’t go too far into giving software specific guidance. Don’t want to be taken the wrong way.
Dave Cramer: it sounds like we need to be in the advocacy business… that we have a good sense of how to construct an a11y FXL epub, but that the commonly used tools don’t do that
Gregorio Pellegrino: +1!
Dave Cramer: i was thinking that it would be cool for us to make some samples of super high design FXL that are also quite accessible
… publish it side-by-side with what common tools like Adobe currently output to document the gap
… use this as a way to approach the various authoring tools
… to start some change
Laura Brady: going back a little, I know that adobe is not currently working on improving the FXL export tools
… i had a meeting last week with WIPO and they are talking to Adobe engineers to improve their export too
Ken Jones: my CircularFLO is a tool to export FXL epub relies on INDD
… i would love to help make a11y books (it doesn’t have to be my tool though)
… so I am offering
… could maybe make a selection of different types of FXL (cookbook, picture book, etc.)
… because we’re all used to looking at code, the code in Adobe INDD FXL looks bad to us, but as long as it works and is accessible, I don’t necessarily see a problem
Wendy Reid: so i think we know what the problem is, and some of the areas we want to tackle
… samples are a great idea
Dave Cramer: would it make sense to do some of this in conjunction with the Daisy KB?
… there is some epub FXL info there, but its not highly detailed
… no idea how Daisy sees that info, but it seems like the place people are already going for this sort of info
Gregorio Pellegrino: +1
Dave Cramer: we might be able to get some volunteers to build it out with common a11y pitfalls
Ken Jones: +1
Dave Cramer: might be able to reach a broad community without reinventing a bunch of stuff
Matt Garrish: that’s the ideal, yeah
… that’s why we have it in github
… a lot of the issues we’ve talked about today hasn’t been enumerated for us, we’re not aware of the granular issues
… but we’re interested in helping to find whatever solutions are necessary
… e.g. for tables, i’m thinking of using ARIA labels on top of the generated output to start layering in a11y
Ken Jones: i’ve started listing problems and potential solutions in this document
… i haven’t tackled tables and hierarchy in this doc yet
Wendy Reid: very helpful
Gregorio Pellegrino: what about partially sighted people?
… PDFs are an issue because font size, spacing, etc. cannot be changed currently
… all of these benefits are lost with FXL
… those PDFs can be “validated” with WCAG, and so can FXLs
… also, we should mention contrast requirements (at least foreground and background)
Dave Cramer: there have been people who have brought up finding a middle ground between full FXL and reflow epub
… e.g. something like FXL, but with viewport set to 100vh x 100vw
… and allow reflow within the constrained page
Gregorio Pellegrino: responsive design? :)
Dave Cramer: not really sure how that would work in practice
… but CSS has also progressed quite dramatically since the FXL spec was originally released (e.g. flex, grid, etc.)
… probably the fewer FXLs that exist in the world the less of a headache they will be
Ken Jones: i recommend that people don’t make FXLs, but i think there’ll always be the need for that sort of highly designed, magazine-style content
… two things
… PDF seems to be alive and well and not going away
… the INDD design people interested in a11y seem to be talking about a11y in PDF
Gregorio Pellegrino: I produce both accessible EPUBs and PDFs :)
Ken Jones: and the other thing is that rather than trying to massage fixed layout content into a reflow/semi-reflow document, could we not generate a secondary reflow output?
Matt Garrish: I think dauwhe’s approach is what the PDF people have been thinking about as a way of complying with the latest WCAG
… having that HTML become reflowable as you zoom would help us for those low-vision users
… we have to be careful not to lump the concerns of blind users with those of the low-vision users
Wendy Reid: we have a couple things to work on
… we could look at Daisy KB and say what is not here and what needs updating?
… based on what someone who is producing FXL might want to know
… we are talking about future ways to do FXL better, like what can we do with grid, flexbox, etc.
… might be experiments we can run with this new CSS
… and then just distilling some best practices and suggestions for people producing FXL
… like making sure your reading order is correct
… and when FXL is necessary, vs when content can really just be turned into reflow
… use of mixed modalities is currently under-implemented by RS, and therefore not used by publishers, and so on and so on
… but mixed modalities is kind of a middle ground
Matt Garrish: we should also look at improvements that can be made using existing spreads using ARIA, etc.
… ARIA flow might be useful within a multi-page context for example
… though might not be well supported in RS (or at least not as well supported as in AT)
Gregorio Pellegrino: in DAISY we are testing DPUB aria in different browsers… very low support
Wendy Reid: those threads (e.g. Daisy KB, best practices document (starting with Ken’s doc + some ARIA stuff), and fun experimental stuff)
Dave Cramer: also documenting things where we’re not sure what to do
… even just identifying the problems is useful
Wendy Reid: absolutely
Ken Jones: about the tag hierarchy, i’m unsure how valuable that is when each page is a separate entity
… e.g. does the hierarchy restart at the end of each page?
… just not something I’m an expert in right now
Dave Cramer: this is also a problem that HTML has not faced
… each is generally its own DOM, its own a11y tree
… it sort of seems that if a screen reader is identifying headings, then as long as you are avoiding nesting issues, then you are fine
Matt Garrish: this is one of the big problems with FXLs, when you break things across pages
… e.g. how do you break a table across pages?
… it also makes trying to navigate these sorts of documents unenjoyable for users
Gregorio Pellegrino: i tried to validate FXL using ACE, and I didn’t get any errors for the FXL itself (i.e. the viewport or other things)
… we should try to create a FXL with zero errors in ACE, at least
Ken Jones: that’s definitely possible, as long as its a valid epub with alt-text then it should be valid (even if the alt-text is trivial)
Wendy Reid: i’ll send out a summary email to this group
… we’re going to need a couple more meetings to figure out what to do about this
… and who has the bandwidth to do some of these things
… please keep in mind these topics, and which you are interested in contributing to
… does this time work for everyone? Or should there be a separate Doodle?
Gregorio Pellegrino: let’s run it by george
Wendy Reid: will do
… thank you all