EPUB 3 Working Group Telco — Minutes

Date: 2021-01-15

See also the Agenda and the IRC Log


Present: Matthew Chan, Juliette McShane, Avneesh Singh, Ivan Herman, Masakazu Kitahara, Wendy Reid, Toshiaki Koike, Tzviya Siegman, Gregorio Pellegrino, Ken Jones, George Kerscher, Ben Schroeter, Charles LaPierre, Matt Garrish, Hadrien Gardeur, Bill Kasdorf, Laurent Le Meur, Teenya Franklin, Brady Duga, Garth Conboy


Guests: Dave Cramer, Bert Bos, Neil Soiffer, Bruce Miller, David Farmer, Steve Noble, Sam Dooley, Deyan Ginev, Louis Maher, Sydnee, Brian Kardell, Murray Sargent

Chair: Dave Cramer

Scribe(s): Wendy Reid


1. MathML CG/WG

Dave Cramer: We have a last minute change to the agenda
… members of the soon to be MathML WG visiting today
… their charter is being written and reviewed
… we thought this would be a good time to meet with the EPUBWG and share information on the scope and nature of the work
… there’s many publishers that use MathML
… and math conventions
… it makes sense to share information on our use cases
… I am very excited by this new energy around math

Neil Soiffer: Two main prongs to the charter

Dave Cramer: https://mathml-refresh.github.io/charter-drafts/math-2020.html

Neil Soiffer: MathML Core: Igalia is leading this work
… getting MathML into browsers
… integrate with the existing tech stack
… we are running into the constraints of the browsers
… other prong is to improve accessibility, search, and computability
… allowing author intent
… disambiguate what something means
… if you had vertical bars around content, it could be an absolute value, or something else
… screenreaders can guess, but its a guess
… we want to allow authors to clarify that meaning
… that’s the high level overview
… open to questions

Brian Kardell: Just wanted to correct one thing, the current and near-future of browsers are the target of MathML Core level 1
… as things grow and change, we can grow with them

Bill Kasdorf: Thanks for joining, do you want to make any comments on backwards compatibility

Neil Soiffer: There are things that aren’t in Core
… there is a level 2 planned
… we want to get something out now instead of waiting
… there are polyfills for the differences
… there’s a few things in MathML that are obscure and never implemented … there are polyfills for things not in core including some of the MathML features that were deprecated. … many things still work, but some javascript may be required
… the idea of core is that the stuff people use will be supported
… fractions, square roots, etc
… line breaking is in level 2
… mfenced is going away to be replaced by a polyfill
… if you’re willing to use those, it will be the same
… for Core, it covers about 95-98% of what you want

Brian Kardell: We based a lot of this on what MathML is in use in the market

Neil Soiffer: Bevelled fractions, for example, I thought it was common, but in the research we found it’s rarely used

Tzviya Siegman: I have a few questions
… first about authoring
… are the workflows maintainable?

Neil Soiffer: David Farmer is on the call, he works on PreText

David Farmer: The Math in PreTeXt, designed to convert to any output, PDF, HTML, EPUB, braille
… the way we do HTML is to use PreTeXt and then add mathjax
… we are working on an easier way for authors to write content
… many textbooks in math
… learning LaTeX is a hurdle, we want to make it simpler
… for accessibility we want to simplify

Tzviya Siegman: I don’t want a new tool
… if an author wants to write in LaTeX, they will
… what would be helpful for me (Wiley Scholarly), is to talk to us directly
… how to incorporate into the tool chain

Brian Kardell: The goal is not to lose author information
… flow from author, to latex, to the screenreader
… if the author has some intent, let them express it

David Farmer: It’s still LaTeX, but we’re considering the steps between

Neil Soiffer: Please get in touch

Brian Kardell: In terms of existing tool chains, things should work as before if not better

Ivan Herman: Two questions
… one: in terms of the standard and implementation, are you keeping the content markup and presentation markup?

Neil Soiffer: We have no intention of doing anything to content
… presentation and content MathML will continue to exist
… dealing with author intent, incorporates all of content MathML
… threads the content tree
… that may at some point replace content MathML, for now we are keeping it

Ivan Herman: Other question is practical
… can you speak to the timeline?
… when do you plan to publish the new recommendations, and implementation?
… this WG relies on what you do, as do the reading systems
… they would love to incorporate core

Neil Soiffer: Going from memory, roughly Core will be out first
… mostly already written, another year or so to release
… CR in 6 months
… we’ve been exploring other ideas, and the publication schedule is closer to the end of the charter
… math player accessibility, working on the replacement for that now
… developing as we finalize intent
… NVDA will pick up all of that
… as will ORCA, possibly JAWS, I have asked
… Apple is a mystery
… Also spoke to the mathjax folks, the core work doesn’t replace their work, but it might result in a new backend renderer, faster
… likely won’t switch until browsers are implemented

Ivan Herman: What can this group do to help?
… help with testing, review, support?

Neil Soiffer: I don’t think there’s anything before going to charter, aside from sign off
… but as we get into development, comments on publisher needs
… trying things out once implementations are out there

Ivan Herman: Is EPUB WG listed as a liaison?

Neil Soiffer: You’re listed

Dave Cramer: We’re doing it already!
… A lot of people who use EPUB are educational publishers at all levels
… is there a lot of support for pedagogical notation?
… I mostly work on novels
… just wondering about the teaching additions

Brian Kardell: There’s a lot of different textbooks
… there’s not elementary math

Dave Cramer: Long division with things crossed out to show places and things

Neil Soiffer: That was in MathML 3, but there’s no browser supports, it might change if there’s need

Brian Kardell: The thing with that is finding someone to express the priority and achieve them

Neil Soiffer: Fund them

Dave Cramer: If you want stuff in MathML, Brian can make them happen, with some help

Brian Kardell: It can be difficult to know what is important without people asking for them
… standards is a small group of people doing the work and owning it forever,
… from a practical standpoint, we need to know what is important
… given the parameters we have now

George Kerscher: At EPUBTest we have a math sample book
… we are reporting results
… support by reading systems and screenreaders
… visual presentation and AT support
… I can’t wait to see the changes as it’s better supported by Chrome
… we still have chem on the web in publishing group happening
… great to see the semantics of chemistry included
… great work!

Tzviya Siegman: In terms of elementary math vs other math
… this group doesn’t have elementary (primary) mathematics
… I could help get samples from Cengage, McGraw-Hill
… I have samples ranging from college to nobel prize winners
… we can give you the samples, I know you have done research, but we can provide even more
… we have a large community

Neil Soiffer: Please support the charter!

Dave Cramer: If you’re an AC rep, or ask your nicely
… bake them cookies

Bill Kasdorf: Quick, if you want elementary math, I can get the people from the conversion/pre-press houses
… some are now members

Neil Soiffer: There’s a polyfill for it, so I’d love testers

Dave Cramer: We’re here and we’re happy to help

2. Testing

Dave Cramer: moving on to the regular agenda
… I wanted to talk about testing again
… we had hoped for a testing champion
… in the short term it is me
… I have a strange fascination with writing tests for EPUB
… I have started
… I am not sure if these will be used as CR exit criteria
… but we can use them to learn about the spec
… and make it better
… I have been going through the RS spec and writing tests
… working on a script felt premature

Dave Cramer: https://github.com/dauwhe/epub-tests

Dave Cramer: https://github.com/dauwhe/epub-tests/blob/main/epub-test-notes.md

Dave Cramer: EPUB is weird and there’s a lot of interaction
… this is in a fork of the testing repo
… extracting normative statements from the spec
… write a test for it
… and add it to the list
… I’ve been running into some interesting issues that I wanted to discuss
… one thing that this has done has reminded me how much of the spec has not been implemented
… biggest example is manifest fallbacks
… at least among the commercial/trade oriented RSs
… very little support
… not supported in Thorium, Apple Books, etc
… I have tests that show RSs aren’t processing manifest fallbacks at all
… one interesting consequence is looking at the way RSs handle foreign resources
… test where I had a JSON content doc with an xhtml fallback
… I have a few RSs where they display the JSON
… throw it in a webview and see what happens
… this seems incongruent with the spec
… RSs are just trying to display the content no matter what
… I made a bunch of manifest fallbacks with XML
… they’ll show the content, a tree view
… that seems odd
… I suspect I’ll run into the same thing with linked metadata records
… order of priorities of those
… haven’t seen support for this
… I wanted to mention to the WG that we have these areas of little implementation
… another issue is that we have a number of normative statements that feel hard to test
… we had to set it up this was, a EPUB 3 RS must process the container etc
… meta safe statement

Garth Conboy: Assuming you’re right and few commercial reading systems use fallbacks
… what do you think that means for us?
… we remove it from the spec
… or the opposite?

Dave Cramer: I honestly don’t know
… it touches on the concept of the core media type
… which is foundational to the idea of EPUB
… a CMT doesn’t require a fallback
… but you also expect it to work
… this finding blurs that a little
… it’s not a CMT, but the RS tries and fails in an odd way

Tzviya Siegman: I have a similar question to Garth
… I think we need to be careful that testability isn’t the only thing to look for
… we’re going to discover untestable things
… but EPUB works
… and we don’t want to destroy something that works because it’s hard to test
… I’m not sure what the best solution is
… maybe wordsmithing can help but won’t help with the testing

Dave Cramer: It’s a good illustration of remembering our larger goals

Ivan Herman: Answer to Tzviya first, absolutely true that testability should not be the only criteria
… even in the audiobooks spec or others, where vocabularies were defined and testing unclear
… if there’s features EPUB has, and are used, they must be part of the spec, it’s on us to prove that for a feature there’s a use
… that’s a criteria we need to fold into our testing
… in the happy days of HTML5, and with versions of SVG, we did find features in both that were never implemented
… there’s some nice features people planned, but they were removed from later versions because they were never implemented
… we would have to have a policy for features that haven’t been used
… and aren’t implemented practically in any reading system
… what do we do with them
… and what do we do with the backwards compatibility
… we declare them as deprecated or dropped
… we cannot hope for every feature implemented by every RS
… but at least have reasonable coverage

Charles LaPierre: When I was wearing my developer app
… for Java we had JUnit
… testing framework
… I am wondering if we need a unit test reading system
… testing for hard-to-test features
… we need to build a test reading system maybe

Matt Garrish: I think we probably need to look at or consider what features are foundational
… look for implementation
… removing manifest fallbacks opens up the spine, which could be a problem

Dave Cramer: +1 to mg

Matt Garrish: I’m all for removing things we don’t need, but we have to be careful

Dave Cramer: That’s my concern as well

Tzviya Siegman: +1 to mgarrish

Dave Cramer: this feels foundational
… but also neglected

Ken Jones: +1 to mgarrish

Dave Cramer: I don’t want to remove this
… I want to find a way forward that acknowledges the reality but promotes interop

Hadrien Gardeur: I’m wondering if fallbacks are not used as extensibility for some environments
… I wouldn’t be surprised where specific reading systems craft specific epubs with this in mind
… I do wonder what is the intent
… I don’t remember where fallbacks came from
… if there’s a new version of something like images or video, the fallback could be used to maintain compatibility
… but also could have extensibility
… reading systems might have these use cases, but they don’t allow sideloading

Dave Cramer: I hope we get to a point where we can pass these tests
… ADE supports the fallbacks
… and there’s probably others
… gives us the opportunity not to mangle the spec
… and allows for the extensibility

Garth Conboy: I agree with Matt
… we need to be careful not to rip out foundational things
… CMTs and the fallback chain may not be foundational
… you may not have to lose CMTs with losing the fallback chain
… I think it’s worth questioning

Matt Garrish: Certainly there’s ways of making things work
… it becomes complicated
… potentially invalidates content in the market
… I think we have to balance priorities
… further to Hadrien’s point
… there’s other systems that aren’t on our radar
… libraries may be using linked metadata records
… it’s a means of having the record travel
… there’s lots of requirements on merging and consolidating that may not come into play
… there’s use cases that may not be on our radar

Dave Cramer: I think we’ll be having lots of discussions like this
… we don’t have access to all user agents

Ivan Herman: This reminds me of a question I was asked
… with web and HTML, a big search engine (e.g., google)
… could come back with data that says “25% of websites use x feature”
… there was an established way to have data like that
… but is there an equivalent for EPUB

Dave Cramer: We’ve done a little bit of this
… thanks to the goodwill of group members who have access to this data
… we can only do this in certain ways
… it’s a lot easier to do on the web

Brady Duga: The web and EPUB are just really different
… there’s no “open EPUB” equivalent for the open web

Wendy Reid: It just takes a lot of processing

Ivan Herman: I expected that
… we may have to think about ways to achieve something like this
… we need to make some decisions feature by feature about what should or shouldn’t be normative
… is it implemented or not
… implementability is not the only criteria
… we need ways to determine this

Dave Cramer: Thanks for the good discussion, I’ll be filing issues on the things I find to discuss them
… this will come up again!
… Thanks everyone, see you in a week!