W3C

Publishing Maintenance Working Group Telco

23 July 2026

Attendees

Present
Avneesh Singh, Dale Rogers, Brady Duga, Gautier Chomel, George Kerscher, Gregorio Pellegrino, Hadrien Gardeur, Ivan Herman, Daniel Kimberg, Masakazu Kitahara, Shinya Takami, Susan Neuhaus, Toshiaki Koike, Wendy Reid
Regrets
-
Chair
Susan Neuhaus, Wendy Reid
Scribe
Brady Duga, Wendy Reid

Meeting minutes

CFIs

George Kerscher: Lars talked to me at a meeting telling me how great CFI are and that he has an open source implementation
… and I am trying to understand the relationship between CFI and EPUB
… are CFI in 3.4? Are they referenced by epub? what is the w3c status

Ivan Herman: The answers are no, no, and no
… they are not even a w3c note, and it is not referenced from 3.4
… in the EPUB WG, Brady Duga and I looked at it, it seems half baked
… e.g. there is no processing model
… we talked about what to do about it, and at the time we let it alone
… we could change our mind now, that is just where we are
… CFI is very bound to xml, so we may be in a better position, since we previously had html serialization as a posibility
… but that has changed, we have decided on xml
… I think a discussion with an eye on annotations
… is a good topic. Let's not discuss now, but in the future we could

Wendy Reid: Chairs have discussed in relation to annotations
… so we likely have more to discuss

George Kerscher: Given that there is an open source impl, if two epub impls use it, does that count as two?

Avneesh Singh: Since it isn't in the spec, the 2 implementation threshold doesn't exist
… some reading system people are very into CFI. At the same time, some people hate it

Ivan Herman: I agree with Avneesh Singh, but the answer to George Kerscher's question is no, we would need another implementation
… but we need a separate full discussion on this topic

EPUB 3.4 is now in CR

<Wendy Reid> https://www.w3.org/TR/epub-34/

Wendy Reid: We are in CR! (3.4 that is)

<Wendy Reid> https://w3c.github.io/epub-tests/contributing

Wendy Reid: time to talk testing

<Wendy Reid> https://w3c.github.io/epub-tests/results

<Wendy Reid> https://w3c.github.io/epub-tests/reports/xx-template.json

<Wendy Reid> https://w3c.github.io/epub-tests/

Wendy Reid: the test suite is complete, I believe
… we added scrolled layouts, there are mixed layouts recently added
… anything else?

Ivan Herman: Not that I know, but if someone would check that it would be helpful
… we have 175 tests, inheriting old ones and adding some new ones

Wendy Reid: We still have the 3.3 test results. We have some new tests for 3.4, and some of the old tests have changed
… it was hard to get all the tests run last time, and now we have even more
… ideally, we would have everyone run all the tests, but if that is too hard what should we ask for
… ? Maybe just new tests, or not all platforms, etc

Ivan Herman: There were a number of tests we had to rename, so the old results don't match up

<Wendy Reid> ak Gregorio Pellegrino

Ivan Herman: We may want to manually update those

Gregorio Pellegrino: 3 or 4 years has past since the last test round. I think the testing was very useful for the community, so I think a full run is very useful (regression checking, etc)

<Ivan Herman> +1 to Gregorio Pellegrino

Gregorio Pellegrino: We did do some of this at TPAC last time, maybe we can do it again

Gautier Chomel_: I agree. I will be testing all the tests on our engines (reader and mobile). Even though these are the same engines, there have been forks
… Maybe we can mark the old passing tests that have not been rerun should be marked as outdated (or something)
… In terms of timing, I plan to test on thorium by dec, is that too late

Ivan Herman: We only have a minimum deadline. We said we will not go to rec until Sept, but we can go later
… our charter ends in Feb, so if we get the results in Dec, we can vote in Jan, and go to rec before Feb
… and that leaves a lot of time for others as well

Shinya Takami: we plan to test new and modified tests before tpac. If we need all the tests, we will need to discuss. I don't know how long that would take

Wendy Reid: We have reading platforms (ios, android, web, etc), and a rendering engine. We should clarify if the engine is the same, you probably don't need to retest on all platforms
… so that might take some pressure off

Brady Duga: I have a question on process, we have EPUB 3.3, it has tests, we have 2 implementations for all of those features, in EPUB 3.4, what do we need to test to get out the door for W3C's requirements?
… do we only need to test the new features?

Ivan Herman: Formally, we have to do the whole thing
… we could hand wave, but technically this is not a delta spec, so we should do the whole thing

Dale Rogers: You mention that a lot of systems use the same engine. My understanding is that on the Mac, everyone uses the same engine

Ivan Herman: The Mac comment isn't correct. On iOS it is, but not on the Mac
… We never really broke out a matrix like that
… This is not an authority that verifies things - we just believe what the reading system tells us
… We do make it clear that if two implementations use the same engine they should not both report success

Dale Rogers: Do we not test the reading systems then?

Ivan Herman: No, we can do the test ourselves if we want
… last time we thought it was cleaner and more fair if the authors of the RS do it

Wendy Reid: We have about 3 layers we need to test
… many reading systems use a browser engine underneath, we don't (and really can't) test that
… But they still have a rendering engine on top of that which does epub
… and then there may be an app layer
… So Kobo had a number of platforms, but the engine was shared

Susan Neuhaus: MobileReads has some forums for developers
… do we want to reach out to them

Brady Duga: I do not have an answer for Susan, but I wanted to talk about letting implementers do the testing
… outsiders may not know how the system works
… sideloading may be possible but it may run differently
… there's two different pathways, or more, so it's important to have the reading systems provide the information
… you also don't know about the engines or quirks, no idea what code may be shared, some may not be, and won't know
… it's difficult for anyone except the implementor

Ivan Herman: Two more things - still on the matter of the same engine on different platforms
… So we added the ability for the same company to report multiple platforms results (android, ios, web, etc)
… and the report shows the individual platforms (e.g. 3) and also a single report that shows the `or` of the reports
… and also we generate the OPDS for the entire test suite
… which could help any reading system that supports that

Hadrien Gardeur: Yes, very useful here, and widely implemented

Ivan Herman: and my answer to Susan Neuhaus is, where anywhere we have possible help, adding to tests is the best thing they could do to help

Susan Neuhaus: Thanks, that makes sense. On the other side of things, most of our books go to Kindle
… is it possible for someone else to do that?

Ivan Herman: We have failed over 10 years to get them on the group
… I would take just about any results for Kindle

<Avneesh Singh> Is Kindle considered as EPUB 3 implementation formally?

Dale Rogers: There is a question about what is behind the scenes vs what is available to users
… I would think we want to test what people can actually see

Brady Duga: It's not hidden behind a wall, there are certain publishers that can do certain things, like fixed layout, there are rules about FXL, only certain pathways are enabled for certain pipelines
… different entities might experience different pipeline, readers don't really see the differences in the end experience

Ivan Herman: We also have to understand, we are testing the spec, not implementation
… s So we care whether the spec can be implemented. So we don't care whether it is visible to end users, from a w3c CR process, it just matters that it can be dome

Hadrien Gardeur: Back to Kindle, we know they don't work with epub directly. I think that isn't really a problem, and a lot of apps also do the same thing, even though they are cosidered epub reading systems
… Also, I am collecting requirements for comics stuff. It turns out that if Amazon requires something, a lot of books follow that for all books
… they have an outsize influence
… We could set up testing for Kindle easily, and since we won't get the tests from Amazon, I think it would be interesting

Gregorio Pellegrino: I agree with Hadrien Gardeur regarding epub as b2b
… There are often lots of changes made to epub for each platform. For instance, kindle uses mathjax for mathml
… so sideloading may give different results

Ivan Herman: The reason I would be unhappy to do it ourselves is that it is ideal to have the implementor do it. But since Amazon won't we will have to settle for second best
… I am unhappy we have to do it, but that is where we are

Susan Neuhaus: Amazon doesn't like side loading for testing, they prefer an emulator they provide

Wendy Reid: One other reason we stongly prefer platforms do the testing - in the case that side loading doesn't work the same way as other ingestion methods, we might make it look like there is a bug or missing feature in the platform
… so reading system testing is best, but we take what we can get
… But even just ingesting all the books would be good
… and if they are visible we can test them

Hadrien Gardeur: Back to what Ivan Herman said, there is more than just the spec testing, Amazon has other requirements that are being implemented by authors

Ivan Herman: At the moment the results are the old results. I propose that we remove those results, so the tables would all be empty
… old results will still be archived. Objections?

Susan Neuhaus: Makes sense, just so long as there is a link somewhere to see the old results

Ivan Herman: Ok, it is there, but there is no link. But I could add a link for old tests

Susan Neuhaus: That works

Gautier Chomel_: Do we have a field to say who ran the test? For instance, was the tester from the same org as the reading system

Ivan Herman: So we can add it, but what do we do with that data? I don't think we should say who ran the test
… I am not sure if we should display a name

Wendy Reid: we don't need to show it, but we could store it

Brady Duga: It seems like we could just simply say "is representative of company" boolean? We just need to know platform a's tests were done by someone else, platform b was done by an employee

Gautier Chomel_: Agree with Brady Duga.

Ivan Herman: Should I add a flag that is official test or not?

Wendy Reid: Something like that

Ivan Herman: I will do a PR, we can discuss there

AOB

https://www.w3.org/news-events/tpac/2026/registration/

Wendy Reid: reminder - tpac registration
… early bird ends soon. Invited expert don't pay, and there are fee waivers if needed
… Also, some people have vacations coming up - should we reduce meeting cadence?

Susan Neuhaus: Can we do this in slack?

Wendy Reid: maybe I will do a poll

Minutes manually created (not a transcript), formatted by scribe.perl version 248 (Mon Oct 27 20:04:16 2025 UTC).