Meeting minutes
<Ivan Herman> pesent+
EPUB 3.4 CR
Wendy Reid: Should we go to CR with 3.4. But what does that mean?
Ivan Herman: Without the details, the group declares we are done with the technical work and we consider the technical work done
… this is just a declaration of that
… but we can't make further technical changes unless they are clear errors
… we need to be at CR for some amount of time before we can go to rec
… the main goal would then be testing
… we already have about 100 or 150 tests in the form of books
… so a RS should run all those books to verify they pass
<Wendy Reid> https://
Ivan Herman: We have to show that at least every test has 2 passes
<Wendy Reid> https://
Ivan Herman: And they should be independent
… So we need at least 2, but 4 or 5 would be great
… Once all tests have passed and no technical errors have been found, then we can go to rec
… There are more details on the specifics, but Matt and I can handle those
… we will need all horizontal reviews done
… They are all done except security, in other conversations we have discussed and feel we can continue without it
… we can just call time on it, since there aren't many changes
… We have discussed with ISO, they will informally review the document to see if there are any hurdles for publishing at ISO
Hadrien Gardeur: Some questions on open issues
… There was an issue on multigranularity that applies to structural semantics - is that doc covered?
Ivan Herman: No
Hadrien Gardeur: We also have informative text that could impact the core spec
… I also did not open an issue on light novels
… But I am doubtful that mixed layout has 2 implementations in the wild
… I haven't double checked the tests for 3.4, but I don't think it is there
… If we were to add testing for it, I doubt we would pass
Ivan Herman: One releveant item - if a feature doesn't pass tests, then that feature should be removed
… In our case we are special, so we use one of our specialized labels but leave it in
Dale Rogers: Is there a time limit on CR? Since we are depending on other people
… Is the testing done by third parties or is it this group?
Ivan Herman: For the first, there is no limit. Some groups stay in CR for yearsa
… As for the testing, good question
… Previously, we said anyone can do it
… But it looks better if the company self tests
… Because the tests can be used by the reading systems to identify their own problems
Brady Duga: on the mixed format publications, we've sort of known its been an issue, certainly not new now
… there's reading systems that pass each, but nothing requiring a combination, ideally we should, but if we do it, and fail, we'll need to rewrite a lot of the spec, but it's not one feature
… we'd need to introduce the concepts of a just fixed or just flowing document, it may be worthwhile, but we should be prepared for work
… but if we want to do it, we need to undertake it as a project
… decide on it now
Wendy Reid: I am interceding as chair
… this is in the charter for 3.5, and I agree this is a big project
… I think we need to be more research, etc. I don't think it is worth hinging CR on it as we need to spend more time on it
… and it isn't in the charter for 3.4, but it is for 3.5
Susan Neuhaus: Are there consequences for the security review?
Wendy Reid: We did security for 3.3, very thorough. We expect minimal things if anything
Brady Duga: I did look through the security reviews repo, there are multiple open reviews with no activity, lots of groups in the same position, this is a W3C problem
… we're not to blame for this
Hadrien Gardeur: On mixed layout, the work we have done makes it easier. It is a single feature
… there are not that many references to it, so I am not convinced that it will be that hard to remove
… I disagree with pushing this off. Is it responsible of us to keep pushing this off?
Ivan Herman: I don't have all the facts, so I would like to see what it means to remove it. I would like to see a PR of what needs to be done to remove it
Wendy Reid: To clarify, from out of the blue, I mean that this was never discussed as something we want to handle in 3.4
… EPUB 3.4 came out of problems found in 3.3, we are seeing that again - 3.5 is intended to address the things that came out of 3.4
… I think this deserves the time and consideration, and shouldn't be hurried in now
… And we can make it part of new features, like full bleed images
… we aren't pushing it out that far
… since we are discussing recharter now
Hadrien Gardeur: I am not saying we should work on a solution now, but we should warn people about it and test it properly
Wendy Reid: We will have tests for this feature
Hadrien Gardeur: So if we fail, then we need to remove it
Shinya Takami: I agree with moving it to 3.5. The situation hasn't changed, and we should take the time to address it
Brady Duga: So we will have tests for mixed documents, if we don't get 2 implementations, will we deprecate the issue?
Ivan Herman: We have to check something, but the current suite is an extension of 3.3
… So we are adding tests on top of 3.3
… We need to check to see if this was tested at 3.3
… so I don't really have all the info
Gregorio Pellegrino: For the tests, how do you test if it passes or not?
… if you open a mixed doc the cotent displays
Wendy Reid: we have 2 up to date tests, and we have older tests
Ivan Herman: we have some test results from the previous run
… We have JSON file with the test results
… What do we do with the results this time?
… do we have everyone re-run all the tests, or just the new ones?
Brady Duga: I disagree with Gregorio, there are specific rendering guidelines for FXL, something displaying doesn't necessarily pass the test
<Susan Neuhaus> +1 brady
Avneesh Singh: We have one implementation
… We had a similar issue, we didn't remove it but we just made it non-normative
Dale Rogers: It sounds like the process is more than just passing epubcheck
… there is a rubric, so who makes that?
https://
Ivan Herman: If you look at tests, the first page explains what you should see
… so the rubric is in the book itself for each test
… Avneesh Singh, you are both right and wrong
… epubcheck is separate, and it is not part of this testing
… there is work to do to make epubcheck work with 3.4 , but that is ongoing
… The charter does require a new epubcheck with 3.4, but it separate from the traditional CR testing
Wendy Reid: We had a process for backwards compatibility, we can mark a feature as at risk
… This isn't just a single feature, we don't really call it out. We could add a note at the spine override
… But I would like to worry about this later
… we may actually have the 2 passes
… We actually have 4 passes on similar tests
… If we decide we have to solve it we may need an immediate re-charter
Hadrien Gardeur: I can point where it is referenced - 6.2 and 6.3.1
… The are short, then there are some examples
… so it is pretty narrow
… when it comes to specialized reading systems, it requires very precise conditions to work, they are not the general case we expect
… so it is fine to wait until the test fails, we just need to make sure the test is accurate
… so the documents have to have real HTML in them
… so we just need to be extra careful about what we mean with specialized reading systems
Wendy Reid: That is a completely valid implementation of that feature. We don't tell people what to put in their content
… so while it is narrow, it is valid
Hadrien Gardeur: What I described is a partial implementation
… but moving to CR is fine, assuming we have tests
Ivan Herman: I would ask that someone add that test who is knowledgable on the topic
… we shouldn't wait for this to go to CR, we can deprecate later if we need to
Shinya Takami: If some features are only in Japanese reading systems, and we deprecate those features, we would have to object
… So if 2 or more implementations exist it is valid
Ivan Herman: The source of the implementation (e.g. country) is irrelevant for CR
… so if there are 2 Japanese implementations, that is fine. It is considered passed
Wendy Reid: For that reason I don't even want to add an at risk to this
… so we may not have enough implementations to be comfortable, but we would still pass CR so shouldn't use at-risk
Ivan Herman: The goal of CR is to show a feature can be implemented. That is why we have the low bar of 2. It just shows it is possible to do, not that everyone does it
Hadrien Gardeur: To clarify, if we have 2 implementations then it isn't an issue. I just want to make sure the scope of the tests is correct
… We need to test the spec, and not design the tests for the reading systems
Wendy Reid: I will take an action item to review the tests in that section
… we will have a more robust set of tests
PROPOSED: Progress EPUB 3.4, EPUB Reading Systems 3.4, and EPUB Accessibility 1.2 to Candidate Recommendation.
<Shinya Takami> +1
<Brady Duga> +1
<Ivan Herman> +1
<Toshiaki Koike> +1
<Grigorily Manucharian> +1
<Avneesh Singh> +1
<Susan Neuhaus> +1
<Wendy Reid> +1
<Masakazu Kitahara> +1
<Gautier Chomel_> +1
<Dale Rogers> +1
<Murata Makoto> +1
<Hadrien Gardeur> +1
RESOLUTION: Progress EPUB 3.4, EPUB Reading Systems 3.4, and EPUB Accessibility 1.2 to Candidate Recommendation.
<Gregorio Pellegrino> +1
Ivan Herman: Procedurally, this becomes valid in 1 week, so people not on the call can object if they want
… and we may not get the actual CR until end of July
Murata Makoto: Can SC34 distribute the text before our meeting in August [SC34 meeting]?
Ivan Herman; The text is public
Murata Makoto: So it will be available at the beginning of August?
Ivan Herman: Yes, but the format will change so that isn't final
Murata Makoto: But I shouldn't say it is CR now, I should wait a week
Wendy Reid: Yes