Meeting minutes
Wendy Reid: APA meeting is next week
Extended Descriptions - https://github.com/w3c/publ-a11y-extended-desc/blob/main/README.md
Wendy Reid: so we should discuss the proposed Extended Descs
George Kerscher: Trying to get new values into ARIA is difficult
… we have been working on 2 attributes that identify extended descriptions
… Gautier is the author of the proposal
… APA came back with comments, and we added them to the explainer
… this technique identifies 2 options. For EPUB we suggest a link after the thing that needs an extended desc
… likely to the end matter that is out of spine
… there the image is replicated and there is a back link
… For the web, you can use a details element that expands and collapses. We know that can be a problem in EPUB
… You could also place it directly underneath the image, but that doesn't work as well for mainstream uses
… There is an ARIA details that can point to things on page, in this case it would point to the link
… Also when people encounter this link there is no agreed upon way to identify the link or the content once you follow it
<Charles LaPierre> role="extendeddescriptionref" to mark the forward link (the anchor in the primary content that points to the extended description).
<Charles LaPierre> role="extendeddescription" to mark the container that holds the extended description (which can be in the same document or in an external resource).
George Kerscher: we propose two attributes to solve this called [???] and [???]
Matt Garrish: This is bigger than dpub ARIA, extended desc is bigger than epub. That is why we are talking to APA and not just ARIA
George Kerscher: Gautier Chomel_ and Gregorio Pellegrino won't be able to attend next week, but I and Matt Garrish will be in attendance
Wendy Reid: I know how resistant ARIA is to adding new roles. I wonder about roles when there isn't a specific semantic behavior
welcome colleagues from dpfj : )
Wendy Reid: adding more semantics when there isn't a known behavior can be a problem
… I think I am on the side of ARIA. What is the benefit of new semantics?
<Gregorio Pellegrino> https://
Gregorio Pellegrino: We already have guideline for these semantics.
… We have checked with RS devs and they prefer a more declarative way
… with ARIA details you have to walk the DOM, sometimes in different files
… For the expected behavior we have some suggestions in the explainer
… This could be in a sidebar or popup, etc. So we expect behavior
Matt Garrish: I raised this previously. And this is why we are bringing it to APA first
… they have concerns
… so we are in the process of presenting at this point. The meeting next week will help flesh out the proposal from both sides
… this is just the beginning of the conversation
Hadrien Gardeur: I mentioned in email, but we have already impolemented extended desc in our iOS app
… We have an image viewer that allows you to manipulate the view of the image (zoom, rotate)
… And there we have an extended desc
… We have mostly kept the algorithm, though slightly simplified. You get the extended desc there
… this is done via button, with the label taken from the link
… there can even be multiple
… it works. I don't like the use of non linear resources, but otherwise the affordance works
… We do see some bad descriptions, which are content problems. But when they are good the experience is really nice
… There are some questions about read aloud, and whether to use extended desc
… so we are interested in guidance there. It is a little too verbose
… You can try it now on the Apple ecosystem
George Kerscher: Accessible name and desc. I assume desc is the alt text?
Hadrien Gardeur: It's more complicated than that. Alt text is one thing to use, but the algorithm calls for other things that are more important
… so we kept to the spirit of it
<Gregorio Pellegrino> https://
George Kerscher: I am not familiar with it
Gregorio Pellegrino: Sending the link now
Hadrien Gardeur: This is now in the toolkit so other apps can use it
George Kerscher: Today we don't know what the link is to
… down the road I think a lot if pubs will want to use an icon as the link
… a standard icon could help in identifying it
Wendy Reid: I think the use case is very clear
… the way we are saying to do this in EPUB makes sense, but we have to make the case for the web in general
… things are different in epub
… extended desc being a separate file makes perfect sense for EPUB. it isn't as clear for the web in general
Matt Garrish: That is why we want to go to APA first
… we don't want to start in dpub-aria to avoid surprising people
… we just want everyone to get involved to see, we don't want to do it in isolation
… I agree if it doesn't work for the web, that is fine
Brady Duga: where did this document come from?
Gregorio Pellegrino: the joint TF
Gautier Chomel_: I want to add a use case around 1 file per chapter [I missed the first bit]
… we have been working in this for a while, sent the first version to APA. I suggested we do the same for the web use case (open an issue, etc)
Charles LaPierre: side note to Hadrien Gardeur non-linear comment
… I don't think we are hard and fast on that
… pubs don't want to see extended desc at the end of the book
… it is up the to the publisher
Susan Neuhaus: Gautier Chomel_ said we should develop a use case for the web
… is that outside our role?
Wendy Reid: There are areas of publishing that have browser experiences that are still publishing scenarios
George Kerscher: Publisher use of LMS is extensive
… if they choose to use details under an image it is still useful to know what it is
… as opposed to something that is frequently use for details (e.g. answers)
… so this would clarify what the content of the details is
Wendy Reid: We are meeting with APA next week, plus some things they are bringing to us
w3c/epub-specs#3079
Brady Duga: This has come up a few times recently, that we don't have a processing model
… the question is that we don't say what reading systems should do when they ingest one of the documents
… what if it is missing a resource? what if it isn't jsonld?
… what do you do if there is a target that isn't in the document
… I think we need to make sure that these solutions are interoperable because these things will be common
… we should tell reading systems what they should do when they encounter this
Laurent Le Meur: we have an expectations section
… to create a real processing model we need some thing that is normatice
… one about the shape of the model, one about the processing, one non-normative expectations of reading systems
… we can add more to a non normative section, but we can't call it a processing model
Ivan Herman: I am in favor of doing that normatively
… I wonder if there are similar issues coming up from the other end
… and there is also the writing end
Brady Duga: That is true, too. Right now, we talk about implementers but don't make it clear if we are talking about reading or writing
… we say that implementers should support the other kind do we say that they should output the other as well?
… we should clarify that the output should be this, but you are allowed to read other things
… html makes similar things clear, like improper nesting of tags. We need to clarify when we talk about implementers if we are talking about reading or writing
Ivan Herman: To step back, we say we do not define how the implementation works in the epub. We only define export
… we are only standardizing the export, not the import
Laurent Le Meur: Brady Duga you are talking about collections, we should maybe remove it
… once stable we will create a json schema
… so that will be part of the judge
… so anything that doesn't fit the schema will be rejected
… so it must match the schema
… it doesn't solve everything, like if the anchor is missing
Brady Duga: To respond to Ivan Herman, I don't read that section of the spec the same way, we don't want to define what happens inside the reading system
… but import is still covered here. We should say "you SHOULD reject" or you "MUST reject" an import that doesn't match the schema
Hadrien Gardeur: It sounds tricky to do strong requirements
Brady Duga: we should come up with a way to specify what should happen if an annotation doesn't match a file like "chapter1.html" but is still relevant, like if the annotations are traded between a "Moby Dick" from Standard Ebooks and another from Project Gutenberg
Hadrien Gardeur: you could theoretically find the annotation if it is text fragment
… I also think un-anchored is fine, but not everyone will. So I don't think we need hard requirements
… it is a balancing act
Brady Duga: I'm not proposing any specific text right now, I like leaving things open. Maybe we should require that any annotation should be shown. But a reading system might not want to do that. So we shouldn't be too specific
… but maybe people haven't thought of this. We could use "SHOULD" because if makes sense to try and find a text fragment even if the target content document is lost.
… we can't make it precise, but we should add more than is there
… I can raise individual issues on all the things I have listed, or do we want someone to take a crack at it for now
Ivan Herman: We should this out. I will prefer to see the format fully fleshed out first
… and I would like to see the skeleton of a section for this
… it makes it easier to do PRs
Laurent Le Meur: My only question now is, if we want something normative and we say to reject, can we say that if the external schema is non-normative?
Ivan Herman: We can make the schema normative. It is stable enough now
… we can at least try
… so the schema will be part of the spec, perhaps indirectly
… It won't be easy to produce
Wendy Reid: Next steps?
w3c/epub-specs#3078
Wendy Reid: Something to think about for two weeks out
… writing tests
… it is different than epub
… So please think of ideas
Brady Duga: we only test normative rfc2119 statements? There aren't that many in the spec
… one of the reasons for the processing model is that there are normative statements that are missing
Laurent Le Meur: The files you mention are good for import, but we also need to test export
… so we need a single epub with a set of annotations to make
Wendy Reid: Right, so we need export and import tests, so we will need a canonical epub
… we might be able to automate some of this, like the export
Ivan Herman: The problem is, if I want to annotate I can't choose the right selector
… the RS will choose. There might be two different selectors for the same things
Brady Duga: we can automate some
Ivan Herman: so auto testing that the selectors id the same thing is hard