W3C

Publishing Maintenance Working Group Telco

16 April 2026

Attendees

Present
AvneeshSingh, charles, CharlesL, dale, DaleRogers, duga, gpellegrino, hadrien, ivan, kimberg, mgarrish, rdeltour, shiestyle, toshiakikoike, wendy
Regrets
George, suneu
Chair
wendy
Scribe
mgarrish, wendyreid

Meeting minutes

wendyreid: looks like there is the possibility of making some decisions on smil issues

Playback model for media overlays - w3c/epub-specs#2943

Hadrien: a bit of context - I started reading media overlays when we voted to add rolls - and we're implementing this in readium swift implementation
… there are a number of issues in the spec that is not completely clear and can be problematic
… we have a discord channel where we are discussing some of the issues
… looking at this first issue there is very little text about what a reading system should be doing
… one question is what happens when you follow a smil document when it moves around in the reading order between spine items
… the spec only says to play the next document after finishing
… a bigger question is that the specs says to load the next content document - in some cases this can get you stuck in a loop
… I've written about the problems in the issue including an alternate way to handle it
… in the discord channel we got into where do you start playback and how do you navigate within the publication once it's started
… we are only focusing on the second issue here - the current approach is spine-centric
… media overlays are quite different from the spec - you can move around out of reading order as they are not linear
… what I'm proposing is another way - to make a media overlays order different from the spec/spine
… it could avoid the cases of being stuck in loops
… people who have implemented media overlays in reflowable have done it differently from each other and from the spec
… the other issue is about how to handle audio playback

duga: I suspect you are right and we didn't look at the spec when we implemented for play books and we figured it out for ourselves
… but the question is whether we can find a playback model that suits everyone - children's books may be optimized and have audio be primary
… it is a hard question to solve and all the spec has is a "should" - maybe we should remove that line and just leave it to implementers

Hadrien: I don't think it is problematic for the use cases you define
… audio might be more problematic - you can have smil files without audio
… so what is the reading order and does the smil or spine define the order

wendyreid: I think the problems are real but in working on accessibility and spreads one of the things we covered is content across spreads - there are smil cases for this - but I worry about not going too far about moving around with user control
… media overlays often advance without user control - what happens to the user experience when you get moved all over the book
… there are ways to manage that - like a warning - do we need guardrails like having user interaction before this happens

Hadrien: there are no guardrails today - the current spec language doesn't handle the cases we've talked about so far
… it's common to go across spreads more than jumping around the book - the spec is not clear on how to handle spreads
… if we drop what we have and don't replace we won't have interop

wendyreid: do we have a proposal for how we would present the smil reading order?

Hadrien: I propose walking the spine and creating a set of the smil documents referenced - it's detailed in the issue
… it's not a complex way of compiling the order

<Hadrien> w3c/epub-specs#2943 (comment)

Hadrien: the link is to use cases and examples

AvneeshSingh: I think this is more of an implementation guideline - one way would be to make this a note - but do we really need to change the spec?
… what change is required?

<wendyreid> https://w3c.github.io/epub-specs/epub34/rs/#sec-behaviors-loading

Hadrien: we need to change the reading systems spec
… this is more than an implementation detail - it's core logic

DaleRogers: my understanding in epub is that the markup document is the primary source of information and smil documents are subservient - they are called by the markup file
… so can a smil file be standalone and call what it wants?

Hadrien: yes, the smil has par elements that indicate the content and audio to load
… this is a technical issue but also why we have a technical spec for reading systems

ivan: I'm suprised the original smil spec doesn't say anything about this - it defines par, etc.

Hadrien: the problem isn't within one document but that we use multiple documents and tie them to the spine - it's how epub uses smil

wendyreid: smil was probably also architected more with the web in mind

ivan: no, smil allows several html document - the html documents are part of the structure but here they are adjacent

Hadrien: what we're saying is that smil is designed to drive the whole thing but the spec doesn't cover this

ivan: isn't it possible to go the whole way and let smil be the boss?

Hadrien: yes, I think it works better this way

ivan: the problem is the spec is from 2008 and we don't know how much testing it has gone through

Hadrien: no matter what we use we'll have the same issues

AvneeshSingh: one thing I'm worried about is changing the architecture of the spec late and not being able to anticipate what new issues we'll encounter
… the spine gives the reading order and smil goes back to the spine when it completes

Hadrien: I think the spec is broken and everyone has a different implementation - going back to spine won't solve it

wendyreid: hard to get my head around what this will look like
… both in spec language and reading order
… we do not tell reading systems how to do their interface - what would readium look like for example if we changed the language
… there are reading systems that care about this more than others

ivan: putting on my staff contact hat - do you have any idea of the amount and depth of work to make this happen?

Hadrien: right now it's a single paragraph but would need expanding - limited to the RS specification

ivan: more generally about all the issues you have raised

Hadrien: these two issues are mostly about reading systems

ivan: I'm looking at the charter and these are features that will have to be reviewed - will it require adding a year to our charter to handle all this
… our charter ends in less than a year and I want to avoid problems with this going on too long

Hadrien: I don't think it's big changes but I understand the time issue - it is a departure from the current spec

wendyreid: if we want to do this well we should do it right
… there is no denying this is a problem and that overlays are a powerful tool
… I don't want to push in small fixes in the hopes it scaffolds to a better experience
… I wonder if we should recharter with a specific focus on smil - push out 3.4 and then do this as a 3.4.1

mgarrish: You covered most of what I wanted to say. Everyone has implemented, but everyone has different ideas
… there will be implementations that won't conform if we change now. I worry about doing this in the current charter, we need to make sure everyone is on board
… don't want to reach a point where we make the change and someone comes and finds they don't conform
… I'm happy to make the changes around variable bit rates and smaller fixes, but this bigger change is worrying to me
… I do think this needs to be addressed, but worry about where it leads

Hadrien: one of the struggles we'll have if we do another dot release is that many people who have knowledge will not be participating in this group
… we need to think about how to get more people involved - there is a lot of interest between EAA and audiobooks taking off
… the line between audio and text is blurring - people want to implement this asap and waiting on chartering and new specs leaves people unsure what to do
… it wasn't on my radar, either, but I keep hearing about it from people

DaleRogers: I just wonder if this is a bug fix or a new feature

ivan: I don't think anyone wants to push this under the carpet but we need to optimize our manpower to get this right
… we don't have to suspend our work on this even if we move this to the next release - maybe getting 3.4 out should be a higher priority so we can start again quickly
… we could get a quick recharter if we get 3.4 out

wendyreid: I want to echo the same - by rechartering with this as a second track of work it gives us time to do the outreach
… we can talk to some of the larger platforms we don't have feedback from yet
… there might be other issues that we'll find out if we have more time to engage people
… we also have the community group to get ahold of people - discord can also be useful
… we can also improve the future spec language if people are incubating solutions to these problems
… once 3.4 is out of the way we can focus more of our attention

AvneeshSingh: the scope is larger - there are a lot of people in the daisy community who would like improvements but we haven't brought in because the scope of this revision seemed more limited
… if we have a more flexible charter to tackle media overlays that will be positive for daisy members

Hadrien: if we recharter sooner than later I will be more comfortable - we can't push this back a few years considering the state of the spec
… we have three groups of issues - the granularity issue is being done, the audio format and vbr can be tackled, the content model and reading system side, and final are the readin system issues
… if we can get the minor changes in 3.4 and the other issues in a future revision that would work but we need to keep people engaged
… we wrote specs without being sure about how implementations will work and we need to fix those problems now

ivan: I agree that the small issues are the ones we should settle in 3.4 - in parallel we should try to get a call with W3C folks to deal with the administration of whether to recharter or update our current charter
… if we do this, we have to have a hard look at what to do with annotations because we are behind on it - 3.4 is practically ready for CR absent it

Hadrien: we know that some parts are problematic - should we add a warning?

ivan: that should be okay - identifying underimplemented or unclear features is fine

wendyreid: yes, we've done this before with links to issues

Hadrien: I've also seen a recent issue with full-bleed images that is going to be problematic to solve

AOB

w3c/webai-roadmap#30

wendyreid: we have been asked by the web and ai interest group for feedback on how ai technologies and llm have impacted on our group/epub
… we'll discuss this on the next call

ivan: we have got some feedback about ISO and seems that the obstacles are gone and we can go through the PAS process to publish our documents

Minutes manually created (not a transcript), formatted by scribe.perl version 244 (Thu Feb 27 01:23:09 2025 UTC).