Publishing Working Group Telco — Minutes
See also the Agenda and the IRC Log
Present: Tzviya Siegman, Dave Cramer, Mateus Teixeira, Ben Schroeter, Matt Garrish, Deborah Kaplan, Wendy Reid, Ric Wright, Marisa DeMeglio, Laurent Le Meur, Avneesh Singh, Gregorio Pellegrino, Bill Kasdorf, George Kerscher, Franco Alvarado, Joshua Pyle, Charles LaPierre, Garth Conboy, Rachel Comerford
Regrets: Luc Audrain, Benjamin Young, Romain Deltour, Juan Corona
Guests: Lars Wallin
Chair: Wendy Reid
Scribe(s): Mateus Teixeira, Dave Cramer
Ivan Herman: ivan+
Wendy Reid: let’s get started!
Wendy Reid: https://www.w3.org/publishing/groups/publ-wg/Meetings/Minutes/2019/2019-11-04-pwg
Wendy Reid: approving minutes…
Resolution #1: last week’s minutes accepted
Wendy Reid: minutes approved!
1. Road to CR
Wendy Reid: https://github.com/w3c/publ-wg/wiki/%5BPublishing-WG%5D-CR-Transition-for-pub-manifest-and-audiobooks
Wendy Reid: plan for today is to talk about getting to CR
… posting a link… first major thing is the transition document that we are currently working on
… major part we need to discuss and finalize is our test plan
… need not necessarily hard tests, but clear strategies for how audiobooks and pub manifest documents will be tested
… we should finalize the text we have here, hoping we get people to help with the task of writing tests and finalizing before the end of the year
… ivan, anything missing?
Ivan Herman: one more thing… not just what to test and how, but how we ourselves will produce reports for the testing…
… can produce purely manually but it may be tedious… there are ways to speed it up and automatize it
… but depends very much on what we test and how
… not something to decide today, but something to think about
Wendy Reid: I don’t see Tim Cole here today, but he was very kind to show what Web Annotations did for testing
… we can use the structure from Web Platform tests… it also provides a visual way to prove how well testing is going
… might be our best option, might just need to make that clear in the transition document
… we could also produce a test suite people can reference
… I think that’s the best way
Bill Kasdorf: Yay, Tzviya!!!
Tzviya Siegman: I don’t see Lars… who has implemented audiobooks in his reading system (Colibrio)… and he said he can help… so that can be a part of the plan… with Tim and Lars working on automated testing, we have a solid team
Ivan Herman: one other thing that may be helpful… I did an implementation of the pub manifest algorithm and collected a number of tests which are all small manifests that could be a good start with that specific part of the testing
… have organized it in completely ad hoc manner for myself, but that can be changed
Wendy Reid: just noticed that LarsColibrio joined… do you want to say hi?
… seems like you’re muted
Lars Wallin: hello everyone [some garbled audio in background]. Better if I mute, I am in a very noisy environment.
Wendy Reid: Lars is here… have already spoken to him about testing… also gotten word of another implementer who is interested and who will want to test, too
Dave Cramer: dauwhe: how do we know that these sorts of tests pass?
George Kerscher: what is the relationship between these tests and an architecture that would make it easy to move this forward?
Lars Wallin: Good idea George, I would say that you will get this for free basically
Wendy Reid: need to cover the statements made in the specification, pass/fail basis… failures can point us to things we may not have realized would be problematic… some other kinds of tests could be written in such a way that they meet the requirements of the spec
Dave Cramer: looking at ivan’s tests, they are mostly manifests that exhibit properties of pub manifests, but when I think of html/css tests, it passes if there’s a ‘green square’…. it’s obvious whats passes and fails…
… but I can’t tell here… I guess i can point a UA at these manifests and verify that it works, but that is not clear from this json-ld file
Lars Wallin: I would like to suggest that we use our implementation as base for the testing algorithms
Lars Wallin: Our code should be a 1:1 implementation with the spec
Lars Wallin: What you are discussing now guys are not spec tests ;)
Lars Wallin: Spec tests need to be unit oriented
Lars Wallin: (I wish all these people could just leave the room I’m in (an a11y meetup in Stockholm) 😋)
Lars Wallin: https://web-platform-tests.org/
Lars Wallin: Checkout the link :)
Ivan Herman: to be clear, what I have there grew out of testing my own implementation… and I do not claim to do more than that… I have a separate yaml file, too… there is a description of the tests there, but I don’t claim it’s more than that… we have to work on it
… to be more generic and answer George’s question… the infrastructure and tests are closely related… and that being said, we have to realize that an EPUB checker or audiobook checker will test whether a specific manifest is correct or not… the goal of the CR phase is to test whether the spec is ok or not…
… whether there are no problems with how the spec is formulated, and whether the statements can be implemented in the first place
… it’s securing that the spec we produce is okay… in this sense, the implementation I and mattg did, are useful because they revealed inconsistencies in the algorithm
Dave Cramer: ivan, it’s important to make the distinction between the tests we need for CR, about implementation, vs. validating structure like EPUBCheck… that’s a cautionary tale of excellent validation that doesn’t necessarily correspond to support in user agents
… want to avoid tests depending too much on internal states of user agents… hoping the tests for audiobook are very concrete–“given this manifest, the user experiences the following”
Wendy Reid: that’s my hope too… one thing for the manifest to be processed, and another thing for the user to experience the audiobook on a consistent basis… don’t need to provide a lot of user input for those needs to be met…
… we tried to make that clear
… it sounds like we will need to clarify use the webplatform test model and how
Tzviya Siegman: reading Lars’ comments… let’s make sure we keep the distinction between spec tests and other tests clear
Wendy Reid: does anyone have experience writing web platform tests that i don’t know and have overlooked?
George Kerscher: do you know about ACT (tests for accessibility)… rec for writing tests for accessibility
Wendy Reid: i’ll look over those…
Wendy Reid: https://www.w3.org/TR/act-rules-format/
Wendy Reid: alright, i think we’ve exhausted this topic… will make some changes to the CR document this week; send me ideas and comments and i can incorporate them as needed
Lars Wallin: Will do Wendy. I have a bunch of ideas 👍
Wendy Reid: next thing is the matter of shipping LPF concurrently with pub manifest and audiobooks
2. LPF publish with PM and Audiobooks
Wendy Reid: laurent, can you comment on how close LPF is to publishing?
Lars Wallin: I thought you called it LPP which would be an awesome name 😁
Laurent Le Meur: yes, we left the spec almost done, but on the two minor verifications to do, one related to respec warning that i want to work on with ivan… and the other thing is i need to add better links from LPF to web publication and publication manifest specs
… but for the text itself we left it with no new comments… seems good… and would like it to be published as a companion for the audiobook spec, because when we speak to audiobook producers, the first thing they want is a packaging mechanism
Wendy Reid: awesome
Ivan Herman: first, mattg is probably a better expert on respec than me
… the other thing is that we should make it clear that once we have acceptance of director, LPF will need to go through a different path because it has not been published yet… needs to be published by the webmaster with a shortname etc.
Lars Wallin: We have a LPF implementation as well if that helps
Tzviya Siegman: quick reminder of requirements for LPF… do we require it as the packaging format?
Wendy Reid: no
Tzviya Siegman: remind me what the difference between LPF and ZIP is?
Lars Wallin: The name is sub optimal
Lars Wallin: Lightweight Publication Packaging would be much more true to its purpose
Ivan Herman: we need a formal vote about the fact we want to publish, and under what name and shortname
Wendy Reid: we did our naming battles months ago, and I do not want to revisit those dark times
Lars Wallin: Sorry Wendy to be a drag, but LPF is seriously ambiguous. It’s kinda in the “global namespace”
Lars Wallin: And it is not a Lightweight Packaging Format. That would imply that you can package anything. Nope. Its only publication.json. if we change the publication.json to manifest.json
Proposed resolution: Publish the Lightweight Packaging Format note with the name Lightweight Packaging Format, and the shortname LPF (Wendy Reid)
Garth Conboy: +1 (with note that Audiobooks isn’t too useful without LPF) [and lets not reopen this discussion]
Wendy Reid: we did talk about this extensively months ago
… the issue have been closed
Lars Wallin: 😁
Lars Wallin: LPP
Deborah Kaplan: +1, and we shouldn’t reopen.
Ivan Herman: +1
Laurent Le Meur: +1
Gregorio Pellegrino: 0
Matt Garrish: 0
Tzviya Siegman: +1 (agree not to reopen)
Marisa DeMeglio: 0
Wendy Reid: +1
Lars Wallin: -1
Deborah Kaplan: Reminder that the perfect is the enemy of the good
Dave Cramer: 0
Ric Wright: 0
Rachel Comerford: 0
Bill Kasdorf: 0
Ben Schroeter: 0
Joshua Pyle: 0
Lars Wallin: But i’m not in the WG so 😋
Deborah Kaplan: that’s a lot of zeros; we’re short a lot of people
… maybe we should reopen by email
… but I want to remind old-timers that we have a pattern of bikeshedding things for months, and then revisiting
… I would encourage people to not block things that are not blockers
… if there were good easy answers, we would have found them
… just remember why we are negative about bikeshedding
Ivan Herman: I agree with Dave, perhaps doing this by email
… there’s a rule that any resolution can be revisited within 5 days
… so we could resolve today and ask for objections/comments via email, and know before the next meeting
Dave Cramer: works for me
Lars Wallin: 👍
Ric Wright: Deborah and Ivan stated my case
… I’m uncomfortable both with the reduced attendance due to holiday, and lots of 0s
Matt Garrish: I don’t want to bikeshed again
Lars Wallin: +1 matt
Matt Garrish: if we said why we’re calling it lightweight, that would be better, but we don’t explain
Laurent Le Meur: there are many zeros, and it would be good to have comments from those explaining why
… what are people not sure about
Bill Kasdorf: for me it’s the lack of the word “publication.” Publication Packaging Format would be clearer.
Wendy Reid: I’ll send an email explaining the issues to vote on, and I would ask 0 and -1s provide short explanations of why they are voting that way
… is it a problem with the name, etc.?
Lars Wallin: Lightweight Publication Packaging is good guys
Garth Conboy: I was gonna comment, I was a plus 1
… I thought that vote was just on the name
Joshua Pyle: I don’t necessarily like taking a vote on armistice day/veteran’s day
… I don’t like having to explain a zero… for me it’s “I don’t care”
… I’m abstaining because I’m deferring to the better judgment of others
Wendy Reid: that’s fine, that’s an OK thing to say
… I just want to make sure we have all the info out there. Consensus is good. I want to know.
everyone: we’re ok with being ok about being ok
3. Move to CR
Ivan Herman: what I would propose to do… we don’t have anything more to change on the doc
… let’s make it clear, in the email that you will send out, in a week from now we plan to make a vote on CR
… people should look at the docs, as they are now final
… we can publish as /TR/ WD
… people should look at the docs and the wiki, and prepare properly for a vote next week
Wendy Reid: I’ll send an email outlining everything we have to vote on
… and respond to the email with any concerns you have
… or you can even just contact the chairs
… or log issues in github
… let’s get to CR!
Lars Wallin: Yay 🚀
Wendy Reid: is there any other business?
Lars Wallin: Good work!
- Resolution #1: last week’s minutes accepted