See also: IRC log
<scribe> scribe: nigel
Nigel: Today we have TTWG Charter, TTML test repos, not sure what else.
Philippe: A note on IMSC
Nigel: OK, IMSC added to the
agenda
... AOB or anything else to cover?
Philippe: An update on WebVTT
Nigel: OK
Nigel: I note that Philippe has
opened a pull request on the Charter to address the
issues
... opened by me and Pierre, and that's gone round one review
& edit iteration, looking good right now I think.
Philippe: If there are any other
requests for changes let me know.
... The end of the review is end of next week. I don't expect
anything surprising.
... So far 11 responders, 3 abstains.
Nigel: Glenn asked for
ttml1-tests and ttml2-tests, seems like a good idea to me, but
I can't make the repos.
... Any objection to creating them?
Pierre: No objection
group: [no other objection]
Philippe: We need separate repos?
Glenn: Yes
Philippe: OK
Nigel: Please could you create those Philippe?
Philippe: Yes I can do so - will do while you continue...
Pierre: We seem to have the right
people on this call to discuss this.
... Nigel, you summarised this pretty well in one of your
comments.
github: https://github.com/w3c/ttml1/issues/352
Glenn: My understanding is that we want a generic title that does not specify edition, and is flexible?
Pierre: Yes, I think NIgel
summarised this - if anyone wants a generic undated
version,
... the pointer is /TR/ttml1 which always points to the latest
version, but associated with this
... ideally there should be an unversioned title that can be
used to match the "latest" URL.
... The way this came up is it is weird to have a reference
that says "TTML1 2nd Ed" but when
... you click on it you get 3rd Ed.
Glenn: Yes, I don't have any
problem doing that as long as we don't change the title in
one
... of the specific editions. For example the specification
could recommend a generic title
... for use by referencing specs. It could do that in the spec
somewhere, in the Intro or SOTD,
... for example, it could have a note that says "for the
purpose of generically referencing
... TTML1 without specifying date or edition use XYZ". That's
one way to handle it and would
... meet my concern which is it does not change the title of
the specific edition.
Pierre: My thinking was less
intrusive, to use metadata so that tools like specref can
extract
... the name and know that "3rd Ed" is a version.
Philippe: You have 2 ways - you
can put all the info into the h1 or divide into h1 and
h2.
... Respec can deal with it either way but specref only takes
into account the h1 - it doesn't support subtitles.
Glenn: You mean the text content of the h1 or some metadata associated with it?
Philippe: The text content.
Glenn: That's the problem, I don't want to change the content of h1.
Philippe: Fine, it doesn't change
the fact that we are still talking about TTML1. I'm not
... trying to recommend one or the other. It's a matter of
taste. Some people do not like to
... use versions at all. They give the choice to make the
version indication secondary, and
... we tell them that they can omit it or put it in the h2. It
doesn't matter to us.
... You can still use the version URL.
... The latest version is calculated automatically.
... We only have one copy on the server, we don't modify it
depending on how it is accessed.
... That's with or without a date in the URL.
... It's difficult to differentiate the case where the reader
wanted a dated version or a general version, on the server
side.
Glenn: And the latest version changes over time?
Philippe: Correct, those are calculated.
Glenn: From the server's
perspective it just takes whatever comes out of the
filesystem.
... It doesn't sound like the metadata system would work unless
you modify respec.
Pierre: Or specref.
Glenn: Yes, specref is what I mean.
Pierre: The first step is to modify those.
Glenn: How would you change the metadata in the document?
Pierre: That's a w3c tooling choice, it should be the same across all specs.
Philippe: I still don't understand - what is the purpose of the metadata, to give the edition?
Glenn: What I'm hearing is a name
meta element in the head with a generic title, and
specref
... if it finds that uses it instead of the h1.
Philippe: Duplicating the information in metadata is never a good thing.
Glenn: In this case the content would be different because it would not include the version information.
Philippe: If we push the version
into the h2 and teach specref about the h2 would be the
... better option. It knows if you are linking to the generic
latest or a specific edition. It can
... return markup differently depending on how you use it. So
it can merge the h1 and the h2
... if applicable. That's better in the long run.
Glenn: Right now h2 have the top level headings?
Philippe: That's in the body, but
in the head you have h1 with title and h2 with date etc.
... In between some folk are adding another h2 in between with
a subtitle.
Glenn: You can have an h1 inside head in HTML?
Philippe: I didn't mean the html
I meant the div class/id="head" - the one that contains
the
... logo, the title, the copyright info, etc, those kind of
stuff.
Glenn: The top level div, let me
see...
... there's a div class="head" I see.
... OK I see in TTML1 the title contains the 3rd Edition and
the h2 which has the Editor's Draft ...
Philippe: The proposal is to move
the "3rd Edition" in to the h2 and teach specref about that
h2.
... Visually you won't see anything different because the same
text will be there but specref
... can pick the h2 or not depending on what we teach it.
Glenn: Right now there's a "W3C Editor's Draft" in the h2.
Philippe: Yes and we're proposing to add another one.
<plh> https://www.w3.org/TR/pointerevents2/
Philippe: For example the pointer
events spec does exactly that.
... It's called Pointer Events but if you go to the spec it
says it's level 2. The intent is to
... supersede level 1 in the future.
Glenn: I see. So it takes it out of the title, there's a new line there, basically.
Philippe: Correct
... The effect that it has is that in specref if you put
pointerevents2 you will see that it points
... to the working draft but the title is just Pointer Events
because specref doesn't get the h2
... information from our systems.
Nigel: I see, the title is just from the h1 - because specref doesn't know about the h2 yet.
Philippe: Correct.
Glenn: There's a meta issue about what should the specific reference be from IMSC.
Nigel: That's a drift off this topic - it does need to be covered but that's different.
Glenn: If you put a date into specref then it points you to a specific dated version of the doc.
Pierre: That's the point, specref
with a version specific URL will include the version, but
... a generic latest link would just have the title without the
version.
Philippe: If you're talking about
latest version, you need to make sure the right title is
returned.
... That could be in our backend.
Glenn: It sounds like you need changes to specref and we need to change the title to put the edition into an h2?
Philippe: Right, if you want to do that, then we do need to make changes to specref too.
Glenn: Is your proposal to put the version info into the h2 Pierre?
Pierre: Yes that's what I would do if that's the direction we're going in.
Glenn: Even though that breaks our convention today?
Pierre: Yes - do you know of anyone who uses the title with version in the h1 today?
Glenn: Impossible to know.
Pierre: It seems like W3C is
going in that direction generally so I would follow it.
... We're just discussing modifying the TTML1 h1 to allow the
automatic bibliographic references to work.
... It would be to take "3rd Edition" out of the h1 and put it
in the h2.
Glenn: My preference would be to
modify Specref in such a way that we don't have to change
... the title to a different format.
Philippe: No, that's not going to happen and I can't propagate that to other groups.
Pierre: We could introduce extra metadata duplicating the information but that's undesirable.
Philippe: We already have people using respec to create these h2s.
Glenn: I was suggesting that
specref can look for metadata if it is there and if absent
then
... use the h2.
Philippe: That would be a special
case and I'm not going to allow special cases, for sure,
sorry.
... It's not worth our development time unless I can propagate
the change.
Glenn: The problem I see is it
breaks continuity with the past titling convention, that's
all.
... It would also require a change in TTML2. It's just a
formatting, stylistic change. We could
... make that change. Give me a few days to convince myself
it's okay.
Nigel: I think we should make an
assumptive decision to change to version in h2 and then
... you can tell us if you find any problem Glenn.
Pierre: I can just create a pull request for us to review, that's more concrete.
Nigel: Yes
Glenn: Then we can look at it.
Nigel: Pierre, I think you can go ahead and do that please.
SUMMARY: @palemieux to prepare a pull request moving the version into an h2
github-bot, end topic
<plh> https://github.com/w3c/ttml1-tests/ and https://github.com/w3c/ttml2-tests/ are live
Nigel: Do we have an issue about the reference?
Pierre: IMSC 1.0.1 references the latest TTML1
Nigel: It's not raised as an issue.
Glenn: We haven't discussed this - it is an open question.
Pierre: Presumably the difference
between editions is to correct bugs and things that ought to be
fixed.
... So it doesn't seem unreasonable for a profile of TTML1 to
reference the latest edition.
... I could be persuaded otherwise.
Glenn: Up until 3rd Ed we insisted on only editorial changes, but in 3rd ed there are substantive changes.
Pierre: Yes, but regardless of
whether the changes were substantive, they are bug fixes,
... not feature additions. So if someone comes in and says
"which edition should I implement"
... we would always say the latest edition.
Philippe: Yes, that's what our system would say as well.
Glenn: It would be interesting to get other feedback on this, Mike may have a different view.
Nigel: Yes, it's a technical
issue - if an implementor implements the spec precisely
following
... the latest edition and then that changes underneath, then
the state of conformance of
... that implementation is unclear. It could be that it's not a
big enough problem to worry about.
Pierre: Yes I could be persuaded
either way.
... This is confusing from an external point of view too and
causes arguments. Maybe I'm
... persuading myself to point to a specific edition.
Glenn: In the case of TTML1
referencing XML we point to a specific version/edition of
XML
... so that doesn't arise.
Nigel: It seems to be a matter of
good practice to reference a specific version but in doing
that
... to accept that if the referenced spec changes then the
referring spec needs to be updated
... and that should be transparent.
Glenn: In TTML we have normative references to XML 1.0 and XML 1.1 and both have editions in the title.
Philippe: Which link do you use?
Glenn: We use the dated
links.
... The title has the edition and the link is dated. That's the
practice we followed in TTML.
Pierre: Listening to this I'm
starting to lean towards referencing a specific version in IMSC
1.1.
... Maybe it wasn't a good idea in IMSC 1.0.1 to reference
TTML1 without dates.
Philippe: The only problem is if
you reference a dated version and you make a security
... update that you want everyone to pick up then they
won't.
Pierre: The tooling on W3C...
Imagine IMSC 1.1 makes a hard reference to TTML1 2nd Ed.
... Imagine in the meantime it was superseded because of a
terrible security bug. When you
... get to it you would get a huge warning saying "don't use
this"?
Philippe: yes
Pierre: Doesn't that address the issue?
Philippe: If you don't mind this
then realise that at the end you are directing the reader
to
... a specific version.
Pierre: At least it gives the
reader the ability to do a diff and see what changed.
... It also means, to Nigel's point, that as a group we have to
be more active in updating stuff.
... It is more work for us but it's more precise maybe.
Philippe: If you guys believe
there is the use case to be that precise then sure. At the end
of
... the day it does not matter which reference you state,
people are going to use whichever
... XML parser they are familiar with. They're not going to
write their own.
Glenn: On the other hand, from a
point of view of conformance and testing you might have
... an issue there, but that's a separate point. Right now we
have referential integrity
... with respect to TTML references. We went through TTML2
recently.
... When we refer to RFC we don't have this issue, because they
don't change the text.
Nigel: They change the text to point the reader to a newer version.
Glenn: That's true, but it doesn't change conformance wrt that spec.
Philippe: This is topical because the AB is discussing living standards.
Pierre: In some groups people
want to be precise because they are putting stuff on
shelves
... for years and they want to know exactly which versions they
reference.
Glenn: I'm reminded of DOS, and
switches in code to handle different compatibility bits.
... That's what you end up with when you were very precise.
Nigel: There's a precise mirror
here with python, node, java etc with tests for a piece of
software
... passing given a specific set of 3rd party library versions,
and capturing a freeze point
... of those versions. We could do that with our specs, and
associate a test suite with a specific
... set of reference versions.
Glenn: Right now the tests are
based on a specific version of a spec, you could make
that
... the other way around.
... In the case of XML 1.1 it refers to the specific edit in
place date.
Nigel: I'm not sure if we need an issue against IMSC 1.1 to reference a specific version of TTML2?
Glenn: I heard Pierre say he is leaning towards doing that.
github: https://github.com/w3c/imsc/issues/381
SUMMARY: Discussion ongoing, group leaning towards a specific dated reference to TTML
github-bot, end topic
<plh> https://github.com/w3c/ttml1-tests/ and https://github.com/w3c/ttml2-tests/ are live
Philippe: I've created the repos
and put them in the Timed Text team so you guys should
... all have write access to them. I still have some generic
files to add.
Nigel: Thank you!
Philippe: FYI we started the
superseding of IMSC 1.0 in favour of IMSC 1.0.1 as
decided
... by the WG in April. We started the proposal at the AC
level. Assuming everything else
... goes well then by mid-June we can declare IMSC 1.0
superseded, and all of the links
... will be updated as appropriate.
Nigel: Thank you!
Philippe: We are publishing the
CR for WebVTT today so that was the last publication in
my
... pipeline for this WG.
Nigel: Thank you
Glenn: I just closed TTML2 issues
699 and 715 by merging the approved pull requests that
... had been open for a while.
Nigel: Thank you.
... We seem to have covered our agenda for today, so I'll
adjourn. Just a note that in
... 2 weeks' time there's no meeting due to the IRT industry
event.
... Thanks everyone! [adjourns meeting]