<vivien> recording in progress
Bernard starting the meeting
...agenda: content hints, ... (see slides)
<vivien> Draft Charter
Bernard: proposal out, we've made
some changes based on feedback
... See slides for details
Bernard: timeline added to
charter
... that's where we are
... comments?
Cullen: I'd like us to have a f2f
meeting to dig in to proposals on how we get started on
v2
... before TPAC
... not a charter issue though
Stefan: plan to send charter to AC review by March 1st.
<vivien> Issue 478
<jesup> does someone have a link to the github issue where this was discussed a while back?
<jesup> not this issue linked in the slide
Thomas talking about content hints
Thomas: detail vs motion for video
Bernard: need to get to a decision (see slide)
Jesup: I've discussed this a fair
bit, and overall I'm positive of the idea of content
hints
... the question is where they would live and how to keep track
of them (track or sink)
... what happens when data gets manipulated by Web Audio (would
hint be carried through?)
... my inclination is to have it part of the sink
... how to define how you tag things (e.g. we want to add
something new)
Jan-Ivar: I wonder if there is
any precedence in the web platform
... dislike the generality
Thomas: the only hint would be the rate degradation hint, could perhaps use the degradationPreference instead
Cullen: codecs can't figure out
is our experience, so we need a knob
... don't care very much where we attach it, but must be
standardized
Jesup: agree to Cullen
Jan-Iver: worry about web platform tests
Jesup: implementation dependent
hta: we do have tests that will
fail if video is black
... the thing about hints is that it means that you prefer that
you bias things, it's not command
... hard for the codec to guess what it should do
... I think giving the platfomr more info is better than tool
to shoot yourself in the foot
... info about content, so like to have it on the track
Jan-Ivar: Care about consistent defaults
hta: don't agree, consider EC, we don't specify it
jan-ivar: difficult to write a test for that (but we're working on it)
hta: we could have test that checks video metric
vr00m: this is really useful. Prefer it to be on the track. Testing: can be done (losing detail)
(discussion about test)
adambe: if we put it on the track, do we duplicate what we can do with constraints?
Bernard+Jan-ivar: not for video, perhaps for audio
adambe: can you see result by checking settings?
jan-ivar: (scribe lost
detail)
... some overlap
bernard: hints is about content rather than a control surface
jesup: inhertently for hints they
are describing properties, and can have several
properties
... need a list so sink can can use as it sees fit
... can ignore hint if it can't do anything with it
... direct control should be on the sender
Mo: my recommendation: think this a very minor use case, and apps will get it wrong. We should be very careful and clear that this is overriding the browsers guess
Jesup: I think we have reasons
for having this
... this is better than no solution
... app can know more about the content
Mo: I think you can accomplish
the same thing with constraints (set low framerate for example)
and degradationPreference
... we should guide users to use the existing knobs in the
right way
Thomas: we don't want surprising results that are not consistent across browsers
Cullen: the Opus neural net
detector fails
... if the app knows about the content we should allow it to
tell
... we should go with the proposal on the table
hta: reason for existing prop is
failing to detect content type
... sometime the app knows better than browser
jan-ivar: terrible api. not testable, not mandatory, can be lost (cloning) - need to copy along. Many chances for error.
jan-ivar: also needs to be smart
enough to know when to ignore a hint
... also redundancy (we already have constraints)
cullen: the other things don't do
this (hints)
... covers important use cases
... I don't understand why we should not have this
Jan-ivar: not testable
cullen: test as EC
(details)
... codec can use it or not
jan-ivar: highly redundant
cullen: how?
(silence something wrong with webex, we can't hear jan-ivar and jesup)
... jib and jesup are back)
(discussion about overlap between constraint and hint)
jib: lets go for video - framerate
cullen: different thing than motion vs detail
jesup: we can have both hints and
direct control (such as constraints)
... I in favor (some detail to work out though)
... only q is if it should be on the track on the sink
cullen: agree
jesup: I see putting it on the sink superset of on the track
mo: why would a minimum framerate fix the content share problem?
hta: don't know what the framerate should be
jesup: what to do if you cant support mimum framerate
cullen: if we set it at 1 fps you
get mouse movement problems
... hard to find the right fps setting
jib: if this is an hint can be ignored the developer gets surprised
jesup: depends on codec, some codecs makes no difference
suhas: I expect users can use
this
... as long as we explain it's a hint (but can be ignored)
jib: can a hint be removed?
hta: can be changed over time
jesup: must be able to provide multiple hints
jib: seems like a mistake to me
jesup: for this, can help hash out details
mo: for this, but reword (browser override)
cullen: agree
jib, hta: agree
Consensus to do something, need alternative proposals
mo and jesup will work on hashing out details
mo: browsers must try to detect type of content
hta: does not always work
mo: more important that browsers try to do the right thing
<vivien> Issue 29
ScreenShare presented by Suhas
Suhas: what do people think about the proposal in the slide?
jib: likes this
cullen: would this work on most OSes?
bernard: depends on what you want to share (many windows and monitors)
<jib> clarification: jib likes what suhas said which is different from slide: basically add prose that UAs figure it out automatically
cullen: we need to be able to pick a specific ppt window
jesup: be careful about timing of
changes (resizing of windows for example) that can show things
that are not supposed to be shown
... security consideration thing
jib: what is mapping failure?
suhas: full screen + esc sequence. Would let the UA make the decision
<vivien> Issue 31
jib: downscaling necessary, want to use
constraints
...proposal: explicit list, no cropping
jib: comments?
We'll go with this direction
<vivien> Issue 39
jib: not sure what we should do
(always disallow or disallow by default)
... I'd like to hear what others think
suhas: good use case
... but I understand the risk
... is about how we ensure the trust
cullen: lean to allow in some way
(disallow by default)
... seems like a good compromise
jesup: agree
... should we give any guidance on how to try communicate the
transfer of trust
jib: agree, we already have the problem for cam and mike
jesup: also the risk of skipping iframes
<vivien> Issue 43
bernard: proposal is to close Issue
(agreement to close)
<vivien> Issue 49
<jesup> +1 for me on issue 49
support for making the change
<vivien> [meeting adjourned]
<hta> ok, out of time.