- From: Dan Burnett <dburnett@voxeo.com>
- Date: Wed, 9 Nov 2011 04:57:35 -0500
- To: public-xg-htmlspeech@w3.org
Group,
The minutes from our second day are at http://www.w3.org/2011/11/04-htmlspeech-minutes.html.
For convenience, I have pasted a text version below.
-- dan
*****************************
HTML Speech Incubator Group Face to Face day 2
04 Nov 2011
See also: [2]IRC log
[2] http://www.w3.org/2011/11/04-htmlspeech-irc
Attendees
Present
DanB, Michael, Glen, Matt, Robert, Patrick, Avery, Nagesh,
Debbie, Bertha, Milan, Rahul, DanD
Regrets
Chair
Dan_Burnett, Michael_Bodell
Scribe
avery, Matt
Contents
* [3]Topics
1. [4]Architecture picture
2. [5]Requirements
3. [6]Design Decisions
4. [7]Revisit the picture from Dan
5. [8]Planning
6. [9]Finishing the Incubator Group
* [10]Summary of Action Items
_________________________________________________________
<matt> scribe:avery
Protocol issue 1: Minor typo.
Milan: What does the speechinputresult look like when the null event
happens.
... question the usefulness of having result at all.
robert: It should be undefined, meaning it's null.
mbodell: We've discussed this before and agreed to have it this way
so let's not change.
... discussion about what to null out in this case.
Milan: e.g. can you null out the third if there's a fourth?
mbodell: yes,
... More discussion about what to do in this case.
robert: Maybe a new event.
Mbodell: Maybe this should be a new event, when there's an empty
message.
Burn: Clarify this is at protocol level or API level
mbodell: API level
Robert: There should be a new event with the compressed
resultHistory array.
avery: what should we call this new event?
Burn: Requested that Mbodell clean up this issue about whether we
need a new event for the empty message case (nulling out previous
candidate reco).
Robert: This is not Protocol issue 1 itself, but is a result of it.
Milan: Maybe we need something like "reorder'" rather than "delete"
for the new event.
Robert: In the future we could develop a new pattern and call it
resulthistoryupdated event. It could be a can of worms because we
would fire that for every reco.
<scribe> ACTION: mbodell to create the new event for this case. See
discussion above for details. [recorded in
[11]http://www.w3.org/2011/11/04-htmlspeech-minutes.html#action01]
<trackbot> Sorry, couldn't find user - mbodell
<matt> trackbot, list users
<trackbot> Sorry, matt, I don't understand 'trackbot, list users'.
Please refer to [12]http://www.w3.org/2005/06/tracker/irc for help
[12] http://www.w3.org/2005/06/tracker/irc
<matt> trackbot, status
<matt> ACTION: Michael to create the new event for this case. See
discussion above for details. [recorded in
[13]http://www.w3.org/2011/11/04-htmlspeech-minutes.html#action02]
<trackbot> Sorry, amibiguous username (more than one match) -
Michael
<trackbot> Try using a different identifier, such as family name or
username (eg. mbodell2, mjohnsto)
<matt> ACTION: mbodell2 to create the new event for this case. See
discussion above for details. [recorded in
[14]http://www.w3.org/2011/11/04-htmlspeech-minutes.html#action03]
<trackbot> Created ACTION-5 - Create the new event for this case.
See discussion above for details. [on Michael Bodell - due
2011-11-11].
Milan: I'd like Robert's example updated, especially to know how to
use the waveform API.
Dand: We should move on to requirements.
Break time.
<mbodell> scribenick: mbodell
Architecture picture
Dand: Need something simple to show how the API + Protocol + web
server + UA all fit together
<smaug> which diagram?
<smaug> mbodell: could you post a link?
Robert: Should have builtin and proxy as on the same level
Milan: but builtin can point to proxy
mbodell: really 2 different orhtogonal ideas: Use UA default or
author specified; either of which could be on the same device or in
the cloud
Glen: I disagree with TTS+ASR being specified in one thing as par t
of remote speech service, I think they are different
Dand: Yeah too much detail, I'll just taqke out TTS + ASR and leave
it at Remote Speech Service
Dan: Could put Audio Capture API as peer of WEB API
... Default is a better word than builtin
... [audio capture APIs] -> [Web API]
... / |
... [default] [Author Selected]
... | X |
... [Local] [Remote]
Robert: But the default could actually be a hybrid of both local and
remote
Matt: What is the purpose of the drawing?
... Can we start simple and then go to more complex?
Milan: I like Dan's drawing
Mbodell: Where does the protocol come in?
Dan: As the circle between Author Selected and local and remote
Glen: What are the dotted lines? Things we don't control?
Dan: Yeah, and audio capture API and defalut and the arrows to local
and remote
<smaug> thanks
Dan: Change Author Selected to Speech Service Proxy
Group discusses how to make google not take over all his browsers
Requirements
dan: Here are the requirements that the group as a whole started
working on
Link to document:
[15]http://www.w3.org/2005/Incubator/htmlspeech/finalreport/XGR-html
speech.html#requirements
[15] http://www.w3.org/2005/Incubator/htmlspeech/finalreport/XGR-htmlspeech.html#requirements
dan: My suggestion was to take these and include them, perhaps as an
agenda
Robert: My suggestion was to reword requirements in place
Dan: What about when a requirement becomes two different
requirements
Robert: we could just note it with those
Milan: Do we even need these, couldn't we just stick with design
decisions
Robert: I think we do because, it is the only paper trail we have to
the voting we did on prioritization
Dan: I think I like Roberts plan better
Robert: Marc went through and did one pass, and then Marc, Milan,
Robert, Michael and others did one around August
<scribe> ACTION: Robert to fish out the latest version of that
[recorded in
[16]http://www.w3.org/2011/11/04-htmlspeech-minutes.html#action04]
<trackbot> Created ACTION-6 - Fish out the latest version of that
[on Robert Brown - due 2011-11-11].
Dan: That was just for protocol work
... What about for web api?
MBodell: We originally had Raj on tap for that, but he had family
issues so then we distributed it to everyone, only Debbie had any so
there are some from here proposal
Dan: Anyone concerned about me using that to consolodate, I'm not
Group: No one has concerns
Debbie's document
[17]http://lists.w3.org/Archives/Public/public-xg-htmlspeech/2011Aug
/att-0011/ASRPropertiesAPI.html
[17] http://lists.w3.org/Archives/Public/public-xg-htmlspeech/2011Aug/att-0011/ASRPropertiesAPI.html
Design Decisions
Dan: I think each of the things we talked about in requirements also
include some design decissions
Robert: What about something like 74 "Bjorn's email", that is tough
to tell since Bjorn has sent a lot of email
Dan: I could just take this out
Robert: What about linking to email
Glen: The nice thing about that is you can see the archives and see
the linking to other threads of discussions, etc.
<matt> [18]Bjorn's mail for requirement 75
[18] http://lists.w3.org/Archives/Public/public-xg-htmlspeech/2011May/0030.html
Dan: I don't want to go down the path of linking each design
discssion to an email thread
Glen: not for every one, but just when appropriate
Dan: Yeah, when appropriate is fine
... There are some new one's at the bottom
... These (86+) were ones that we'd talked about in phone calls, but
had never made it to earlier drafts of the final report
Mbodell: we did implement these
dan: I don't think we implemented 89
mbodell: we did for nomatch, but not for noinput, noinput is an eror
Milan: Do we have a timeout parameter?
Mbodell: No, but can use your own JS timer.
Robert: Should we add noinput as a separate event and not an error
Milan: I like that, should also add the property
Glen: How do you define what noinput means? There is always some
noice
Avery: On speech starting that stops the timing
Glen: But you as an author can track that with these events
Milan: But cleaner to do that with a parameter
MBodell: we don't today have that, but we are discussing that
Robert: I think we should add it, it is a very common UI
Glen: I think we could add the parameter, not so sure about the
event
Milan: Why not have them?
Glen: Why not just put the noinput into the nomatch event
dan: Not sure
Group: Agreed to add the timeout parameter
... Agreed to add timeout event
Dan: Need someone to add text
Glen: Why not just define this event as speechstart event hasn't
happened in a certain length of time
<glen> and define the timeout this way as well
<scribe> ACTION: Dan to add it to the spec (mbodell to do it if he
doesn't) [recorded in
[19]http://www.w3.org/2011/11/04-htmlspeech-minutes.html#action05]
<trackbot> Sorry, amibiguous username (more than one match) - Dan
<trackbot> Try using a different identifier, such as family name or
username (eg. dburnett, ddruta)
<glen> perhaps named speechStartTimeout
<scribe> ACTION: dan add end to the reco element [recorded in
[20]http://www.w3.org/2011/11/04-htmlspeech-minutes.html#action06]
<trackbot> Sorry, amibiguous username (more than one match) - dan
<trackbot> Try using a different identifier, such as family name or
username (eg. dburnett, ddruta)
<glen> and onnospeech event
group: Agrees to parameter is speechStartTimeout and event is
nospeech
<scribe> ACTION: dburnett to add it to the spec (mbodell to do it if
he doesn't) [recorded in
[21]http://www.w3.org/2011/11/04-htmlspeech-minutes.html#action07]
<trackbot> Created ACTION-7 - Add it to the spec (mbodell to do it
if he doesn't) [on Daniel Burnett - due 2011-11-11].
<scribe> ACTION: dburnett add end to the reco element [recorded in
[22]http://www.w3.org/2011/11/04-htmlspeech-minutes.html#action08]
<trackbot> Created ACTION-8 - Add end to the reco element [on Daniel
Burnett - due 2011-11-11].
<scribe> ACTION: dburnett add the sentence about weights from dd90
into the spec where we describe weights [recorded in
[23]http://www.w3.org/2011/11/04-htmlspeech-minutes.html#action09]
<trackbot> Created ACTION-9 - Add the sentence about weights from
dd90 into the spec where we describe weights [on Daniel Burnett -
due 2011-11-11].
dan: I think we are done with that
Revisit the picture from Dan
<glen>
[24]http://lists.w3.org/Archives/Public/public-xg-htmlspeech/2011Nov
/0049.html
[24] http://lists.w3.org/Archives/Public/public-xg-htmlspeech/2011Nov/0049.html
<matt> [25]DanD's Diagram
[25] http://lists.w3.org/Archives/Public/public-xg-htmlspeech/2011Nov/0049.html
mbodell: Can we relabel speech service proxy as author selected
speech service
dand: No the arrow from Speech web api to speech service proxy is
where the author selected is
... we could also label the speech web api to default with UA
selected
<Robert_> I think this is the latest redraft of requirements from
the protocol group:
[26]http://lists.w3.org/Archives/Public/public-xg-htmlspeech/2011Jul
/att-0023/Protocol_requirements_draft_-_RB.htm
[26] http://lists.w3.org/Archives/Public/public-xg-htmlspeech/2011Jul/att-0023/Protocol_requirements_draft_-_RB.htm
Glen: How about Default Speech and Speech Service with the labels of
UA selected and user agent
<smaug> why is default out of scope?
<smaug> it is not more out of scope than remote speech service
<Milan> Agreed
Olli: The remote speech service is out of scope, it is the protocol
that is in scope
Milan: Agreed
... Didn't agree that default was in scope
Olli: thinks they are equal, fine if default is out as long as it is
the same from remote
Dan: actually the speech service proxy is out of socpe
mbodell: Yes, we used to have conceptual boxes for protocol and api
and the boxes were in the scope, but we've moved the stuff to the
arrows, so now the boxes, except the speech web api are all out of
scope
Group: Obvious, but the user agent and device are out of scope too,
but probably so obvious don't need to do it
mbodell: could invert the dotted lines and dashed lines
dan: Don't like that
group waits for mail
We break for lunch
<matt> [27]Updated diagram take 2
[27] http://lists.w3.org/Archives/Public/public-xg-htmlspeech/2011Nov/0051.html
<matt> Meeting; HTML Speech XG F2F, Day 2
<Avery> Found a bug in one of the examples in Section 7.1.8
<Avery> The example is "Hide possible graphical UI related to reco
element if permission is denied"
<smaug> Avery: what is the bug?
<Avery> NOT_AUTHORIZED is outside the comment but should be inside
it.
<Avery> The first comment in the sample.
<smaug> indeed
<Avery> It's a nit but we should fix it.
<smaug> yup
<matt> Scribe: Matt
Planning
burn: 2 areas of planning we need to discuss: after the group is
finished and what the group needs to do to finish.
... I would like to see W3C create a new WG to work on this effort.
Why?
... I don't think there is another group right now that could
actually productively take this work. The HTML WG is severely
overloaded. Work can happen in that group, I don't think you
actually get any additional productive participation than you would
get from your own WG. No standing teleconference times, or anything.
... It's probably good for small features, you can start a thread
and come to a conclusion in a short amount of time.
... WebApps is another consideration, and it seems reasonable, but
from what I understand WebApps is overloaded and it's charter is in
two parts, a general part that could include our work, and a very
specific part saying what they'll do exactly.
... They would have to recharter and the work is quite different
from what they do today.
... It could work there, but I think there's a reluctance by the
group to take on new work.
... Anyway, I think having our own WG is the best way to be able to
start relatively quickly, scope the IP discussions enough that
lawyers won't take forever to let us join the group.
... It will also make it easier to get the work exposure.
Milan: And what about VBWG?
burn: When this XG started, there was intense interest in it
explicitly not being in VBWG.
... So, I never really considered it.
mbodell: MMI is another candidate, but they don't seem to have the
participation we need.
burn: I don't think it'd be appropriate as the sole goal of the MMI
WG.
glenn: How hard is it to create a WG?
<Milan> matt: Write a charter
<Milan> matt: ... which is straightforward
<Milan> ... submit to w3c management
<Milan> ... goes to AC for a vote
<Milan> ... which takes 4-6 weeks
<smaug> what about protocol? W3C or IETF?
<Milan> ... minor edits
<Milan> ... mainly depends if we have agreement in this room
<Milan> Dan: To address Ollie and Avery's point, yes protocol isn't
an easy fit with W3C
<Milan> Matt: What about websockets?
<Milan> ... probably very little resistance to moving protocol to
IETF
burn: They definitely consider Web Sockets a transport, so it'd be a
good candidate for IETF.
... Working through a charter, we'll have to find if anyone in the
XG right now wants it to happen at W3C?
... That's something we'd have to decide as part of the charter
work.
... My initial thought was that we would say that we would work
together with it at IETF.
Avery: There's other details about the audio in/out.
matt: There is a part of a charter that details interdependencies. I
put VBWG, HTML, AudioWG, WebRTC, WebApps, and others.
rahul: Web and TV.
matt: I'll add that too.
Milan: Google would be amenable to having it be a new WG?
Glen: I think we'd be interested in it.
Milan: But Bjorn wanted it in WebApps.
burn: The big issue was they wanted Web developers participating. If
that happens, it's good.
... Talking around the w3c it seems that whether it's in WebApps or
not is not significant to decide if that happens.
... Not sure Bjorn believes that.
... But we should make sure we get Web developers.
Glenn: We don't want it in a group that bogs down and takes forever.
... What I liked about having our own WG is that we can streamline
the process.
... And having our own WG means not having to educate another WG on
speech again.
burn: We'd have to educate them, and then convince them that the
work belongs there. It's not easy.
dand: And getting them to prioritize the work.
burn: One of the questions would be are there staffing concerns that
could affect the creation of the WG?
... This isn't something we have to decide, but where would this
work live, where would the group live?
matt: We have this domain structure that really just kind of
reflects our management structure. My gut says that the work would
end up in the UbiWeb domain alongside the VBWG and Geolocation, etc.
Glenn: Is this WG a sufficient number of companies?
burn: It's sufficient but --
mbodell: We'd certainly rather have more.
... The charter also applies to what we can work on, so we can't
move into other spaces. It's what we bring to lawyers to do patent
evaluation stuff.
burn: The more clearly focused the work the better.
... I think the work of our XG really informed the focus.
<smaug> I think Mozilla would participate
burn: Straw poll, do you think if a WG were to form for this, would
your company be interested in joining?
<unanimous>
burn: Olli has some of the best input on end user protection, stuff
we really need.
... Matt and I sat after lunch and began hacking on a charter.
... Activities after this week, what else?
Robert_: What's the IETF plan?
burn: At the last IETF meeting, I spoke to some people there about
it. I explained what we were doing and what the work was, and gave a
headsup.
... The reason to do that is that at IETF you start with a BOF
session. Everyone sits around and works on a charter.
... Oddly enough those session started as a way to get people
together to chat, but now in order for them to be a success you have
to work months in advance to get a charter going and get everyone to
agree.
... I am willing to help anyone who would like to do that, but I
would really really appreciate that if this is something you would
like to see happen that you could find your existing IETF
representative/s and let them know about this.
... I'll speak with them and work together with them.
mbodell: Those happen at the F2Fs?
burn: Yes, not the one next week, but the one in March in Paris.
... So, now would be the time to talk to someone at your company who
is going to that and put me in contact with them and we can have
initial discussions at the meeting.
Robert_: You've mentioned Bernard yesterday.
burn: I talked to him at the last meeting, but you should talk to
him directly.
Robert_: He's in the building next to us, we should go talk to him.
mbodell: How does the structure and organization of the IETF compare
to W3C? I know their wiki page says there are no members of the
IETF.
burn: IETF there are no organizations, everyone represents
themselves.
... Largely a meritocracy.
... Discussions are primarily on mailing lists. You can do
everything via email.
... Any decisions that are made at F2F's are tentative until
approved on the mailing list.
... IETF meetings tend to just be 2 hours long, and that's all the
time you get for a week.
... Individuals submit "internet drafts" and say "hey, this should
be a standard"
... That is one of the possible ways to get work started.
Glenn: When are internet drafts submitted? Presumably have authors
listed with their company affiliations, correct?
burn: Yes.
... Drafts go in -- people tend to participate on their personal
accounts btw.
... There is no membership, anyone can join anything at IETF.
mbodell: What about IP?
burn: At every IETF meeting, they present a Note Well document that
has some standard language that approximately says: Anything you
contribute by speaking, typing or any information you volunteer then
IETF owns it unless you disclose otherwise.
... It's pretty strongly worded and quite clear.
... When people send drafts, like I mentioned, there are still tons
of internet drafts and people say "let's use this one as a starting
point" and if there is rough consensus the group adopts it as a WG
document and it changes from having an individuals name to a WGs
name and proceeds on spec track.
... IETF works on "rough consensus" and "running code". They don't
like votes.
... Maybe you are sitting in the room and your boss is sitting in
the room with you, and you have differing opinions and are
uncomfortable raising your hand. People will hum instead.
... It's often very clear.
... Large amounts on both sides means no consensus. A large amount
on one side and few on the other means rough consensus.
... If people agree and implementations come out, it's a go.
... IETF very highly values implementation experience.
... I don't plan to do anything in particular right now with the
protocol.
... If someone wants to see it go forward, I am happy to help. If
you have reps there I am happy to work with them. Voxeo likes this
work overall, and for me to say "yeah, me too" isn't that hard.
Right now, I'd be the person doing it, and I don't have enough
cycles to create it myself.
Robert_: We'll talk to Bernard Aboba on Tuesday and see if he's
interested.
burn: Google has a number of people involved in IETF, from Google
Harvald, the co-chair of WebRTC is also at IETF.
mbodell: I think if we are doing the protocol we should try to kick
it off at the same time.
burn: Some of you were talking to Cullen, he's also an IETF guy.
<Milan> Harald Alvestrand <harald@alvestrand.no>
Finishing the Incubator Group
burn: Anyone have opinions on teleconferences between now and the
end of the group.
mbodell: If we don't take a week off next week, there are 3 weeks
before the end of November. 10th, 17th, 24th. 24th is Thanksgiving
Day.
... If we were on a normal schedule, we'd only have the 10th and the
17th.
... I think we want to finish with the conference on the 17th.
burn: Yes.
DanD: In terms of going beyond commitment to this work. Are there
any plans to prototype what we have here before we go to a WG or
not?
mbodell: We're not really standardizing anything here.
DanD: There are solutions already, adapting the solutions is as much
as changing what you have with or without an intermediary step. If
we can have a way to start evangelizing the concept, within the
community in general with a working solution?
burn: Early on Google said they had implemented the "HTML Speech
Standard"
DanD: Not how we defined it though.
mbodell: It's still not a standard though, so have to be careful to
say that.
burn: I can't answer that question.
Robert_: Undoubtedly we'll have prototypes.
burn: Part of the Recommendation process is that there are
implementations.
mbodell: There are two ways this stuff gets done. From the standards
POV it gets done later in the standardization process. That said,
the industry is always innovating and prototyping all the time and
are free to do so.
burn: When you issue the CR document it also says it's a call for
implementations. It doesn't mean that people can't implement before
that. Once you get to LC you really should be implementing, long
before CR.
mbodell: So, the timeline. Call on the 10th and the 17th. Going to
be a lot of work, but the 17th should be the last call we have.
... Hopefully everyone will sign off on it then, or figure out small
changes over email.
matt: You can also work not he charter in parallel.
burn: It's not a deliverable of the XG, so you probably won't see it
on the list, as that might be inappropriate.
... Michael and I have to get all of the changes in before the next
call on the 10th.
mbodell: If you submit any changes or examples, it would be really
good if you could double check and get them as ready as possible. We
can work this on the list too don't wait until the call.
burn: Please do work on this now, as mbodell and I will be.
... In addition to the report we have to provide a few short
paragraphs of what we were about and what we did and what we might
do next.
... I'll be working on that, it may go the list or not, maybe a
version. Depends on timing. Keep in mind that this isn't a document
that is going to be edited over and over.
mbodell: If no one objects, two weeks from today we're done. Final
report and summary are sent out.
burn: After that, we're done.
matt: You can leave this ml open for a while, but you can also have
an ml start up when you get the charter firmed up.
burn: Sounds good. What should we call it? HTML Speech?
DanD: When it came to markup, we said no markup, but we say "HTML
Speech"?
mbodell: We do have speech bindings.
DanD: When I hear HTML Speech, I think markup.
burn: It's been argued the other way that HTML 5 is more about the
JavaScript than the markup.
mbodell: I think it is the appropriate name, even if it's not
markup.
... It is part of the HTML ecosystem.
nagesh: HTML 5 Speech?
<objections>
burn: S.peech E.vents E.xtending C.ommon H.tml
... I think we are done.
... Thank you!
<mbodell> Mbodell: S.peech P.rotocol for E.vents and E.xtending
C.ommon H.tml == Speech
Received on Wednesday, 9 November 2011 09:58:11 UTC