<trackbot> Date: 12 September 2012

wtf, zakim

sigh

<scribe> Meeting: W3C Audio Working Group Meeting

<chris> conference code?

<jussi> chris: 26631

<chris> jussi: thanks

-> Agenda is http://lists.w3.org/Archives/Public/public-audio/2012JulSep/0740.html

<ChrisWilson> "the conference is restricted at this time"??

Chris, the code is 26631

<ChrisWilson> mute jussi

<joe> I cannot connect

joe, the code is 26631 today

<joe> thx]

<joe> sorry, now Zakim is not even answering the regular number

Interface naming (update)

Olivier: should we rename notegrainon to startgrain?

<joe> Is there an alternate Zakim number, cannot reach 617 761 6200

cwilso: sounds consistent

crogers: we could overload start

… depending on number of arguments

… but the name would always be start

<jussi> +1

cwilso: sounds better

<tross> +1

<ChrisWilson> +1

<joe> +1

joe, did you manage to connect or are you on IRC only?

<joe> irc only, Zakim appears to be down to my phone

<joe> "non-working number" recording

sorry to hear, will be taking notes

<joe> np, let's go w irc

<scribe> ACTION: ot to add overloading of start to bugzilla [recorded in http://www.w3.org/2012/09/12-audio-minutes.html#action01]

<trackbot> Created ACTION-49 - Add overloading of start to bugzilla [on Olivier Thereaux - due 2012-09-19].

Olivier: crogers, any idea when the changes could get into the spec

crogers: busy with implementation at the moment, but will try to get changes there in next week or two

Olivier: sounds good.

[skipping loop start/stop points for now]

Revisiting the circular graph issue

https://www.w3.org/Bugs/Public/show_bug.cgi?id=17326

Olivier: do we need more knowledge/understanding to go forward?

crogers: think we have a good understanding

… problem is when you have a delay in a feedback loop and do block processing, the block size is the minimum delay

… we've defined that value to be 128 sample frames

… depending on sample rate, this value defines a time in seconds

… no restriction on delay time when handled in seconds

… think we've decided not to throw exception

… and silently clamp to the minimum

Olivier: what do people (here) think of the proposal?

[no objection]

<joe> it seems fine

thanks joe

Olivier: does this need to be added to the spec?

crogers: I think so
... ROC was also asking what if you have several such delay nodes

… may not need to all be clamped

… opinion that we should not require optimisation, each node would be taken separately and be clamped if they are in a feedback loop

Olivier: are there cases where it would be difficult to detect a feedback loop?

crogers: not a problem, seeing our implementation

… but I think Marcus was worried about question of minimum delay in samples or seconds

<joe> q

Proposed resolution: no exception if delay too low in feedback loop, clamp silently to minimum

crogers: the problem is solved if we define the minimum in terms of sample frames

… uniformly 128 sample frames

Olivier: seem to recall the indecision was because this would induce differences between devices with different sample rates

crogers: probably won't get to a conclusion without Marcus and Robert in the discussion
... setting the limit in sample frames will make implementations easier
... you will hear differences when you are running at different sample rates anyway

(joe, will ack you in a sec, write your question on IRC?)

Olivier: are there any precedent?

crogers: think most such systems do it by the sample block size

Olivier: would either make it easier to test?

<joe> (ot, I am on the call live now)

(joe, oh!)

crogers: not a problem if tests are run in a consistent environment

joe: perhaps we could incorporate caveat on ability to perform certain tasks with feedback loops
... IOW, leave support for feedback loop out

crogers: seems important to have though, fairly basic feature of modular systems

… ok with not talking about hard limits

crogers: I think we can define it at least in terms of sample frames

… and browser can get equivalent implementations that way

Olivier: so proposal would be to keep the clamping, but implementations can choose their minimum

joe: essentially, yes

… for v1, seems wise

Use Cases and Requirements review

Call for review -> http://lists.w3.org/Archives/Public/public-audio/2012JulSep/0665.html

any objection to publishing as a WD with view to be a Note?

crogers: sounds good to me

… important to get this out

Resolved: publish webaudio-usecases as WD

Loop start/stop points

-> Proposed: https://www.w3.org/Bugs/Public/show_bug.cgi?id=17390#c6

joe: proposal is to augment current capability with loop mode enumeration

… additional behaviour would be to loop through range specified, until the stop time

… could have additional loop modes in the future

crogers: why can't we just have a loop start and end which, if 0, will loop the whole sample

joe: was the original proposal

… would be ok with it, but this proposal is a little cleaner and extensible

crogers: wondering if we could just overload start and end and add loop mode later - would that be dangerous

… if we are able to add loop mode at a later time, would that get rid of danger?

joe: think so. may end up with having to support an odd legacy behavior

… what if today someone uses start and end to specify a loop mode

… what happens when we add the loop mode parameter

crogers: we still have the loop attribute, right?

joe: yes

Olivier: would one proposal make it easier for the generic case of full loop

joe: both solutions would be quite simple for that case

crogers: what if we remove the mode of "loop entire sample"

… that's specified by having the parameters being the whole thing

joe: in which case we don't need the loop mode

crogers: yes, we could add loop modes (start/end and backwards) later
... which is why I suggest having loop start and end be 0

<jussi> maybe use infinite?

crogers: what if the loop end is beyond the size of the sample

joe: I have clamping in proposal

crogers: what if the new buffer is shorter

joe: if you set start and end points that are invalid, clamped values are used in the playback

… this clamping does not affect visible/readable values of loop start and end

crogers: edge case if both are invalid, what would it sound like
... like starting simple

jussi: what if the loop end is at infinity? would that be better than 0?

crogers: unsigned long, so there is no infinity

… brings up next question of looping at sub-sample values

crogers: wondering if it's worth adding

joe: not sure it's worth it

crogers: probably not worth if the likes of soundfont do not do it
... it is possible, but not too fond of it

joe: would favor going to a float now if we suspect we may have sub-sample looping

crogers: then should we go with infinity?

<jussi> nope

joe: would prefer 0 and 0 myself

crogers: do you feel strongly about infinity

jussi: no

Resolution: make loop start and end be floats, and 0 and 0 are default

crogers: rounded if fractional?

joe: yes

Thierry: the charter has been updated with small suggestions

… should be vetted soon

Thierry: registration for TPAC is ongoing, you have a few more weeks

<jussi> Cheers!

This is scribe.perl Revision: 1.136  of Date: 2011/05/12 12:01:43  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Succeeded: s/valies/values/
No ScribeNick specified.  Guessing ScribeNick: ot
Inferring Scribes: ot
Default Present: +1.510.334.aaaa, ot, tross, jussi, gcardoso, +1.650.214.aabb, ChrisWilson, crogers, +1.617.600.aacc, joe
Present: +1.510.334.aaaa ot tross jussi gcardoso +1.650.214.aabb ChrisWilson crogers +1.617.600.aacc joe
Agenda: http://lists.w3.org/Archives/Public/public-audio/2012JulSep/0740.html
Found Date: 12 Sep 2012
Guessing minutes URL: http://www.w3.org/2012/09/12-audio-minutes.html
People with action items: ot

WARNING: Input appears to use implicit continuation lines.
You may need the "-implicitContinuations" option.