Pointer Events WG Voice Conference -- 26 Mar 2013

<ArtB> ScribeNick: ArtB

<scribe> Scribe: Art

Getting started

AB: I posted a draft agenda a few days ago http://lists.w3.org/Archives/Public/public-pointer-events/2013JanMar/0212.html. The main subject is to discuss the LC comments for which we have no recorded resolution regarding what, if anything, should be done about the comments.
... any change requests?

JR: + extensions to Element interface

AB: OK
... would someone please agree to scribe with the proviso others will help fill any gaps and make corrections?

<scribe> Scribe: Rick

<scribe> ScribeNick: rbyers

Pointer event behavior across windows

<ArtB> AB: Rick Byers submitted this comment http://lists.w3.org/Archives/Public/public-pointer-events/2013JanMar/0180.html

<ArtB> AB: it appears an IE bug was identified but not clear if the spec needs to be changed e.g. a non-normative note about PEs across windows.

RB: probably outside the scope of the spec

JR: also discussed similar issue on Dom3 events and didn't do anything there

RESOLUTION: Cross-window issues are out of scope for this spec

3D Pointers

OP: Agree, out of scope

<ArtB> AB: Bill Fisher submitted this comment http://lists.w3.org/Archives/Public/public-pointer-events/2013JanMar/0185.html

<ArtB> AB: we discussed this feature request before more generally in the context of the pointerType extensibility thread and agreed we need more experience before adding normative text.

<ArtB> AB: Jacob replied as such in http://lists.w3.org/Archives/Public/public-pointer-events/2013JanMar/0189.html. Bill's reply to Jacob, he mentions the need for a z-coordinate http://lists.w3.org/Archives/Public/public-pointer-events/2013JanMar/0200.html

JR: Intriguing discussion, tempting to just jump and support. But this is a tricky point in the specs lifetime - need to resist the temptation.
... We're probably going to want to have a long discussion, don't think it's as simple as he describes
... It will be awhile before we see implementations
... If we rush this into V1 we'll probably get something wrong

RB: Seems like if we get extensibility right, this should be successful being implementation-specific first before being standardized
... To what extent are his concerns here justified?

JR: LeapMotion has shown that prototypes are reasonable
... not beyond the realm of posibility to extend the interface with additional properties for experimentation

AB: Tend to agree that without more experimentation, this would end up blocking going to candidate
... Feels like the right thing to do is to do it in V2

SG: I don't think this is necessary in v1. Better to start close to mouse (what we have now), then once we have people on pointer we can focus more on how we handle other newer types of input.

OP: I agree

RESOLUTION: 3d pointers are out of scope for Pointer Events v1, but will consider for v2 when we have more experimentation

Last Call comments from Yandex

<ArtB> AB: Chaals' comments are in http://lists.w3.org/Archives/Public/public-pointer-events/2013JanMar/0184.html. The comments are from "people at Yandex who implemented Pointer Events for our services". (BTW, Yandex joined the PEWG a few days ago.)

AB: Yandex has joined this working group

<ArtB> AB: the e-mail raises 3-4 different issues

<ArtB> AB: Jacob's reply to Chaals/Sergey http://lists.w3.org/Archives/Public/public-pointer-events/2013JanMar/0197.html and Sergey's reply http://lists.w3.org/Archives/Public/public-pointer-events/2013JanMar/0199.html.

<jrossi2> rbyers: first issue raised is that they prefer preventDefault() over touch-action

<jrossi2> rbyers: we've talked several times

<jrossi2> rbyers: jacob commented on the reasoning

<jrossi2> rbyers: yandex asked about other browser vendors

<jrossi2> rbyers: I replied with links to a test page and G+ post about it

<jrossi2> rbyers: at this point, no way we're going to change PE to use preventDefault() model

<asir> Agree with Rick's position

AB: Agree this isn't something we should be considering changing

OP: Agree

RESOLUTION: declarative-only control over browser default action (touch-action property) will remain the only mechanism for now

<ArtB> AB: "Tilt angles are very difficult to work with, why not use standard

<ArtB> spherical coordinates?"

<asir> Gist from Jacob's response: Tilt angles. This is based on USB standard

AB: Jacob responded pointing at USB documentation

RB: Not a significant issue, good as is

RESOLUTION: keep tiltX/tiltY units as defined

AB: next issue - "pointer capture doesn't add any value, artifact of ie6"

RB: is there a compelling reason why capture needs to be in v1?
... JR: a substantial difference between these APIs and the capture APIs in IE5/6 is that these handle multi-touch
... using the capturing phase it's tricky to track specific pointers to specific elements
... could imagine slider scenario extended to mixing board with multiple sliders, each capturing a pointer
... setCapture makes this easier
... latest comment saying this makes it problematic for component author:
... don't really buy this for 2 reasons:
... 1) there are got/lost events that will let them know when this happens
... 2) there are a million other scenarios where it could be effected
... multi-touch is the primary place this is useful
... I'd like to run this API by our folks working on Web components
... any encapsulation issues here

JR: Agree we want to make sure it plays well there

RB: sounds like worst case and we decided we wanted to think about this a little longer, it's not a big deal if we wanted to pull from v1

JR: main concern is that it could affect someone's review of the spec - that it would reset things in the last call process

AB: If we decided we didn't want to include section 9 then we would have to go back to last call

<scribe> ACTION: rbyers to follow-up on web component implications of pointer capture [recorded in http://www.w3.org/2013/03/26-pointerevents-minutes.html#action01]

<trackbot> Created ACTION-28 - Follow-up on web component implications of pointer capture [on Rick Byers - due 2013-04-02].

JR: for folks wanting a more straight forward migration from touch events, this is simpler
... lets you emulate implicit capture model of touch events

RB: agree
... but not sure how important this is in practice - often implicit capture causes problems

OP: why does the API take pointerId and not a PointerEvent as it's parameter?

JR: not necessary to pass the entire object - just need to identify the pointer

OP: is the ability to capture a random number a problem?

JR: if it's a valid pointer id then it will get captured (can setcapture outside of a pointer event_
... if it's not valid then it throws an exception

AB: any other feedback? If Rick doesn't come up with any large issues with leaving it defined, then we'll leave it in.

OP: agree
... this is kind of related to pointer lock, but only designed for mouse events
... assume we'd want support for pointer events at some point

<smaug> https://dvcs.w3.org/hg/pointerlock/raw-file/default/index.html

JR: pointer lock is a lot different than just capture, it also hides the cursor, gives you deltas, etc.

OP: yes it does more, but at some point we need to think about pointer lock for pointer events

JR: agree, that's probably pointerlock v2. There's nothing in PointerEvent spec preventing a 'pointersLock"

<jrossi2> also, there are some sites already using pointer capture APIs I believe

<ArtB> AB: "Why should the mouse have pointerId == 1? There is no need for this, since we have a pointerType for detecting input device type, and it makes it impossible to use two mouse devices simultaneously."

jrossi2: If you could provide a specific example of a site using it, that might be helpful...

JR: we've talked about this before
... generally mouse is persistent
... felt it simplest to reserve a pointer id for the mouse
... multi-mouse is possible with the spec, but it's not a scenario most implementers are going after
... don't see this becoming an important thing

RB: does code special-casing 1 protect us or hurt us if we add multi-mouse in the future?

SG: seems strange to say write for multiple pointers for touch, but there can only ever be one mouse

RB: also strange to say that pointer Id is an opaque integer, don't interpret it in any way ... unless it has the value 1

JR: for multi-mouse we'd have to define which pointer fires mouse events

AB: seems like this may not be the perfect model, but meets the "I can live with it" test

RB: Alex Russel brought this up

<asir> the WG discussed on Mar 3 and Mar 12

RESOLUTION: supporting multi-mouse is out of scope for v1, will tackle in v2. The primary mouse having id 1 won't prevent this.

AB: Last comment on the thread ...

JR: Multi-touch handling and the convenience of having a touch list

RB: there's good reason to encourage more thought on the right way to do a pointer list API
... has security implications (IE originally had one and it was removed)

AB: let's add this comment to the thread

JR: have get pointer list API on v2 list

RESOLUTION: an API to return active pointers is out of scope for v1, but will be tackled in v2

Distinguishing input from multiple users

<ArtB> AB: Sangwhan Moon submitted this comment on March 24 http://lists.w3.org/Archives/Public/public-pointer-events/2013JanMar/0210.html

<ArtB> AB: first, there is a bit of procedural issue here. Since the LC comment period ended March 19 and Sangwhan's input was submitted on March 24, strictly speaking we _could_ say this isn't a LC comment.

<ArtB> AB: however, I recommend against that i.e. I think we should consider Sangwhan's comment as a LC comment.

<ArtB> AB: I say this for a couple of reasons but mainly because we have what I will call a "social contract" with the Public. We should always welcome feedback at any time in the process and then on a case-by-case basis decide what, if anything, to do about the feedback.

<ArtB> AB: any comments on the process-related aspects (separately, we will talk about the technical nature of Sangwhan's comments)?

<asir> Agree this should be a Last Call comment

JR: no objections, I have one of my own...

RB: agree

JR: want a way to associate pointers with a particular device (or 'user')
... eg. multiple users interacting with the same content [via multiple input devices]
... one example is the Wii browser
... specific proposal: add a deviceId member to PointerEvent
... is the issue clear to everyone?

AB: I think so

JR: it's not clear to me that pointer events are the only type of input you want to differentiate
... don't have anything wrong with the approach in principle
... but probably belongs on UIEvent, not on PointerEvent. Eg. to support multiple keyboards.
... Secondly, we want to make sure we're not exposing a unique identifier for users (something that persists across pages). needs to be generic, reset for each page, no guarantees that you get the same ID for each device after a navigation
... prefer this would be in the scope of the UI Events spec

OP: Fully agree

AB: should we propose to WebApps working group that this get added there?

RB: seems reasonable

JR: I can reply on the thread and see what Sangwhan thinks

<ArtB> ACTION: Barstow reply to Yandex comments (Chaals) and include link to 26-Mar-2013 minutes [recorded in http://www.w3.org/2013/03/26-pointerevents-minutes.html#action02]

<trackbot> Created ACTION-29 - Reply to Yandex comments (Chaals) and include link to 26-Mar-2013 minutes [on Arthur Barstow - due 2013-04-02].

<jrossi2> ACTION: jrossi2 to reply to Sangwhan on the list [recorded in http://www.w3.org/2013/03/26-pointerevents-minutes.html#action03]

<trackbot> Created ACTION-30 - Reply to Sangwhan on the list [on Jacob Rossi - due 2013-04-02].

<jrossi2> ACTION: jacob to reply to 3D pointer thread [recorded in http://www.w3.org/2013/03/26-pointerevents-minutes.html#action04]

<trackbot> Created ACTION-31 - Reply to 3D pointer thread [on Jacob Rossi - due 2013-04-02].

extensions to the element interface

JR: a glaring omission from the spec here
... section 6 defines extensions to the Element interface, but they should also be on Window and Document
... wish we would have caught this earlier, don't expect objections
... pretty obvious that's what the model should have been
... but think it's something we'd want to fix

<asir> sort of like a documentation issue

RB: agree, I was assuming that too

AB: consider this a bug in the IDL personally

OP: Definitely they should be on document and window too

RESOLUTION: Update spec to add all Element extensions to Document and Window as well

Testing: status, plans

AB: Matt agreed to be test facilitator, but he's not here today
... Cathy has done preliminary work on assertions

<ArtB> http://www.w3.org/wiki/PointerEvents/TestAssertions

Any other Business

AB: standing action for the group to review and contribute to these assertions

<ArtB> AB: does anyone have any new implementation status to share?

RB: no change on Google side to report this week

AB: Jacob, do you have last call tracking doc?

JR: Yes, I'll add to it for this week and will check it in
... see change for element extensions as not a substantial change, right?

AB: Yes, non-substantive
... just a bug in the IDL
... None of the changes we've discussed so far result in substantive changes
... everyone agree?

<asir> Yes

AB: We will have a call next week

RB: I also agree, no substantive changes

<jrossi2> yes

<jrossi2> no substantive changes

- DRAFT -

Pointer Events WG Voice Conference

26 Mar 2013

Attendees

Contents