Minutes from Pointer Events WG call 7 August 2019

For some reason, the RRSAgent generated minutes ... don't seem to have 
been generated (should have been at 
https://www.w3.org/2019/08/07-pointerevents-minutes.html)

A rough dump from IRC is below:

Meeting: PEWG
Chair: Patrick H. Lauke
Agenda: 
https://lists.w3.org/Archives/Public/public-pointer-events/2019JulSep/0029.html
Scribe: patrick_h_lauke
present+ patrick_h_lauke
smaug: present+ Olli_Pettay
ella: present+ ella
NavidZ_: present+ NavidZ

Topic: tiltX and tiltY aren't as helpful as just having altitude angle 
   https://github.com/w3c/pointerevents/issues/274

Navid: do we want this polyfilled, or provided by platform?
Patrick: concern is polyfilling client-side could be performance hit
Navid: browsers could internally calculate as needed/accessed
Patrick: would it make sense to have both these sets of variables? does 
it go against some kind of API philosophy?
Olli / Navid don't see any major problems
Olli: DOM API has many ways of achieving same thing, not a problem
Navid: do we need to spec what happens if an author changes just one 
like tiltX, that it needs to be recalculated?
Olli: but these are read-only values, so can't be modified
may need to specify the calculation between the two
Navid: what about constructor? do browsers need to guarantee if 
constructed that both sets are equivalent
Olli: if you pass them all, i would expect all those values to stay, but 
if author only provides one set or the other, then browser calculates 
the missing pair
Navid: what if we don't guarantee anything? either they do it properly 
and provide everything, or not. browser doesn't try to backfill
Olli: maybe that's fine. if it's not trusted events, created by webpage, 
then just use values passed in constructor. it's simple, and then up to dev
Navid: so formula we present in spec is only for trusted events
Patrick: who wants to have a stab at PR
[discussion on github process]
Patrick: I'll make an initial commit, think I have the formula for 
conversion
we'll make an initial PR with a commit and iterate over that
so that Microsoft can also comment/contribute, don't want to move ahead 
without them

Topic: Setting pointer capture on an element in a parent frame when the 
pointer was made active in a child frame 
https://github.com/w3c/pointerevents/issues/291

Navid: we had criteria that pointer needs to be active for release etc, 
but we haven't defined what activeness means
i.e. if we have iframe, and pointer goes over it, does the iframe see 
the pointer as active?
does the iframe see the pointer as active and can capture it, or not?
Olli: i'd say this is defined in spec. 10.2 is quite clear
not against behavior of blink, but it's different from what the spec says
Navid: where in 10.2?
Olli: activeness is not defined, there's no limitation similar to what 
blink does currently
blink has some limitation / custom behavior, but spec doesn't define it
Navid: correct blink goes beyond what spec says
Olli: not against adding definition to spec, but need to be clear about 
which document we're referring to
Navid: the case about passing pointer id via postmessage?
Olli: if you have same origin, you could do it directly
Navid: if you're out of process/cross origin, and somehow get ID, you 
should not be able to just pointercapture? if you allow that, an iframe 
could just randomly capture random numbers
Olli: could be useful, could be valid cases, but also possible to do bad 
things
Patrick: i seem to remember early discussions around advertising iframes 
doing nefarious things
Olli: for same origin, it could be ok...but yes blink's behavior seems 
more consistent with other things
Navid: could we explore if there are valid use cases? delegating 
permissions/capabilities to iframes?
does it have enough value?
Olli: or we can limit API (to behave like blink?)
Navid: can we change that in future though if valid use cases DO crop up?
Olli: always safer to start restrictive and then open them up more later
safe to change spec now, and then later can change spec
Navid: capture only allowed if pointer is seen as active in the current 
document
Olli: defining current owner document may be tricky
before event dispatch
maybe DOM spec has some kind of hooks how to define this stuff
there's something like a target's associated document, somewhere in the 
algorithm, that could be useful
Navid: ok we'll explore further, and i'll comment on issue
smaug: https://dom.spec.whatwg.org/#concept-event-dispatch

Topic: Changing the DOM hierarchy while handling a "pointerenter" event 
produces significantly different results across browsers 
https://github.com/w3c/pointerevents/issues/285

Navid: this seems more relevant to UI spec
we do talk a bit about how we handle mouse fallback events, but we don't 
do it for the pointer events themselves
should we write more in PE spec, or wait for UI spec to make it more 
clear, and we then piggy back on top of that
Olli: UI spec needs to be clearer
Navid: will discuss at TPAC

Topic: AOB
NavidZ_: Too many other behaviors interrupt pointer events (dragstart, 
touch-action)<https://github.com/w3c/pointerevents/issues/213>
Navid: that's done
not much else from last meeting that was progressed
would welcome further feedback though - particularly use cases etc
(the above in reference to PR Add single-directional touch action values 
https://github.com/w3c/pointerevents/pull/294 )

P
-- 
Patrick H. Lauke

www.splintered.co.uk | https://github.com/patrickhlauke
http://flickr.com/photos/redux/ | http://redux.deviantart.com
twitter: @patrick_h_lauke | skype: patrick_h_lauke

Received on Wednesday, 7 August 2019 16:19:14 UTC