Minutes from PEWG meeting 14 November 2025 (TPAC)

Dear all,

apologies, only just realised that I hadn't posted the minutes (and 
since RRSAgent refused to turn up in the channel, for some reason, these 
are purely a manual dump straight from IRC):

PEWG
14 November

Agenda: 
https://www.w3.org/events/meetings/19833134-cd0a-426a-8c78-a086c27c7a29/

Attendees:
Patrick_H_Lauke, smaug, various observers

Chair: Patrick H. Lauke
Scribe: Patrick H. Lauke

Patrick: quick intro, group has been operating for a few years. last 
stable spec is PE Level 2. we've been focusing on PE Level 3. the spec 
has been submitted for REC track, we're currently working on making sure 
WPTs are in order
Patrick_H_Lauke
Patrick: we're on a charter extension at the moment 
https://www.w3.org/2023/11/wg-pointer-events-charter.html
ananya has joined (~ananya@6b13c2d6.publics.cloak)
ananya
present+
Euclid
present+
Patrick_H_Lauke
Patrick: as mentioned, we submitted a tweak to our charter that is 
currently being discussed/evaluated 
https://github.com/w3c/charter-drafts/pull/715
smaug
present+
ananya has left IRC ()
Patrick_H_Lauke
Patrick: short overview of the rationale behind our recharter, including 
the relationship between PE and UI events (spec is currently abandoned, 
there are lots of areas where PE needs to reference aspects of UI 
events, we're now looking at incorporating them directly in PE - not a 
power grab, but since UI events is currently not maintained, and we rely 
on things being defined algorithmically e.g. for "retargeting
Patrick_H_Lauke
events when a node disappears")
Patrick_H_Lauke
Patrick: also explanation of gesture stuff - we're not trying to define 
the low level "how do UAs identify/decide what is a gesture", we're more 
interested in sketching out some basic high level API for authors to be 
able to detect things like "a swipe left/right" without having to code 
these from basic principles. similar to something like the old hammer.js 
library
kagami has joined (~kagami@6b13c2d6.publics.cloak)
Patrick_H_Lauke
Patrick: we're a small nimble group, but always happy for more 
participation. workload is fairly low atm, though may ramp up a bit with 
new areas we want to look at. Our repo https://github.com/w3c/pointerevents
smaug
https://github.com/w3c/pointerevents/issues/559
mmocny has joined (~mmocny@6b13c2d6.publics.cloak)
Patrick_H_Lauke
Patrick: was hoping to have mustaq here for more clarification, but...
Patrick_H_Lauke


# TOPIC: Web Platform Tests - making a list of which tests need to be 
reviewed/revised

kagami has left IRC ()
kagami has joined (~kagami@6b13c2d6.publics.cloak)
Patrick_H_Lauke
Patrick: we still have a lot of tests that are red. partly this is due 
to tests that relate to features that we've now removed (directional 
touch-actions, which we've moved out of PE3 again and deferred to PE4 
due to lack of support beyond Chrome), and some issues that appear to be 
down to test harness itself
Patrick_H_Lauke
Olli: yes, from Firefox's side, anything to do with "pen" pointer has 
issues / fails
Patrick_H_Lauke
Patrick: was hoping to have mustaq here with us today, as he's been 
doing sterling work on compiling our list
Patrick_H_Lauke
Patrick: at last meeting, we discussed possibility of tagging tests 
somehow, so we can get a subset that is relevant just for PE3 to 
facilitate the path to REC
Patrick_H_Lauke
Patrick: so what's best way forward? should we try and go through things 
here now?
Patrick_H_Lauke
Olli: possibly better to do offline first. If we remove/exclude things 
related to pen, remove the tests related to touch-action values we 
removed, remove the ones that appear to be due to broken harness...
Patrick_H_Lauke
Patrick: ...and if i recall, some tests like pointerlock are in here, 
but they don't actually test anything from PE, just "it happens to use 
pointer events, but testing aspects that aren't"
Patrick_H_Lauke
Q from observer: so what is the plan for making separate list of tests? 
folders? marking as tentative?
Patrick_H_Lauke
Olli: we're looking at a few options, hoping to see what other groups 
are doing
Patrick_H_Lauke
Patrick: "tentative" was discussed in our last meeting, but it's perhaps 
not quite right - it's reserved for things that don't have a spec yet, 
whereas we want to use if for "these are for level 4". doesn't appear to 
be a concept of "this test only for PE 1, 2, 3, but not 4"
Patrick_H_Lauke
Olli: some work may need to be done with filenames, but then we have 
problem of breaking references
smaug
Some of the tests are part of interop-2025, and 
https://wpt.fyi/results/pointerevents?label=master&label=experimental&aligned&view=interop&q=label%3Ainterop-2023-events 
looks rather green. (Yes, that is 2025, even though it says 2023)
Patrick_H_Lauke
ACTION: for next meeting work on a plan forward to tackle list of WPTs - 
excluding ones that are PE4, check which ones are down to harness 
problems, etc
kagami has left IRC ()
Patrick_H_Lauke
Patrick: not sure when Mike Smith (tm) is planning to pop in, but he 
said he had some information regarding UI events way forward. might see 
him in coffee break, so let's leave that item until then
Patrick_H_Lauke


# TOPIC: touch-action logical values (including directional, but worst 
case just existing PE3 ones)

Brandel has left IRC ()
Patrick_H_Lauke
Patrick: did some early work on this ages ago, got a branch, but likely 
conflicting by now
kagami has joined (~kagami@6b13c2d6.publics.cloak)
Patrick_H_Lauke
Patrick: relates to https://github.com/w3c/pointerevents/issues/505 and 
https://github.com/w3c/pointerevents/issues/272
Patrick_H_Lauke
Patrick: context is that PE contains touch-action values, these have 
been unchanged since what became PE Level 1 at the time microsoft 
submitted this as spec
Patrick_H_Lauke
Patrick: includes things like pan-x/pan-y. but those may not make 
sense/aren't flexible enough when working in non-LTR writing systems
Patrick_H_Lauke
Patrick: even though i'm not a super expert, i took a naive approach of 
looking at how some other CSS specs used logical values, rather than 
absolute directional values, and what that could look like in PE
adekker has joined (~uid727463@6c65f1b9.public.cloak)
Patrick_H_Lauke
Patrick: not *replacing* the existing values, but complementing them 
with something logical that doesn't relate to direction, but rather to 
column/row (which changes dynamically based on writing system)
Patrick_H_Lauke
Patrick: draft PR for it was here 
https://github.com/w3c/pointerevents/pull/496
smaug
https://www.w3.org/2025/04/10-i18n-minutes.html#215e
Patrick_H_Lauke
Olli: question if this is an actual real-world problem - what are the 
use cases - or if it's more a case of spec "purity"
Patrick_H_Lauke
Patrick: I could envisage it making sense even from a practical 
application. for instance, on touch device / mail client, swipe left on 
an email to get more options ... when it's LTR. when it's RTL, may make 
more sense to have the swipe the other way to the right. or say you want 
to do something "along the axis of the text", being able to define it so 
it works both for horizontal laid out text and vertical text s
Patrick_H_Lauke
ystems
kagami has left IRC ()
Patrick_H_Lauke
Patrick: from memory, when we last discussed this in our group in Jan 
2024, Rob (Flack) had some concern about the direction-specific ones
Patrick_H_Lauke
Patrick: as it wasn't clear if the direction (e.g. pan-block-start) 
meant it's a pan TOWARDS the block start or AWAY FROM the block start
Patrick_H_Lauke
Patrick: but arguably the same problem already exists with touch-action 
direction values ... like pan-left/pan-right/pan-up/pan-down ... does it 
mean panning TO or FROM those
Patrick_H_Lauke
Patrick: problem was that i took inspiration form CSS specs for 
margin/padding/etc, and those are...static, they only refer to the position
Patrick_H_Lauke
Patrick: perhaps next sensible action would be to flesh out this draft 
PR some more, include actual simple diagrams (animations?) that make it 
clearer, then trying to sync up again with i18n and csswg about these
Patrick_H_Lauke
ACTION: revisit #496 to check for conflicts, and to add even rough at 
this stage diagrams/animations
Patrick_H_Lauke
Olli: we may need to then pass extra info, as panning happens in 
different process. CAN use things like (dir) to craft CSS that's 
specific to LTR/RTL already...
Patrick_H_Lauke
Patrick: ...but would be good not to require authors to rely on jumping 
through burning hoops
smaug
https://issues.chromium.org/issues/457794569
Patrick_H_Lauke


# TOPIC: bonus...a chromium bug flagged by PLH

kagami has joined (~kagami@6b13c2d6.publics.cloak)
Patrick_H_Lauke
Olli: [explains the problem - relates to touch-action: auto and how much 
of the user agent's magic - there's nothing TO pan here, so should not 
swallow/digest pan events - is spec'd/left up to UAs)
Patrick_H_Lauke
ACTION: Olli and Mustaq to duel it out
Patrick_H_Lauke
Michael Moony: specifically keen on some of the gesture metrics. in 
event timing spec we have concept of interaction, interaction to next 
paint (popular core web vital). two ideas: are gestures are just new 
things, would be cool to treat these as interactions. but even for 
old-fashioned tap, many events fire ... would gesture be an umbrella for 
all these discrete events? and should we be working closer
Patrick_H_Lauke
Olli: our planned event are VERY high level, so not sure we can go much 
further
Patrick_H_Lauke
Michael Moony: i would love some way to logically group a stream of 
events to treat it as a single "interaction"
Brandel has joined (~Brandel@6b13c2d6.publics.cloak)
Patrick_H_Lauke
Michael Moony: would love gesture id
Patrick_H_Lauke
Patrick: so you'd be able to relate individual events to gestures, and 
vice versa. almost sounds a bit like what we did for coalesced events
Patrick_H_Lauke
Olli: you may not get the individual events though, if the UA consumes 
the gesture
mmocny has left IRC ()
Patrick_H_Lauke
Patrick: we're at the stage currently where everything's still up for 
grabs. I'd suggest if you have use cases or ideas, feel free to send 
them to us. even if you wanted to join one of our meetings (without 
formally joining the group), feel free to reach out, we're always happy 
to have guests
Patrick_H_Lauke
[Michael and Olli discuss some more specifics]
m-alkalbani has left IRC ()
kagami has left IRC ()
Brandel has left IRC ()
saji has joined (~saji@6b13c2d6.publics.cloak)
Steven has joined (~Steven@6b13c2d6.publics.cloak)
Patrick_H_Lauke


# TOPIC: Dan Applequist (ABQuist) to talk to us about AB initiative

m-alkalbani has joined (~malkalbani@6b13c2d6.publics.cloak)
Patrick_H_Lauke
present+ torgo
Patrick_H_Lauke
Dan: used to be TAG - the "what"; now AB - the "how", process
Patrick_H_Lauke
Dan: Process document has already been through reviews to simplify, but 
can simplify further. addressing criticism of it still being difficult 
to engage with
kagami has joined (~kagami@6b13c2d6.publics.cloak)
Patrick_H_Lauke


# TOPIC: UI Events and way forward to integrate at least parts of it 
into PE4

Patrick_H_Lauke
Baran (meta): question about pointer events for AR/XR devices
Patrick_H_Lauke
Baran: we have two pointer devices that support hover AND multi-pointer 
events
Patrick_H_Lauke
Baran: we use pointerType=touch, but can cause confusion when 
dispatching certain events and websites break
Patrick_H_Lauke
Baran: what are thoughts on allowing direct manupulation for non-touch. 
either for mouse, or even another new pointerType?
Patrick_H_Lauke
Patrick: touch-action has always been a misnomer. we do think that 
direct manipulation should be possible with any pointerType. but tricky 
for compat...particularly if sites don't expect it
Patrick_H_Lauke
Patrick: problem with inventing new pointerType is that a site may not 
recognise it, if they're coded to listen to specific pointerTypes only
Patrick_H_Lauke
Patrick: [summarising a bit - discussing with Baran how the problem with 
things like "how can we support multiple pointers that hover" is if you 
still listen to legacy mouse events, which are explicitly spec'd to only 
work on the primary. but if you abandon those and move to just listen to 
pointer* events, then it's up to YOU as author how you deal with 
multiple pointers]
Patrick_H_Lauke
Patrick: [this is actually when you then may have opposite problem ... 
you listen to pointer events and then check the primary boolean to 
determine if it's the primary pointer]
Patrick_H_Lauke
Olli: checking the inout device capabilities spec (that Baran mentioned 
as a possible workaround) 
https://developer.mozilla.org/en-US/docs/Web/API/InputDeviceCapabilities_API 
- spec last updated 4 years ago, and Chromium only
Patrick_H_Lauke
Patrick: i think that came about before pointer events was even a thing, 
i remember looking at it in my "getting touchy" presentation as part of 
the "how to beat the 300ms delay between touchup and click" to de-dupe 
events firing
Patrick_H_Lauke
https://wicg.github.io/input-device-capabilities/
Patrick_H_Lauke
Olli: i'll see if Mike (tm) Smith is around. we did discuss UI events 
spec on Monday already, and in another meeting...we also need mustaq, so 
we can probably leave this topic for another time
Patrick_H_Lauke
Patrick: so just to mention that, as discussed, I did create a new 
branch for PE+UI events stuff 
https://github.com/w3c/pointerevents/tree/pointerevents%2Bui-events
Patrick_H_Lauke
Patrick: based on main/PE4
Patrick_H_Lauke
Patrick: main reason i made it a separate branch is that this gives us 
freedom to experiment and start putting bits in place that may be a bit 
rough initially, without the need to always make sure it makes sense 
(that's what PE4/main is for)
Patrick_H_Lauke
Patrick: once we're happy with the shape of PE+UI, we can then merge 
that back into PE4/main
Muadh Al Kalbani: suggestion to include examples of how to use pointer 
events correctly/well to deal with multi-pointer cases - e.g. example of 
"basic multipointer drawing application"


-- 
Patrick H. Lauke

* https://www.splintered.co.uk/
* https://github.com/patrickhlauke
* https://flickr.com/photos/redux/
* https://mastodon.social/@patrick_h_lauke

Received on Wednesday, 3 December 2025 15:25:01 UTC