Minutes from PEWG meeting 18 November 2026

Dear all,

in a break from tradition (to go with the break in my arm), the minutes 
this time around were done by zoom transcript. I tried to then paste 
these into irc, but I think that killed RRSAgent in the process... so 
only email-based minutes this time around:

present: Patrick, Rob, Mustaq, Olli

So, I think we only had one topic, uh, and let me just find it in the thing…
I think it's this one here.
So, I think this was filed as a…
Very much a placeholder kind of issue by Ollie.
Olli Pettay 15:07:29
Yeah, after we had that one meeting about this, that it's unclear…
Patrick H. Lauke 15:07:33
Yeah.
And the idea, or the plan, was that Ollie and I at TPAG would kind of…
corner people, uh, between sessions, and maybe have, uh, have a little 
chat with them, how they thought.
We should be looking at, uh, the, uh…
How to deal with mouse, basically, and initializing mouse events, and 
things like that, but obviously that didn't happen.
So, that is probably how to basically work the, uh, the mouse algorithm 
part, uh, back into the spec in a more sensible way, is kind of our next 
big push that we need to do, uh, on…
the specification.
Uh, so, it's probably… yeah, it depends how we want to tackle this, uh, 
in terms of…
Do we want to just start writing things during one of our meetings, like 
today, or…
does somebody want to make a start? Is there a… some prior art that we 
can have a look at, where you think, oh, that spec did a really good job 
of defining the…
Initialization for this type of event, let's use the same kind of thing.
And it… I mean, to me, it's always, uh, you know, the blank page kind of 
thing, uh, is always problematic. Once you've got something, it's a lot 
easier to…
refine and, and start kind of putting bone, uh, meat on the bones. 
Bone-related jokes, always good. Uh…
But yeah, so probably an open question now to everybody. How do we want 
to kind of tackle this initially?
Olli Pettay 15:09:24
I think there aren't any other cases which are quite as complicated 
because here we have pointer events and mouse events and you have 
multiple mouse events. I mean, you have mouse move, which is causing 
boundary events and so on.
It's very complicated, but I think it should be one algorithm somehow 
defining that all.
flackr 15:09:46
Well, the boundary events isn't part of initialize the mouse event, though.
Olli Pettay 15:09:49
No, no, no, but I mean, the overall, how events work, it's the algorithm 
that, okay, you dispatch, let's say, there's a move, user moves pointer, 
that triggers.
Pointer move, mouse move, those trigger boundary events, and all those 
events need to be initialized somehow.
flackr 15:10:02
Right, right.
Olli Pettay 15:10:09
And it's… it…
Patrick H. Lauke 15:10:09
Yeah, and the stuff we had before that we removed, which was maybe sends 
this event, maybe sends that event.
Doppelgänger (ca-wat-brt1) 15:10:19
So, my question is, like.
By algorithm, you don't… do you mean just that…
Patrick H. Lauke 15:10:27
Bye.
Doppelgänger (ca-wat-brt1) 15:10:28
The event init structure, you just populate it and…
construct the event, or there is a method called East Mouse Event.
Like, do we still want that?
Like, in, in, in Mao Civil…
Olli Pettay 15:10:41
No, yeah, we don't want that, and I think initialization… yeah, Rob is 
right that, yeah, if we just… if we are just talking about the 
initialization, then that should be…
too complicated. It's like…
I would try to use the the init dictionary, basically, if possible. You 
initialize that, and then you construct using that.
flackr 15:10:57
Yep.
Yeah, I think so.
Olli Pettay 15:11:06
It's not the same as what JavaScript would do exactly, because 
JavaScript can't…
trigger Trusted Events, but it's almost similar, something like that, or 
Intel.
Doppelgänger (ca-wat-brt1) 15:11:16
Okay, so my question is even more basic than that, right? So, sorry.
So the inner dictionary thing, right?
isn't it implied somewhere that, okay, every event you construct, every 
unit dictionary you populate it and construct the event, right? Why do 
we need an algorithm for that, that part?
Olli Pettay 15:11:36
Oh, but we need to say that what are those values that we passed to the 
init dictionary?
Doppelgänger (ca-wat-brt1) 15:11:44
So, when the event comes from the low level, what values we see set in 
the unit dictionary?
so that you can consume the event, right? That's the idea.
Olli Pettay 15:11:53
Yeah, I think so.
And then that initialization algorithm would be called from this bigger…
What happens when user moves mouse mouse or clicks or something?
I think DOMSpec is a bit… DOMSpec has a way to initialize events, but 
it's also a bit vague there. It's like, go through the dictionary.
Wherever it is.
Doppelgänger (ca-wat-brt1) 15:12:29
So, we have… I'm pasting it… we have this initMouseEvent method in 
MouseEventSpec.
Right.
So…
I guess this is…
We'll keep this, or we'll just build on top of this?
Olli Pettay 15:12:57
Give me the legacy in it, Marcia.
Doppelgänger (ca-wat-brt1) 15:13:00
Yeah.
Like, I always had the thought that this is not needed because this is 
implied. That's why I'm asking the question, right? Like, it is essentially…
Olli Pettay 15:13:02
No, I, I was…
And I I I I'm thinking about the any dictionaries which you when you 
contract.
flackr 15:13:11
Yeah, I think we want to use a dictionary, and we want to use a 
dictionary that's like…
because a Pointer event is a mouse event, the dictionary should have 
both the Pointer event in it and the mouse event in it. Sorry. And then 
it can do the initialization for the Pointer event and the 
initialization for the mouse event from the same init dictionary.
Doppelgänger (ca-wat-brt1) 15:13:24
Sorry.
flackr 15:13:35
Something like that.
Doppelgänger (ca-wat-brt1) 15:13:44
So essentially we need a.
We need…
The initialization of the point driven in it.
That should cover everything because Pointer Raven is bigger one, right?
that has… part of it is mouse availability, so how do we initialize 
PointerEvent init.
Based on some other input, like, say, low level input or whatever, 
right? Isn't…
flackr 15:14:11
Yeah, there'll be a, like, when the mouse moves…
Doppelgänger (ca-wat-brt1) 15:14:11
Oh.
flackr 15:14:15
initialize a Pointer Event with… and then, you know, the various fields 
of Pointer Event in it that are set.
And the initialize a pointer event will internally call initialize a 
mouse event.
Something like that.
Olli Pettay 15:14:46
I think that is what, I pass it to IRC, but I think that DomSpec is 
doing that.
And there is also the event fire. I wonder, should we just use that? 
It's kind of…
On top of that other algorithm in DOM spec.
Oh, but it doesn't take it in it.
Doppelgänger (ca-wat-brt1) 15:16:29
So, I think the DOM event fire happens much later, right? We don't need 
to worry about it.
We… this assumes that we have the event constructed, and say, for 
example, one mouse move.
Called the creation of multiple over entire.
Pointer mouse events, like, every event is constructed, and after that, 
the events fire one by one.
So this firing is talking about that one.
No?
Olli Pettay 15:16:59
Yeah, it could be, uh…
Doppelgänger (ca-wat-brt1) 15:17:04
Because it is taking as a parameter something event E, right?
I presume there's an event.
I don't know.
Olli Pettay 15:17:17
If I run, I can never do it.
This might be something where asking Anas' feedback might be good, that 
what he… because, yeah, maybe we want to tweak DOM spec a bit here, so 
that it's easier for us to…
Maybe. Or…
Patrick H. Lauke 15:18:04
Yeah, because I'm just looking on section 32, we've got those 321, 322, 
323, which are very, very high level.
Initializing a Pointer Event, creating a Pointer Event.
Placing point event from mouse event.
So, uh… are we thinking about basically expanding that, and then making 
it also…
Uh, basically work for mouse events, which are…
Pointer events now, by definition.
And kind of maybe splitting that a little bit, saying, you know, here's 
how you initialize a pointer event. If it's just a mouse event, do this bit.
Well, there shouldn't actually be any difference, really, should there?
Well, apart from some of the values, if it's a pure mouse event, then 
some of the values will be null.
Like the, uh…
uh… what's it called? You know, the various angle things, and I think…
pressure and stuff. I can't remember if that, by definition, those 
basically don't even exist, or if they exist if they're null.
Well, basically, some kind of handling of that.
But then generalizing it into something that works for both.
Doppelgänger (ca-wat-brt1) 15:19:28
So, are you talking about Section 3.2? Like, we have… okay.
Patrick H. Lauke 15:19:31
Yeah.
And then I'm wondering if for the… for the legacy, uh, initializer that 
we have in the spec for mouse event, if that basically just makes it…
It makes it redundant, because, uh, the…
Well, we don't have init pointer event, do we? So, we still keep init 
mouse event, but make it basically like a, you know, when this happens.
Doppelgänger (ca-wat-brt1) 15:19:58
Yeah.
Patrick H. Lauke 15:20:04
it goes through the same steps as the, you know, in the 3.2 section, how 
to initialize a Pointer Event.
with any…
So it's almost like a mapping of… if you do this, it's basically the 
same as instantiating a…
You know, calling the constructor for a Pointer Event kind of thing.
This is what I'm imagining. Correct me if I'm wrong.
Doppelgänger (ca-wat-brt1) 15:20:31
No, I think my impression is that we don't need 600 3.2, like, this is 
just, again…
saying how… just something we know, right? Except that the last one, 3.2.3.
Why it is there? Create Pointer Event from Mouse Event. Maybe this is…
This was added with the impression that low-level events will be the.
General Pointer Event, then out of that will be 4 Mouse Event to 5 Mouse 
Event.
I don't know why we'll create…
Sorry, create Pointer Event for Mouse Event, from Mouse Event.
Patrick H. Lauke 15:21:08
Yeah, that sounds a bit, uh…
Doppelgänger (ca-wat-brt1) 15:21:10
Opposite, right?
Patrick H. Lauke 15:21:11
Yeah.
Doppelgänger (ca-wat-brt1) 15:21:12
You don't have any of the fields.
Patrick H. Lauke 15:21:13
Minute.
I mean, if anything, we'd have a thing that says, basically, you know, 
assume that you're starting with Pointer Event, and then saying, at this 
point, you know, after you fire this Pointer Event, you may also fire 
the legacy mouse event.
Doppelgänger (ca-wat-brt1) 15:21:30
Yeah.
Patrick H. Lauke 15:21:32
Uh, and basically have that in the algorithm, uh, and, you know, that's 
where you then clarify, okay, after the pointer move happens, you do 
the… you can also send a mouse move for compatibility reasons.
Here's what you need to do specifically with the target in that 
particular case, and…
here's where the quirks might be compared to, uh, you know, the original 
Pointer Event was doing this with a boundary, but for the Mouse Event, 
uh, it needs to do this thing.
So it's kind of, yeah, smooshing it all together into one single 
decision tree of what algorithm, what the user agent is supposed to do. 
Start off with the pointer events and here's the bits where it then 
needs to send the legacy events.
Because I think, in general, that would then also mean that… I think up 
to now, we've kind of had the mouse events a bit, like, at arm's length 
in our description, saying, well, here's also the…
the suggested, uh, order that you then get, you know, the interleaved 
Pointer Events, Mouse Events.
Ah.
and kind of left it up to, you know, we don't say in this spec what the 
exact order should be, and it's up to UAs to do things.
Doppelgänger (ca-wat-brt1) 15:22:55
Yeah. Okay.
Patrick H. Lauke 15:23:01
But by specifying it more in the algorithm, we'll probably tighten that 
and come up with something that is much closer to a, well, here's what 
you really should have as the order, because if you follow the algorithm.
That's what you'd get.
Uh, so we're… we're kind of…
specifying further and clearer what the situation will be, rather than 
keeping mouse as a hand-waved, oh yeah, yeah, and then the UA will do 
something relating to mouse, uh, kind of thing.
Which is good. I mean, that then probably opens up various new ways of 
testing things, some more…
Uh, you know, web platform tests…
for more consistency, for the greater good. So it's good, but yeah, it's 
definitely a big, chunky lift. It's not just a small little, let's just 
do something, but we probably need to start somewhere.
So I'm wondering, does somebody want to make the first move a 
disadvantage of just start a branch and…
See, you know, what could be done, make a start, or…
Do we want to canvas for opinions first?
Doppelgänger (ca-wat-brt1) 15:24:23
Yeah, like, so something I wanted to mention before I followed up a few 
minutes before. It's not very related to what you mentioned. One of the 
things that we are missing in this spec.
I think, is that…
We presume that low-level events, low-level events are.
Like, at least in Blink, right? Are only…
in, uh, sorry, mouse down, mouse up, mouse move. Like, moved up, down, 
and move.
And from those, we can kind of…
In Blink, we make up these over and out live events.
Right? I mean, capturing even, like, just the basic ones, right? So 
then, the… whatever construction algorithm we have.
for…
We'll have one algorithm that says, okay, given the low-level event, how 
to construct.
See this mouse move down up, pointer move down up events. And another 
algorithm that says, OK, how we construct over and over the implied 
events, let's say, from those low-level — maybe not the low-level event.
from the existing Pointer Revenue already constructed.
Like, at least how that works in Link. Again, the low-level events make 
the down, move, and up events.
And from those, we derive the other events over on interleave out 
events. So I see two kinds of constraints here.
First, they're defining that, okay, this, we expect only these global 
events.
And from that, we construct these core events.
And then the second algorithm that says we construct the secondary event 
from this primary event, something like that.
Olli Pettay 15:26:10
Yeah, yeah, I agree. And then there is this hopefully simple thing, 
which is only this initialization of one kind of event part, which is 
maybe we should…
We can't.
Doppelgänger (ca-wat-brt1) 15:26:21
Probably, yeah.
Olli Pettay 15:26:23
Start from there. I wonder if I could just…
try out something here, basically, ask Claude, can you create me 
something here? Because, yeah, it's like playing a bit with what's the 
right structure, specular.
Patrick H. Lauke 15:26:34
Oh.
Doppelgänger (ca-wat-brt1) 15:26:40
Oh, okay, okay. One other point. Sorry. This is related to what I said, 
right? I claim that in…
down, move, and up are core event, and every… others are secondary 
event. There's one exception that when the mouse moves out of the 
window, see, you alt-tab, you send the leave event directly. So, yeah, 
this is… this is…
Olli Pettay 15:26:59
Yep, yep.
Yeah, it's the same in GetCall, yeah, yeah.
Doppelgänger (ca-wat-brt1) 15:27:05
Okay.
Olli Pettay 15:27:10
But, yeah, but I think, yeah, if you have a good way, a good algorithm 
to initialize events, I mean, the DOM events, then we can, on top of 
using that, we can then have these other algorithms.
for actually handling them.
low level events, how to convert basically the OS level events to DOM 
events.
That's one part, and then there is this how to create the other DOM 
events, I mean, the boundary events and so on, based on the low-level 
events.
That's roughly what we need in this spec.
Yeah, but I… different topic, a bit different topic. I was wondering, in 
mouse events versus pointer events, isn't there something… some 
difference? Wasn't something about button… or buttons either one of 
those, which we… which is a tiny bit different, so we need to…
Doppelgänger (ca-wat-brt1) 15:28:05
Yes.
Yes.
Olli Pettay 15:28:09
Claire, that needs to be a special case in the initialization algorithm, 
that, okay, now we are dealing with mouse events, do this, otherwise do 
that.
Doppelgänger (ca-wat-brt1) 15:28:12
Mm-hmm.
Olli Pettay 15:28:17
But that's hopefully so much simple thing.
But I can't remember if there are other…
That's kind of a special case.
Yeah, let me try to get some kind of, uh, whip.
patch for this. It's like just…
Maybe it's easier to go to group.
Patrick H. Lauke 15:28:58
Yeah, just to get rid of the blank page, kind of, we could do it this 
way or that way, or… just do one thing, it might be a throwaway, that as 
soon as we look at it closer, we say, oh, actually, that makes it…
Olli Pettay 15:29:04
Yeah, yeah, just an idea. That's what we'll maybe do.
Yeah.
Patrick H. Lauke 15:29:12
More complicated, let's do this instead, but yeah.
If you could, that would be amazing.
Olli Pettay 15:29:20
Yeah, I'll try to do it this week.
Patrick H. Lauke 15:29:22
I will do an action here over in IRC for that.
Olli Pettay 15:30:00
I wonder if we are missing something.
Oh.
All right. Some of the attributes there are defined in some CSS spec, 
whichever is the CSS OM, maybe.
This is.
Patrick H. Lauke 15:30:18
Yeah, I think it's OM.
Olli Pettay 15:31:05
It has its own view module. Yeah.
Patrick H. Lauke 15:31:28
But yeah, I would say then, uh, if…
you're happy to do that, Ollie, that would be amazing. Uh, is there 
anything else you need from others? Opinions at this point? Or…
Olli Pettay 15:32:04
Yeah, yeah, yeah.
Patrick H. Lauke 15:32:10
Are the others happy to just sit back and have a look, and then once 
we've got something, we can…
more specifically say, oh yeah, I like this bit, or how about this bit, 
and then we've got something we can collaboratively work on.
That's a bit more stable.
Cool.
Olli Pettay 15:32:32
Yeah, and in the spec level, I mean, it might be somewhat garbage. It's 
just that having something there, so that's… Because, yeah, I need to…
Patrick H. Lauke 15:32:39
Yeah.
I mean, that's the thing, otherwise we could be here and kind of build 
castles in the sky.
all day long about what it should really be, uh, but, you know, once we… 
once we settle on something, we can always expand it and move it about.
Olli Pettay 15:32:55
Yeah, and yeah, I'll need to take a look at that CSS spec too, that how…
Because it adds page X and Y and stuff like that, and it…
I'm not sure it's that…
If that is very well defined, it's true.
Patrick H. Lauke 15:33:13
Knowing specs, probably not, yeah.
I could imagine it says something about whatever low level you get the 
coordinates and then look at how far the page has already scrolled, add 
that to it, and you basically get the actual coordinate.
On the page itself, as it could be scrolled any manner of way, but that 
gives you the absolute within the actual entire page versus what's on 
the screen at the moment kind of thing.
You'd hope that that's what it says.
Olli Pettay 15:33:58
Yeah, maybe. Yeah, maybe it's fine. Actually, yeah, I need to take a look.
Patrick H. Lauke 15:34:04
And if it does explain it in a reasonable way, I mean, it would be good 
to just defer to that, rather than reinventing the language in our spec 
for that.
And no, we're not taking over CSESOM view module in our spec.
Olli Pettay 15:34:29
Well, I wonder if we should actually take the mouse event part of that. 
I don't know why.
That needs to be there. I'm not sure. Maybe Rob has his opinion on that.
Patrick H. Lauke 15:34:43
your Finnish world domination plan coming to fruition.
Okay.
flackr 15:34:52
Um, I think it's defining…
it as double instead of…
Int.
It doesn't need to be there, like, it probably should be in Pointer 
Event, but that was…
Olli Pettay 15:35:06
Yeah, I think, yeah, maybe it's just legacy thing that it's there.
Patrick H. Lauke 15:35:10
Cool.
Doppelgänger (ca-wat-brt1) 15:35:11
Sorry, what is… I missed the discussion about dollars. It is about the 
couriers.
Olli Pettay 15:35:11
I know.
flackr 15:35:16
The coordinates in the CSSOM view mouse event are upgraded to double 
instead of int.
Doppelgänger (ca-wat-brt1) 15:35:23
Yes, it was upgraded a while ago, and we tried to ship.
Mouse events, it double, and it failed. Just click. We couldn't click.
Point click was a mouse event. We tried to switch the codecs to double. 
It didn't work.
So then the way…
Chrome works that everything is double by definition in ideal, but we.
Anything mouse event?
we truncate the.
Corea 28.
Anything, by the way, we don't truncate, except click, which we still 
truncate.
Olli Pettay 15:35:57
Yeah. Yeah. I think Gecko is doing the same. Yeah. But I was just 
wondering the overall thing that why CSSpec needs to have a partial 
interface mouse event.
flackr 15:36:07
Yeah, that was just like a place to put the upgrade to a double.
Olli Pettay 15:36:12
Yeah, yeah, yeah. And given what my UI event used to look like.
flackr 15:36:12
I don't know.
Olli Pettay 15:36:18
Yes, yeah.
Patrick H. Lauke 15:36:19
Yeah, it's kind of they did what we did, kind of, you know, a separate 
spec to build on top of it kind of thing.
Doppelgänger (ca-wat-brt1) 15:36:28
So, is this something we can independently look at? Like, say, for 
example, let's… we can have the doubles defined in the parameter spec 
and drop the CSM view spec for doubles?
flackr 15:36:38
Yeah, yeah, we should do that.
Doppelgänger (ca-wat-brt1) 15:36:42
I can…
Take a stab at it.
Patrick H. Lauke 15:36:45
That'll be cool, yeah, if we divide and conquer the bits.
Doppelgänger (ca-wat-brt1) 15:36:46
should be delivered.
That should be different from…
Patrick H. Lauke 15:36:51
So, I will… I will bless that with a ring, and make it official with an 
action.
Doppelgänger (ca-wat-brt1) 15:36:58
Maybe we don't have any issue for that in Pointer of Inspect, or maybe. So…
Patrick H. Lauke 15:37:05
Yeah, maybe even just a placeholder issue, just to basically, you know, 
a one-liner that says, look at, you know, moving the definition.
Doppelgänger (ca-wat-brt1) 15:37:09
Yeah.
No.
Patrick H. Lauke 15:37:14
from CS.OM to Pointer Events.
And once we've done that, we could, uh, you know…
ping somebody in, obviously, CSA's working group and say, hey, we've 
done that.
Can you remove it, or cross-reference it, or however they want to do it?
Olli Pettay 15:38:08
Or we could ping Rob.
Shoot.
Patrick H. Lauke 15:38:12
Oh, sorry.
Well, there we go. Even better.
You can have Rob with different hats on and he can do like a Gollum and 
thing like talk to camera from different angles.
My precious, my precious doubles.
Cool, so I think we've given ourselves a giant set of tasks. Anything 
else we realistically can or need to discuss?
In the meeting today, or…
Can we call it a bit early? However, people want to…
tackle this.
flackr 15:39:10
I'm happy to call it a bit early this week.
Patrick H. Lauke 15:39:13
Yeah, cool.
So yeah, I would say I'll try to actually send reminders like weekly now 
and not like send something the night before and then we all go, I said 
I would do, you know, I said I would rewrite the entire spec. Oh no, let 
me just do it five minutes before the thing.
But yeah, at least once we've got something, then we can, bit by bit, 
hopefully iteratively work on that.
in parallel, similar to what we did with the initial… here's what a 
merged spec would look like. That gave us something to… to kind of work 
towards, until we're happy to…
merge that, make that main, and uh… carry on from there. But no, that's… 
that's cool. So, thank you all. I think what I will do this time around 
is I'll just grab the transcript here from Zoom, and very brutally try 
and paste it into IRC.
it probably won't like it, because it always truncates my lines, so I'll 
see what happens. So, this week's minutes might look a bit funky, but, 
uh, going forward, I'll work out.


ACTION: Olli to set up a WIP branch to start exploring a better 
algorithm/initialisation for both pointer and mouse

ACTION: mustaq to look at moving the short mouse event interface 
extension that upgrades values to double from CSSOM to PE


-- 
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, 18 November 2026 16:06:45 UTC