08:02
<charl>
good morning
08:15
<iandevlin>
Morning
08:46
<annevk>
hsivonen: https://news.ycombinator.com/item?id=8950243
09:28
<hsivonen>
annevk: thanks.
09:30
<hsivonen>
annevk: little endian being niche that goes agains tradition is pretty much like the Web being niche
09:31
<annevk>
uhuh
12:04
<annevk>
JakeA: Domenic: how do you feel about adding Response.prototype.ok = function() { return this.status => 200 && this.status < 300 }?
12:06
<jgraham>
>=
12:09
<jgraham>
annevk: So requests has that, but I have never been convinced that the semantics are quite perfect. But I guess it's not obviously a terrible idea
12:10
<annevk>
jgraham: what problems did you run into?
12:11
<jgraham>
Well I haven't, but I can imagine situations where getting a 204 would be just as unexpected as a 100
12:12
<jgraham>
(well unless 1xx responses are handled at a lower layer I guess. Does anyone actually produce those?)
12:12
<jgraham>
Anyway I guess on balance it's a good idea
12:13
<gsnedders>
AFAIK it's only unknown status codes that should be treated as the base one
12:13
<gsnedders>
and we know that all of those in HTTP/1.1 exist
12:13
<annevk>
jgraham: I believe 1xx is ignored and not exposed
12:13
<gsnedders>
whether you want to lump 201–5 in with 200 is a separate choice
12:13
<annevk>
jgraham: mnot would like us to expose them...
12:14
<annevk>
CORS has a 200-299 check somewhere so there's at least one place in the platform where we use this
12:15
<Ms2ger>
annevk, do you know if HTTP defines what to do with status codes that aren't three digits?
12:15
<annevk>
Ms2ger: parse error
12:16
<annevk>
Ms2ger: maybe that turns it into a HTTP/0.9 response?
12:16
<annevk>
Ms2ger: would be an interesting test for web-platform-tests
12:18
<Ms2ger>
Uhuh
12:19
<gsnedders>
in most things it turns into a HTTP/0.9 response, IIRC
12:19
<Ms2ger>
How do you detect that?
12:20
<annevk>
Ms2ger: 0.9 means the entire response is the body
12:20
<gsnedders>
You're aware HTTP/0.9 has no headers?
12:20
<Ms2ger>
No :)
12:20
<annevk>
It's why we have sniffing
12:20
<annevk>
And why <plaintext> is a thing
12:21
<annevk>
HTTP and HTML used to be intertwined a bit
12:23
<Ms2ger>
"Well-behaved servers should retrict line length to 80 characters"
12:25
<ms7821>
does http/2 mention simple-request?
12:47
<annevk>
Fullscreen API and <dialog>. Anyone from Chromium other than falken and philipj who can give input to this bug? https://www.w3.org/Bugs/Public/show_bug.cgi?id=27865
12:47
<annevk>
Also, anyone know plans from other vendors regarding <dialog>?
12:49
<gsnedders>
ms7821: I presume not.
13:39
<zcorpan>
Ms2ger: it seems to me it would be about equally hard for blink to add taintEnabled as it is for gecko to drop it
13:39
<Ms2ger>
zcorpan, seems plausible
13:39
<Ms2ger>
Stuff sucks
13:40
<zcorpan>
also no incentive apart from "interop!!"
13:41
<zcorpan>
i guess the spec should make it optional or something
13:44
<Ms2ger>
Yay interop :)
14:05
<annevk>
Warnings haven't helped?
14:43
<annevk>
GPHemsley: we should probably add a warning to https://wiki.whatwg.org/wiki/Sanitization_rules that it is not maintained and should probably not be used as reference without scrutiny
14:43
<annevk>
GPHemsley: did you have a template for that?
14:58
<rubys>
annevk: +1 to warning on wiki/Sanitization_rules ... it was never more than a work in progress
15:29
<Domenic>
annevk: seems ok, a bit sugary and possibly encouraging bad habits (e.g. being too lenient, when you expect a 201 a 200 is generally not ok). If another spec uses the check that does make me feel better. Probably a good idea on balance.
15:30
<annevk>
Domenic: yeah, I'm hesitant too, really depends on the protocol
15:31
<Domenic>
I think devs will like it. Just us grouchy pedants who grumble.
15:32
<annevk>
The issue mentioned in that thread about filtering out non-2xx does exist and this does help a lot with that so...
15:32
<annevk>
We'll see, it's fairly trivial to ignore if it's really only relevant for CORS
16:07
<GPHemsley>
annevk: I added {{obsolete}}. Is there a better place to point to, or is it just abandoned?
16:32
<annevk>
GPHemsley: abandonware though now there's template strings it might be worth looking into having a sanitized parser again
16:32
<annevk>
GPHemsley: though it'd be hard
16:33
<GPHemsley>
annevk: Alright, I was just curious as to whether there might be values for the 'spec' or 'see' parameters.
16:33
<GPHemsley>
(Which would point people in a better direction.)
16:34
<annevk>
GPHemsley: I don't think there is anything
16:52
annevk
reads TAG minutes
17:40
<smaug____>
hmm, is telemetry data wrong, or has mutation event usage gone down
17:40
<smaug____>
where is the tool to see feature usage on chromium?
17:42
<annevk>
smaug____: https://www.chromestatus.com/metrics/feature/timeline/popularity/148
17:43
<annevk>
There seems to be a downward trend but still quite high
17:43
<annevk>
Are you guys going to Edge London?
17:43
<annevk>
I was thinking of going
17:44
<annevk>
27 June
17:45
<smaug____>
yeah, still a bit high
17:46
<annevk>
But yay downward trend I guess
17:46
<annevk>
I wonder if we have to define it
17:49
<smaug____>
oh, chrome doesn't warn about use of mutation events
17:49
<annevk>
Paging fool... why is he not online?
17:50
<annevk>
SimonSapin: hey, I had a question about Servo. Is document.write() there? Is there much parallel code? Is </script> a special code path?
17:51
<Ms2ger>
No document.write()
17:52
<annevk>
smaug____: perhaps we should ask blink-dev?
17:54
<SimonSapin>
annevk: In order: not yet (https://github.com/servo/html5ever/issues/6), yes (especially in layout code), apparently not yet (https://github.com/servo/html5ever/blob/fc515000c0/src/tree_builder/rules.rs#L783)
17:54
<Ms2ger>
Nothing parallel in HTML parsing, though
17:55
<SimonSapin>
right
17:55
<SimonSapin>
In theory we could run character decoding, tokenization and tree building each in a thread, in a pipeline. But I don’t think we do.
17:57
<annevk>
SimonSapin: no parallel tree building?
17:57
<jgraham>
I think that tokenization-in-a-task was mooted as a GSoC project
17:57
<jgraham>
Moving charset decoding to a task seems like it might not be much of a win?
17:57
<SimonSapin>
I have no idea
17:58
<SimonSapin>
"mooted" as in we established it was not a win?
17:59
<annevk>
From another angle, how much hurt would it be to run script for creation and insertion of elements that have a dash in them?
17:59
<TabAtkins>
mooted as in "discussed"
17:59
<TabAtkins>
"moot" has multiple, somewhat opposing, meanings.
18:00
<annevk>
It would mean that part of the code could no longer be run in parallel
18:44
<annevk>
hsivonen: also https://twitter.com/marshray/status/560121170373386240
19:19
<hsivonen>
krijnhoetmer: thanks for the "Previous day" and "Next day" links
19:33
<Ms2ger>
^
21:53
<Domenic>
TabAtkins: can my text macros be empty strings somehow?
21:53
<TabAtkins>
Right now, no, but I can relax error-checking to allow it.
21:53
<Domenic>
There might be a better way... let me open an issue with a bit more details
22:01
<Domenic>
TabAtkins: let me know if https://github.com/tabatkins/bikeshed/issues/335 makes sense
22:02
<TabAtkins>
Yeah, it does.
22:03
<TabAtkins>
Domenic: I'd been planning to add some sort of conditional element, so you could delete elements unless they matched certain attributes of your document. (Probably starting with just status.)
22:03
<TabAtkins>
I think that'd work for you?
22:03
<Domenic>
TabAtkins: sounds excellent, as long as it works with custom metadata lines like `!Version History`
22:03
<TabAtkins>
I'd still need to do the "remove empty <dd>s" pass.
22:04
<TabAtkins>
Oh wait though. The problem is that you can't pass !Version History conditionally, right?
22:04
<Domenic>
Right
22:04
<TabAtkins>
Let's just figure out how to fix that directly.
22:04
<Domenic>
If I could pass empty strings then I could end up with an empty <dd>
22:04
<Domenic>
which visually is fine
22:05
<Domenic>
TabAtkins: actually, this might be unnecessary. I think I can just add a "back to the living standard" link and substitute that where the "go to commit snapshot" link would be
22:05
<Domenic>
Much easier.
22:05
<TabAtkins>
Heh, kk. And useful!
22:06
<TabAtkins>
but also, I think I can make custom metadata work just fine.
22:06
<Domenic>
Yeah, I figured it would be redundant with the giant red floating banner, but it's nice for symmetry... this way the TOC doesn't jump up and down a line as you go between LS and snapshot
22:13
<TabAtkins>
Domenic: Say I was wanting to integrate ecmd into Bikeshed directly. Most of the inline markup turns into plain HTML, but some turn into ecmarkup elements. What's the significance of these, and how should I reproduce their effects in HTML?
22:13
<TabAtkins>
Specifically, <emu-const> and <emu-nt>.
22:14
<Domenic>
TabAtkins: the idea is they are semantically better than <span class="const">, could potentially be custom-elemented in the future (although it seems unlikely you'd be able to do anything useful there), and that the EMU stylesheet will style them specifically.
22:14
<TabAtkins>
So I should stick with them, then?
22:14
<Domenic>
I think so
22:14
<Domenic>
See also https://bterlson.github.io/ecmarkup/
22:15
<Domenic>
Also: are you really looking into this!? I would be sooo excited.
22:17
<TabAtkins>
Yeah.
22:17
<Domenic>
^__^
22:17
<TabAtkins>
I've got more control over what kinds of markup shorthands can exist in a given document, so that gives me more freedom to add additional types of shorthands.
22:17
<TabAtkins>
Like, for example, all the JS shorthands.