00:07
<Krinkle>
annevk: Are we getting any closer to figuring out the fetch/abort situation?
00:08
<Krinkle>
I'm starting to think that something is better than nothing. I rarely go down that path, but it seems we're looking for a holy grail that doesn't exist.
00:41
<miketaylr>
zcorpon: oops, missed ur pings
00:41
<miketaylr>
and you're not here so unless you read logs, i'll find you tomorrow
06:01
<annevk>
Krinkle_: slowly
06:01
<annevk>
Krinkle_: JakeA is in charge
06:01
<annevk>
miketaylr: he reads the logs
06:31
<JakeA>
Krinkle: we're investigating cancelable promises first. If that doesn't work we have back-up plans. At the very least, the underlying streams will be terminateable
06:34
<daurnimator>
Krinkle_ / JakeA, got a summary of the problem?
06:36
<JakeA>
daurnimator: the problem is the request part of fetch(url) isn't abortable, neither are the body readers such as response.json()
06:37
<daurnimator>
ah okay; I was thinking of trying to solve my GC observability woes by performing requests; which according to the fetch spec: will be aborted if the request is collected; sounded like it might be related
07:03
<daurnimator>
JakeA: it certainly is an interesting question in general
07:04
<daurnimator>
http requests can be hard to abort => for HTTP/1.1, what if it's already sent, and something has been pipelined after it? the responses still has to be read off the socket
07:04
<daurnimator>
for http2 you can at least send a RST_STREAM
07:05
<daurnimator>
but at the high level; yeah, you'll want cancels to cascade; or the code isn't composable
07:05
<daurnimator>
which in the case of JS, implies that promise chains are cancelable
07:07
<JakeA>
daurnimator: yeah, from a js point of view you'd receive no further data, although it may not actually cancel the underlying request
07:08
<daurnimator>
JakeA: which brings up the question of why you want to bother cancelling in the first place if the server still has to send the reply, and the client still has to process it :P
07:10
<JakeA>
daurnimator: if it's a large resource such as a video, you'd want to cancel that. And as you say, http/2 has a good answer here
07:11
<daurnimator>
JakeA: true. the answer I was thinking of was it means that you won't need to bother following any redirects
07:16
<daurnimator>
this has brought up some good points for me to ponder on
07:17
<daurnimator>
'composibility of cancelability' => what is the smallest unit you can cancel => how can you propagate => how should it flow across error handling
07:35
<annevk>
Interesting differences between https://tools.ietf.org/html/std68 and https://tools.ietf.org/html/rfc5234
07:35
<annevk>
Latter seems more useful, but former is more permanent reference...
07:57
<MikeSmith>
annevk: I guess that fact you're spending time looking at those means you have plans to do something that would necessitate referencing them
08:05
<Ms2ger>
zcorpan, https://critic.hoppipolla.co.uk/r/5723 is back to you
08:09
<zcorpan>
Ms2ger: ports[http][1] is unnecessary but i suppose it's not a problem to have it there, right?
08:10
<Ms2ger>
Is it unnecessary?
08:11
<Ms2ger>
These run on port 8000 in automation
08:11
<zcorpan>
yeah, the test just needs cross-origin, which domains[www1] achieves already. no need to also have a different port
08:11
<Ms2ger>
Wouldn't {{domains[www1]}} just try port 80, which may or may not be mapped to wpt?
08:12
<zcorpan>
oh. yeah you're right
08:12
<Ms2ger>
Thanks, I like to be right :)
08:13
<zcorpan>
though it's still a different port compared to 8000 :-)
08:13
<zcorpan>
the docs for wptserve says ports[0] might be the only port, but i suppose other tests need another port also
08:16
<Ms2ger>
Calling jgraham
08:19
<zcorpan>
miketaylr: o/
08:23
<jgraham>
Ms2ger: Hmm?
08:23
<Ms2ger>
Ports usage in https://critic.hoppipolla.co.uk/r/5723
08:24
<jgraham>
Yeah, you have to specify a port, but it could be [0] there
08:35
<zcorpan>
jgraham: but [1] is not a problem, is it?
08:35
<jgraham>
No, there should always be a port [1]
08:35
<jgraham>
Except on https, perhaps
08:36
<zcorpan>
jgraham: the wptserve docs doesn't give such great confidence there will be another port :-)
08:38
<jgraham>
It seems like the wptserve docs shouldn't describe the details of the wpt setup since the server could be used in other contexts
08:39
<zcorpan>
i suppose
08:39
<jgraham>
Probably the testtwf docs should be improved
10:53
<annevk>
If anyone has ideas around compensating folks for doing standards work: https://twitter.com/annevk/status/636847756737388544
12:25
<hallvors>
Another quick XHR test review https://critic.hoppipolla.co.uk/r/5758 - hope it's quick anyway :)
13:08
<annevk>
hallvors: r+
13:08
<annevk>
hallvors: also, thank you
13:08
<annevk>
hallvors: if you have ideas btw how to reuse these tests for fetch()... that'd be great
13:30
<annevk>
More evidence why IDL needs internal slots: https://github.com/w3c/mediacapture-main/issues/232#issuecomment-132898987
13:30
<annevk>
heycam: ^^
13:31
<heycam>
annevk: I'm coming around on the idea of internal slots. doing some SVG rewording recently it would've been easier than writing a bunch of prose defining internal values associated with an object
13:33
<annevk>
aah yes, once you start writing proper specs the need is kind of self-evident :-P
13:33
<annevk>
good
13:38
<heycam>
:)
13:38
<heycam>
annevk, however sometimes I also wanted something like internal slots, but not using IDL values. so abstract list types or something.
13:39
<heycam>
which the ES spec does too, but then I'm not sure how well that fits with defining the internal slots in the IDL itself
13:39
<annevk>
heycam: I think in the end IDL has to define the meta-language for all specifications
13:39
<annevk>
heycam: including such constructs
13:39
<heycam>
ok
13:40
<heycam>
maybe we can just reuse the ES ones for those abstract types
13:40
<annevk>
Yeah, with an IDL flavor same as we do with Uint8Array et al
13:40
<heycam>
yeah ok
13:41
heycam
returns to SVG spec hacking
14:15
<annevk>
I wrote up a proposal for compensating folks to do standards work: https://etherpad.mozilla.org/standards-diversity-grants
14:15
<annevk>
I'm not sure yet whether I can actually convince my employer to do this, mind you
14:16
<wanderview>
annevk: I guess my concern is how real is that work without a browser vendor really behind it willing to do implementation?
14:16
<wanderview>
maybe thats not a valid concern
14:17
<annevk>
I actually worked on URL for a while with a grant
14:17
<wanderview>
or maybe there is a difference between archaeology-type spec work and new-feature-type spec work
14:17
<annevk>
I dunno, fixing open issues and such and writing tests always seems helpful
14:17
<darobin>
agreed
14:17
<darobin>
a grant is a good idea
14:17
<darobin>
I think it would likely work best if it's not "come up with a new standard" but rather work on existing issues and the such
14:17
<darobin>
at least at first
14:18
<annevk>
yeah
14:18
<darobin>
you can't grab someone with no clue about standards and make them write one, it wouldn't be nice to anyone
14:18
<darobin>
but you can get them to learn a lot by fixing specifics; then they become employable and you start having diversity without grants
14:19
<darobin>
you shouldn't make that Mozilla-only
14:19
<darobin>
I would expect other companies could chip in
14:19
<annevk>
my thinking was that it would be easier to start this way
14:19
<annevk>
they can always donate to MoFo and earmark it or some such
14:20
<darobin>
I guess yeah
14:20
<darobin>
it sure is easier to start with just the one person and just the one donor; if it works it'll convince people on its own
14:22
<jgraham>
So there are some issues with the details there. "You must have a GH id" is biasing the pool (people that already wok in Open Source, people that live in countries that have high GH usage, people that aren't shy of publishing their projects, people that don't prefer BitBucket or GitLab or whatever)
14:22
<jgraham>
Positive Discrimination is, at least, a way to generate controversy
14:22
<wanderview>
must be willing to have a GH account?
14:22
<darobin>
I think biasing against people who prefer BitBucker is a feature :)
14:23
<wanderview>
seems GH is required by virtue of us hosting the specs there
14:23
<nox>
Why?
14:23
<jgraham>
GH is required to make contributions, but there's no reason to only accept applicants who already have an account
14:24
<darobin>
I would not list strict requirements on any part, you just want someone who'll do whatever it takes to make something useful
14:24
<annevk>
jgraham: seems like it should be easy to create an account?
14:24
<jgraham>
The implication is that your credibility will be judged on the basis of your GH profile
14:24
<annevk>
I see
14:24
<darobin>
yeah
14:24
<annevk>
jgraham: removed
14:24
<darobin>
I mean if the bar is "you must be able to create a GitHub account" I would say it's pretty low
14:24
<darobin>
:)
14:24
<gsnedders>
darobin: does the Great Firewall not block GH?
14:24
<annevk>
yeah, but indeed then it doesn't make sense to have it as part of the proposal
14:25
<darobin>
gsnedders: sometimes it does, but not regularly
14:25
<jgraham>
Asking people to come up with their own plan and timeline is also probably a source of bias
14:25
<wanderview>
in theory anyone should be able to contribute using git offline and send patches to someone else to push for them
14:26
<annevk>
jgraham: other diversity grants asked for the same thing, but open to suggestions
14:26
<darobin>
part of the problem is that the people who are already able to identify useful work areas they could contribute to are already in the loop
14:26
<darobin>
I'm guessing those aren't he ones you actually want
14:26
<darobin>
given how obscure the world of standards is you might want to ask the candidate to demonstrate a willingness to help the Web platform or something like that
14:27
<darobin>
explain what their vision of the Web is, what they'd expect from Web tech, what they'd like to see that they aren't seeing yet
14:27
<darobin>
problems they've had, etc.
14:27
<jgraham>
Well also I expect that the confidence to write up a proposal is correlated with the probability of being in a group that is already well represented
14:27
<darobin>
pick someone with an interesting vision and the requisite tech background; the rest they'll have to pick up on the job
14:27
<wanderview>
clearly experience in bikeshedding should be on there :-)
14:28
<darobin>
yeah
14:28
<darobin>
1. Explain in 2500 words or more which colour the bike shed should be?
14:28
<darobin>
2. Detail how your coding style is vastly superior to everyone else's.
14:28
<zcorpan>
3. ???
14:28
<zcorpan>
4. Profit!
14:29
<darobin>
3. What is the best way to encode CSS and JSON in an XML syntax?
14:29
<jgraham>
Maybe the entrance exam should just be an argument with TabAtkins about why his indenting preferences are insane
14:29
<TabAtkins>
I'LL FIGHT YOU
14:30
<darobin>
I wasn't aware there was a pattern to them
14:30
<TabAtkins>
tabs for indentation, spaces for alignment
14:30
<TabAtkins>
Like a sane person
14:30
<wanderview>
fixing indentation could be a grant project
14:30
<gsnedders>
darobin: how do you best encode arbitrary semantics?
14:30
<darobin>
gsnedders: ooh that's a good one
14:30
<tobie>
by "underrepresented group," do you mean pretty much anyone that's not an asshole or is this a more traditional definition of the term?
14:30
<darobin>
put differently "Create a serialisation format for RDF."
14:30
<gsnedders>
darobin: (I'll argue that the answer depends on the goal)
14:31
<gsnedders>
darobin: BUT RDF/XML IS GREAT?!
14:31
<jgraham>
annevk: Anyway, I would say that if you want to make this work you should have some preprepared projects that people can apply to
14:31
<darobin>
gsnedders: only if you use XML 1.1!
14:31
<jgraham>
Rather than just saying "make something up"
14:31
<gsnedders>
darobin: i prefer 1.0 5th ed
14:31
<jgraham>
annevk: Also talk to dianeb and Mark Surman, perhaps
14:31
<darobin>
gsnedders: Philistine
14:32
zcorpan
prefers Unicode 3.0
14:32
<annevk>
jgraham: we could perhaps agree on some keyword to annotate issues with that would be good to tackle for independent contributors?
14:32
<jgraham>
annevk: web-platform-tests has occasionally marked issues with difficulty:easy
14:32
<gsnedders>
darobin: "convert RDF/XML to Turtle"
14:33
<darobin>
gsnedders: "Define a CSS syntax for RDF."
14:33
<jgraham>
annevk: This is not entirely unlike the TestTWF MDN fellowship that Ben Boyle did
14:33
<annevk>
jgraham: not sure it has to be easy, just something you could work on for a month or so
14:33
<gsnedders>
jgraham: BTW, is there any off-the-shelf way to get screenshots for CSS tests?
14:33
<jgraham>
gsnedders: What do you mean? wptrunner supports CSS tests
14:33
<zcorpan>
annevk: difficulty:easy means "good first bug"
14:33
<darobin>
yeah
14:33
<gsnedders>
darobin: "… as this is a vital missing link, having already replaced SQL"
14:33
<darobin>
lol
14:33
<gsnedders>
jgraham: I mean I actually want some way to get the actual screenshots of things that aren't reftests.
14:34
<annevk>
tobie: I basically want to say non-cis white male
14:34
<darobin>
annevk: I think tobie knew that :)
14:34
<annevk>
tobie: but it was suggested to me this is a better way of putting that
14:34
<gsnedders>
annevk: you missed the hetero part
14:34
<tobie>
I was just being an asshole myself. :)
14:34
<darobin>
non-cis doesn't include hetero these days?
14:34
<darobin>
I mean cis
14:34
<wanderview>
annevk: I commented on your gist about mailing list vs gh issues
14:35
<gsnedders>
darobin: cis only refers to gender, afaik
14:35
<jgraham>
gsnedders: You could hack wptrunner to do that perhaps, but it's not off the shelf
14:35
<gsnedders>
jgraham: right, that's what I thought
14:35
<tobie>
annevk: one of the benefits of more diversity is to have a new look at things, if that person gets do to work pre-decided by someone else, that sort of defeats the purpose
14:35
<darobin>
gsnedders: I guess it depends on if you consider gender to include orientation — it's a very old debate :)
14:35
<darobin>
cf Wittig, etc.
14:35
<annevk>
wanderview: is this how GH notifications work these days?
14:36
<tobie>
^ in reaction to jgraham's suggestions to have topics to choose from
14:36
<wanderview>
annevk: I've never received a notification for a gist before... so I didn't think they worked... sorry
14:36
<gsnedders>
darobin: true, though at least insofar as what I mostly hear people meaning by it, it doesn't…
14:36
gsnedders
hides
14:36
<darobin>
bikeshed all the genders!
14:38
<annevk>
wanderview: they used to, but they got removed at some point
14:38
<annevk>
wanderview: I was just making a joke about that
14:38
<wanderview>
ah
14:39
<wanderview>
annevk: I thought you were asking for feedback on the gist on twitter... but I see now you were just saying that was your thoughts... I should have just tweeted
14:39
wanderview
gets more coffee
14:44
<wanderview>
annevk: it would be nice if we had some central whatwg document that listed other specs to watch... so I could follow that and then know when to follow new repos
14:45
<annevk>
wanderview: there's https://resources.whatwg.org/biblio.json in https://github.com/whatwg/resources.whatwg.org
14:46
<jgraham>
On a somewhat related topic, in the next couple of days I'm planning to put together a session proposal for mozfest on how the open web is made, teaching about the standards process and how we achieve interop
14:46
<wanderview>
thanks
14:46
<annevk>
wow
14:46
<annevk>
that's great jgraham
14:46
<annevk>
will it be taped?
14:47
<jgraham>
I don't think so, mozfest sessions are supposed to be interactive rather than stand-at-the-front-and-talk
14:48
<jgraham>
I'd like to cover the process from having an idea to implementation in browsers, focusing on how discussions happen, to testing of the implementation, ending with an introduction to TestTWF as something that people can go away and explore more afterwardds
14:48
<jgraham>
So if anyone has any good ideas for fun learning activities to help people understand standardisation, please let me know
14:49
<darobin>
that's really cool jgraham
14:49
<astearns>
make sure when you get to testing the implementation you cover how testing uncovers specification bugs
14:50
<darobin>
maybe you should rope in Marcos, he's written a bunch of stuff about that
14:50
<darobin>
"fun learning activities to help people understand standardisation"
14:50
<darobin>
that sentence works if you drop either the first or the last word
14:51
<jgraham>
darobin: Well I asked and I was told that spending an hour telling people to poke their eyeballs with a fork was unlikely to get accepted
14:51
<darobin>
yeah, plus we're all against forking
14:51
<wanderview>
an hour learning the terminology used in specs
14:51
<darobin>
maybe chopsticks?
14:51
<gsnedders>
speaking of TestTWF, has that essentially died? nothing since Portland last year? nor any talk of anything new?
14:51
<wanderview>
"next we will discuss the difference between a 'flag' and a boolean value"
14:51
<jgraham>
wanderview: haha
14:52
<darobin>
well played wanderview
14:52
<darobin>
jgraham: do you have some notions of French?
14:52
<jgraham>
gsnedders: The series of events does seem to be pining for the fjords, yes
14:52
<darobin>
if so I might be able to unearth some pres I've given about standardisation
14:52
<wanderview>
terminology really is a huge barrier... but I have not way to teach it to someone other than to have them work with the specs for a while
14:52
<darobin>
but it's perhaps more geared towards explaining the concept to completely non-tech people
14:52
<jgraham>
darobin: I have the notion that it's a language principally spoken in Frace. I have very little understanding of it though ;)
14:53
<darobin>
yeah, so probably not helpful :)
14:53
<gsnedders>
jgraham, darobin: I thought that was the language of the Democratic Republic of the Congo?
14:53
<gsnedders>
France seems small in comparison.
14:55
<jgraham>
I was wondering who would point out something like that
15:08
<MikeSmith>
annevk: about https://etherpad.mozilla.org/standards-diversity-grants I would suggest not using the word "standards" but instead "new Web technologies"
15:09
<MikeSmith>
"creating specifications and tests for new Web technologies"
15:11
<MikeSmith>
"creating specifications and tests for new Web technologies (aka 'Web standards') intended to work interoperably in all browsers"
15:12
<MikeSmith>
s/tests/test suites/
15:16
<zcorpan>
hmm, i kinda like the idea of killing <applet> altogether
16:18
<annevk>
MikeSmith: you sound like Ian Jacobs
16:19
<annevk>
MikeSmith: but maybe "Web platform diversity grants" is better
16:19
<annevk>
I dunno
16:19
<annevk>
I should have started with, why would it be better?
16:20
<MikeSmith>
Ian Jacobs sounds pretty good sometimes
16:20
<MikeSmith>
he can play the piano
16:20
<MikeSmith>
I just don't think most non-insane people get excited about the word "standards"
16:21
<MikeSmith>
they want to create stuff
16:21
<MikeSmith>
write stuff
16:21
<MikeSmith>
write specs, write tests
16:22
<MikeSmith>
or at least those at the people we want
16:22
<MikeSmith>
they have to be a little insane to want to write specs and write tests
16:22
<MikeSmith>
admittedly
16:23
<MikeSmith>
anyway, I'm bakeshedding
16:23
<zcorpan>
you have a bakeshed?
16:26
<MikeSmith>
yep, for clambaking
21:06
<annevk>
Not really mentioned through Ian's email, there's now https://github.com/whatwg/html and https://twitter.com/htmlstandard
22:15
<ato>
annevk: \o/