03:57
<zcorpan>
mathiasbynens: www.w3c-test.org:82/html/infrastructure/urls/resolving-urls/query-encoding/
03:57
<zcorpan>
mathiasbynens: www.whatwg.org/specs/web-apps/current-work/multipage/infrastructure.html#resolve-a-url
04:03
<zcorpan>
mathiasbynens: https://www.w3.org/Bugs/Public/show_bug.cgi?id=23822
09:33
<annevk>
working with foolip is the best <3
09:45
<MikeSmith>
hsivonen: Can you remind me of there's a reason were sticking with the old version of jetty we're using for the validator
09:46
<foolip>
annevk: I love you dearly too, taking a look at the changes now :)
09:46
<MikeSmith>
hsivonen: I ask because, I think the version were using is not even getting security updates any longer
09:46
<foolip>
or after lunch, *poof*
10:22
<zcorpan_>
i wonder what the state of <legend> is these days
10:24
<zcorpan_>
looks like display:block doesn't work in blink at least
10:25
<zcorpan_>
<li><p>The <code>figure</code> element now uses a new element
10:25
<zcorpan_>
<code>figcaption</code> rather than <code>legend</code> because people
10:25
<zcorpan_>
want to use HTML long before it reaches W3C Recommendation.</li>
10:35
<MikeSmith>
zcorpan_: what is expected to ever change with the state of legend?
10:48
<zcorpan_>
MikeSmith: to be stylable like a normal element
10:49
<zcorpan_>
MikeSmith: at least when not in a fieldset
10:50
<zcorpan_>
MikeSmith: and follow http://www.whatwg.org/specs/web-apps/current-work/multipage/rendering.html#the-fieldset-and-legend-elements when it is
10:51
<MikeSmith>
Ah ok
10:52
<zcorpan_>
so it seems we still can't reuse <legend> for a new element
10:53
<zcorpan_>
and it will probably continue to be P5 because nobody cares about future elements not being able to reuse <legend>
12:58
<zcorpan_>
MikeSmith: i've updated http://platform.html5.org/history/
13:01
MikeSmith
looks
13:02
<MikeSmith>
excellemente
13:03
<MikeSmith>
zcorpan_: nice updates
13:04
<MikeSmith>
hmm the e.g., http://html5.org/r/1115-1153 stuff doesn't work any more
13:04
<zcorpan_>
thx. i didn't include new copies of the spec though
13:04
<zcorpan_>
yeah
13:04
<MikeSmith>
yeah I don't think we need new copies
13:12
<annevk>
MikeSmith: sorry, didn't realize anyone was using those links
13:12
<MikeSmith>
no worries, not sure we really needed them
13:13
<annevk>
SimonSapin: any updates on the IPv4 stuff?
13:16
<SimonSapin>
annevk: not really
13:17
<annevk>
SimonSapin: is Rust attempting to solve this somehow?
13:19
<SimonSapin>
annevk: when you open a socket with a string for the host Rust’s standard library parses IPv4 and IPv6 addresses, and for things that it considers neither uses the system’s thing for resolving names, which in turn also attempts to parse IPv4 and IPv6 addresses but depending on the system may or may not support exotic v4 formats
13:20
<annevk>
good times
13:20
<SimonSapin>
I looked into various POSIX APIs, it seems hard to force a DNS query and bypass the address parsing
13:21
<SimonSapin>
unless you’re willing to implement DNS yourself, I guess
13:21
<foolip>
annevk: not sure if I stumbled onto a bigger problem in https://www.w3.org/Bugs/Public/show_bug.cgi?id=26673
13:22
<annevk>
foolip: not really sure how to solve it yet
13:22
<annevk>
SimonSapin: sigh
13:24
<foolip>
annevk: I think the bigger risk is getting some detached documents with non-empty fullscreen elements stacks, more than spurious fullscreenchange
13:24
<SimonSapin>
annevk: in any case, I don’t know what’s the right thing to do. There’s a couple of 6 years-old Mozilla bugs about this where people disagree with each other
13:25
<annevk>
SimonSapin: pointers?
13:25
<SimonSapin>
https://github.com/w3c/web-platform-tests/issues/1104#issuecomment-48917917
13:27
<erlehmann>
annevk fug, why you break links :(
13:27
<erlehmann>
._.
13:27
<erlehmann>
i certainly used one revision link in a blog post
13:28
<erlehmann>
i don't remember which one
13:28
<erlehmann>
i should make a private wayback machine
13:28
<erlehmann>
link rot is becoming too much
13:28
<annevk>
erlehmann: I need some context
13:28
<erlehmann>
<MikeSmith> hmm the e.g., http://html5.org/r/1115-1153 stuff doesn't work any more
13:29
<erlehmann>
<annevk> MikeSmith: sorry, didn't realize anyone was using those links
13:29
<erlehmann>
maybe
13:29
<annevk>
erlehmann: ah, right
13:29
<erlehmann>
annevk i assumed you broke it.
13:29
<erlehmann>
was it someone else?
13:29
<erlehmann>
i need to fight link rot. hmm.
13:29
<foolip>
erlehmann: you don't need a private one, there's "Save Page Now" on http://archive.org/web/
13:30
<annevk>
erlehmann: all that broke was a sequence of revisions as the tool no longer supports that
13:30
<erlehmann>
annevk so links to revisions still work?
13:30
<annevk>
erlehmann: yes
13:30
<erlehmann>
i think i linked to revisions.
13:30
<erlehmann>
i am relieved
13:30
<annevk>
erlehmann: e.g. http://html5.org/r/1115
13:30
<erlehmann>
13:30
<foolip>
linking to a range of revisions doesn't seem useful unless Hixie happened to work on exactly one thing for a long time
13:30
<annevk>
erlehmann: the hyphen thing was a special feature, not really recommended
13:31
<erlehmann>
it conflated items into a collection
13:31
<erlehmann>
interesting
13:31
<annevk>
foolip: yup
13:31
<erlehmann>
does any one of you have a plain text blogging system and can link me to it? i want to add multimedia to http://news.dieweltistgarnichtso.net and have no idea about file structure
13:32
<zewt>
.plan? heh
13:32
<erlehmann>
s/plain text/filesystem backed/g
13:33
<erlehmann>
zewt i use single level git repositories with html files in it. hipster news 2 generates feed and overview.
13:34
<zewt>
grr, now firefox is randomly (Not Responding) and i can't tell why
13:34
<annevk>
foolip: so for fullscreen.spec.whatwg.org/#dom-document-exitfullscreen you basically want that during the iteration to dispatch the events we check if the situation with regards to the amount of elements on the fullscreen element stack has changed?
13:35
<annevk>
foolip: and if it has changed, we fully exit fullscreen?
13:59
<MikeSmith>
foolip: liking to a range of revisions is probably also useful if you're trying to document what things changed during a certain period of time -- which is exactly what http://platform.html5.org/history/ is for. But I guess that's a corner case that not many other people would want to do
13:59
<MikeSmith>
regardless, I think we could reconstruct it with links to the github html-mirror change history instead
14:00
<MikeSmith>
if we really wanted to
14:01
<MikeSmith>
(that is, construct alternative links to replace the ones we had in http://platform.html5.org/history/ to the html5.org generated changelogs that we removed after they broke)
14:05
<zcorpan_>
MikeSmith: might be more useful to have individual links to commits instead of month or year-long range of commits
14:06
<wanderview>
JakeA: do you have a sense of how close we are to updating the spec language for the expected Cache/CacheStorage changes?
14:08
<MikeSmith>
zcorpan_: yeah that would definitely be a lot more useful
14:09
<JakeA>
wanderview: We're unblocked on streams stuff, just need to get consensus on the API. I'm getting a talk together for jsconf so that's got my focus. I'll see if slightlyoff is free to take over dragging that to consensus for the next couple of weeks.
14:09
<JakeA>
Don't want to hold up the process, it's been held up enough
14:10
<wanderview>
JakeA: ah, ok... thanks for the update... I think I will implement mostly whats currently spec'd with an eye towards the expected changes then
14:11
<JakeA>
wanderview: I don't expect the querying stuff to change, just .add vs .addAll, .delete vs .deleteAll
14:11
<wanderview>
I guess i thought there was more consensus on the rest of the spec changes, but I probably skimmed some bits in the thread
14:12
<wanderview>
JakeA: great, thanks... good luck with the jsconf talk!
14:50
<annevk>
JakeA: wanderview: I thought we had consensus on the API?
14:50
<annevk>
JakeA: wanderview: short method names directly on Response/Request, plus a bodyUsed property
14:50
<JakeA>
annevk: Talking about the cache API
14:51
<annevk>
JakeA: wanderview: I guess my plan was to make that change today, might have to wait a bit
14:51
<wanderview>
annevk: I think JakeA was saying other cache API questions beyond the streams issue
14:51
<JakeA>
annevk: I'm happy with the Request/Response stuff
14:51
<annevk>
ok, that's what I get for only reading part of your exchange
15:10
<JakeA>
wanderview: What's the story around debugging serviceworker in Firefox?
15:11
<wanderview>
JakeA: currently not great as far as I know... I believe work is in progress to allow debugging workers at all :-\
15:11
<wanderview>
let me see if I can find the bug number
15:11
<wanderview>
https://bugzilla.mozilla.org/show_bug.cgi?id=1003097
15:12
<wanderview>
JakeA: and here is our bug to support ServiceWorkers in devtools: https://bugzilla.mozilla.org/show_bug.cgi?id=943220
15:14
<wanderview>
JakeA: asking what the plans/timeline for that work is... we definitely want to support it
15:15
<JakeA>
wanderview: Cool! Making sure I represent the state-of-things correctly in this jsconf talk. Is there any way to get console logs out of a SW in the interim?
15:15
<wanderview>
JakeA: console.log is supported in gecko workers, yes
15:15
<JakeA>
wanderview: Where is it exposed?
15:16
<wanderview>
JakeA: uh... I think self... let me check
15:20
<smaug____>
JakeA: type console.log(123) to the first textarea. http://mozilla.pettay.fi/workerconsole/
15:20
<smaug____>
the code will be executed in a worker
15:20
<smaug____>
and log goes to browser console
15:21
<smaug____>
(ctrl+shift+k)
15:21
<smaug____>
s/browser/web/
15:22
<wanderview>
smaug____: thanks! I was trying to verify in my mochitest, but I guess we suppress console.log there
15:23
wanderview
bookmarks smaug's workerconsole...
15:23
<JakeA>
Wonder if that works for serviceworker (since they don't have an owner window)
15:23
<smaug____>
ah, no idea
15:23
smaug____
knows next to nothing about service workers
15:24
<smaug____>
hmm, Blink has odd console object
15:24
<smaug____>
[object WorkerConsole]
15:25
<wanderview>
JakeA: hmm, thats a good point... I was actually assuming it would work since SW's extend workers
15:27
<wanderview>
JakeA: yea, baku confirms that console.log() won't work from a SW because of the missing window... sorry for my confusion
15:28
<annevk>
wanderview: want to file a bug on that?
15:28
<annevk>
wanderview: no debugging would be a major pita
15:28
<wanderview>
annevk: baku is doing that now
15:28
<wanderview>
I agree
15:28
<annevk>
yay
15:28
<JakeA>
wanderview: No worries! Is there any way to get those logs in the meantime, even if they go down to the command line?
15:30
<Ms2ger>
dump()?
15:30
<Ms2ger>
Probably not
15:32
<wanderview>
JakeA: can I get back to you tomorrow? I feel like we need to get our plan straight on this... when is your talk?
15:33
<JakeA>
wanderview: in two week, so no major rush. Want to get as many demos working in Firefox as possible though
15:34
<zewt>
could you proxy through a shared worker as a workaround?
15:34
<wanderview>
JakeA: ok, thanks!
15:36
<JakeA>
zewt: I think shared workers have the same issue. Could send stuff back to the page via a message port, I think that's landing soon in Firefox
15:36
<zewt>
right, you'd have to open the shared worker from a real document to receive them
15:36
<JakeA>
wanderview: will update https://jakearchibald.github.io/isserviceworkerready/ in time for the talk too
15:39
<wanderview>
JakeA: thanks... we have an internal meeting tomorrow morning... going to discuss it there
15:42
<JakeA>
zewt: good idea
15:46
<zewt>
i'd think that "service workers" would just be a type of shared worker, so you could connect to them like any shared worker
15:47
<zewt>
like a shared worker url that the browser also knows how to start up on its own
16:10
<dvorapa>
Hello! Is there anybody, who could I speak to?
16:12
<dvorapa>
Could I ask about my new meta extension?
16:12
<Ms2ger>
You can
16:14
<dvorapa>
I want to propose meta extension with keyword `version`
16:14
<dvorapa>
Specification: https://github.com/dvorapa/meta-version
16:18
<annevk>
dvorapa: if you're asking for a wiki account I need your desired username and email
16:18
<dvorapa>
Okay,username: dvorapa
16:18
<dvorapa>
email: dvorapa⊙sc
16:19
<annevk>
dvorapa: "A randomly generated password for Dvorapa has been sent to dvorapa⊙sc It can be changed on the change password page upon logging in."
16:21
<dvorapa>
annevk: Thank you.
17:37
<Domenic>
marcosc_: why does everyone want these separate WakeLock objects... why do we need to be more programmer-footgun-protective than native platforms are.
17:38
<jgraham>
People have different expectations from browsing to a website vs running a native app?
17:39
<Domenic>
this isn't even on that level
17:39
<Domenic>
it's "what if multiple pieces of code within the same website want a wake lock"
17:39
<Domenic>
in native apps they all coordinate to share a per-app lock; if different parts of the code need to coordinate, then they write code to do so
17:40
<Domenic>
whereas sicking and others think it's important that different parts of the code each get their own lock
17:42
<sicking>
Domenic: i don't feel strongly. But the editors claim that they care about that use case, but the proposed API doesn't support it
17:42
<Domenic>
oh, did not realize...
17:44
<sicking>
but it does seem like unless we specify a standard for how to do this coordination, then the JS world will have to come up with their own standard for it. And not just one standard per library, but rather one "the" standard
17:44
<sicking>
mmmm..
17:45
<sicking>
maybe it could work for everyone to do their own solution, if that solution is based on callbacks to the embedder
17:45
<sicking>
either way, it's a relatively small difference in API complexity
17:46
<sicking>
either "screen wakelock" is a property on the document, or you can instantiate it using "new"
17:47
<sicking>
in both solutions the actual API on the lock will be the same
17:54
<Hixie>
can someone work out what Axel means in https://www.w3.org/Bugs/Public/show_bug.cgi?id=26669 and add a clarifying comment on the bug?
18:09
<Domenic>
not "the" standard, just, if you want to use libraries that use wakelocks, they have to communicate that somehow
18:11
<Hixie>
anyone have a better term for "media source object" that has neither the words "media source", "media stream", "file", "media resource", "media data", or "blob" in it?
18:11
<Hixie>
(meaning something that is either a MediaSource, MediaStream, File, or Blob)
18:11
<Hixie>
maybe "media provider object"
18:13
<SamB__>
too bad interfaces aren't a thing
18:13
<Hixie>
even if it was an interface it'd still need a name :-)
18:13
<SamB__>
or I guess those don't actually all implement the same interface anyway
18:14
<jgraham>
Hixie: I guess he's saying something like <form autofocus="foo"><input id="foo"> would have been a better model
18:14
<jgraham>
But that's a guess
18:15
<SamB__>
jgraham: certainly less questions about how to resolve conflicts than with autofocus=yes on individual elements
18:16
<jgraham>
SamB__: Arguably an equivalent number of questions <form autofocus=foo><input id=foo><input id=foo>
18:17
<Hixie>
not to mention <form autofocus=foo></form><form autofocus=bar></form><input id=foo><input id=bar>
18:17
<Hixie>
jgraham: thanks. commented on the bug.
18:17
<SamB__>
jgraham: That question is nothing new, though
18:18
<SamB__>
Hixie: hmm, okay, that's a better example ;-)
18:23
<Hixie>
("not to mention" is such a dumb phrase)
18:26
<tantek>
not to mention "to be honest"
18:37
<hsivonen>
MikeSmith: the reason to use an old version of Jetty is that I haven't had the time to update Jetty
18:37
<hsivonen>
MikeSmith: we should keep updating Jetty
18:37
<hsivonen>
MikeSmith: fortunately, the security bugs *usually* don't affect the way we use Jetty
18:51
<Hixie>
tantek: "to be honest" is slightly better in that it implies that one might have tried to conceal or sugar coat the statement but has decided not to
18:52
<tantek>
Hixie, good point. I often hear "not to mention" abbreviated these days in speech as "Sooooo… " preceding a statement.
18:52
<Hixie>
heh
18:53
<SamB__>
I guess "not to mention" only makes sense when followed by one or more classes of things that you aren't listing individually
21:48
<JonathanNeal>
Hey https://groups.google.com/a/chromium.org/forum/#!msg/blink-dev/07v_yMErc_A/WcyKGN1A2y0J
21:48
<JonathanNeal>
Nice!
21:49
<caitp>
eh, wouldn't get your hopes up
21:50
<caitp>
the two replies so far seem pretty conservative about adding it to the platform as a native feature, and probably not even blink-in-js. But, it would be fun to work on, so here's hopin :3
21:52
<jsbell>
caitp: If you're thinking blink-in-js, then start with a polyfill.
21:52
<JonathanNeal>
blink in js?
21:52
<jsbell>
JonathanNeal: Implementing bits of blink in JS.
21:53
<caitp>
it might come to that, but the features I've implemented so far have been smaller in C++ than the mix of C++ boilerplate + private scripts
21:53
<jsbell>
e.g. lets say we didn't have 'window.atob()'; take a polyfill, ship it with the browser, but with appropriate security sprinkles.
21:54
<JonathanNeal>
oh you mean like a browser extension?
21:55
<jsbell>
JonathanNeal: "like" as in: same security concerns for JS code needing to run in what we call isolated worlds. But from a browser user's perspective it's a feature that ships with the browser and happens to be implemented in C++ rather than JS
21:56
<jsbell>
https://docs.google.com/presentation/d/1XvZdAF29Fgn19GCjDhHhlsECJAfOR49tpUFWrbtQAwU/edit#slide=id.g3840fe06e_00
21:56
<caitp>
I think, if we do end up getting it to support custom data and custom sort predicates, then a Blink-in-JS implementation is essentially impossible :(
21:56
<caitp>
because of those security concerns
21:58
<Hixie>
custom sort predicates?
21:58
<Hixie>
(custom data is already in the spec)
21:58
<erlehmann>
JonathanNeal how is your polyfill going? :)
21:59
<JonathanNeal>
erlehmann: Further than when last I talked about it in here. When I have free time next, I would like to start adding the more complex sorting algorithms.
22:00
<JonathanNeal>
erlehmann: http://sandbox.thewikies.com/table-sorting/ is up to date
22:01
<caitp>
Hixie: similar to Array.prototype.sort
22:01
<caitp>
letting the script assign a callback to the column header telling it how to sort its keys
22:02
<caitp>
er, sort its data
22:02
<Hixie>
caitp: i don't see how that could work. see the e-mails i sent at the time for a discussion of why.
22:02
<erlehmann>
JonathanNeal date sorting could be interesting
22:02
<Hixie>
(but basically, it's because the first callback could cause a navigation, move the table to another document, put each <td> in a different iframe, and who the heck knows what else)
22:03
<caitp>
sure it could, but who cares
22:03
<caitp>
if a user wants to break their own site, what of it :p
22:03
<Hixie>
it's not about breaking the site
22:04
<Hixie>
it's about breaking the browser
22:06
<caitp>
also I'm not sure the <data value="someString"> really counts as custom data, it's a custom string
22:06
<caitp>
it's not as flexible as say, a JSObject
22:06
<Hixie>
how would you sort a JS Object?
22:06
<caitp>
with a custom predicate!
22:06
<jsbell>
And before someone suggests it: evaluating a short snippet of script in a completely isolated execution context turns out to be something browsers are inefficient at, at the moment
22:06
<Hixie>
then it's a custom predicate, not custom data.
22:07
<caitp>
well I mean to do anything meaningful with it you'd need both
22:07
<Hixie>
jsbell: it also doesn't work when what you have to pass the isolated script is a pair of DOM nodes. :-)
22:07
<SamB__>
jsbell: that's no surprise
22:07
<Hixie>
caitp: you can already assign whatever you want to an element
22:07
<SamB__>
jsbell: building those contexts tends to be expensive
22:07
<Hixie>
caitp: myElement._myCustomData = {}
22:07
<Hixie>
or whateer
22:08
<caitp>
sure, but the sorting algorithm won't consider it when sorting
22:08
<jsbell>
Hixie: Yep. (But someone will suggest passing a serialization or something, which doesn't help)
22:08
<Hixie>
caitp: you already said you need a custom predicate
22:08
<jsbell>
SamB__: Yep. We'd like that for e.g. custom indexers in Indexed DB, but the engines aren't optimized for it (yet)
22:09
<caitp>
I'm saying, without a custom predicate, "custom data" eg attached to a node, wouldn't be much good
22:09
<caitp>
because it wouldn't be used
22:09
<SamB__>
processes, threads, JS sandboxes ...
22:09
<jsbell>
And then there's pulling in libraries to do anything useful... anyway, non-starter right now
22:09
<SamB__>
not sure about lua though
22:10
<caitp>
yeah, the performance and security concerns of a custom predicate are valid concerns
22:10
<SamB__>
jsbell: I wonder if it would be potentially faster to recycle an existing context than to make a whole new one?
22:10
<Hixie>
caitp: that's what i said before
22:10
<caitp>
I am agreeing with you
22:10
<Hixie>
ok good. so why are we still talking about custom data
22:11
<caitp>
but at the same time, without the ability to do that, people might very well not find it worth using the feature
22:11
<caitp>
and would resort to using jQuery-plugin-foo instead
22:11
<Hixie>
the spec allows for a custom sorter for exactly this reason
22:11
<Hixie>
oh hey, good news everyone! I've figured out a way to bypass the spec's pop-up blocker... and chrome implements the spec
22:13
<caitp>
where exactly does the spec allow for a custom sorter? it must not be in the tables document of the multipage spec
22:14
<jsbell>
SamB__: Dunno. I suspect competition on speed and power vs. things like ES6 modules (and blink in JS...) will drive down the cost of contexts
22:14
<Hixie>
<table onsort="...run whatever algorithm you want to sort the table here...">
22:14
<caitp>
oh, I see.
22:14
<Hixie>
it lets the browser handle all the "table changed" logic, "this is the sorted column" logic, etc
22:14
<Hixie>
and just fires the event when the table needs sorting
22:15
<Hixie>
there's two ways to use it, basically
22:15
<caitp>
so you're saying they would sort the table manually and then prevent default to stop the default logic
22:15
<Hixie>
you can either just actually sort the table, then return false
22:15
<Hixie>
or you can set the custom sort data to the rows in the order you want, then return true
22:16
<caitp>
I mean that could work, but that's a lot more complicated than a predicate
22:17
<caitp>
so it seems unlikely most people would use it
22:17
<Hixie>
it's actually only a very few lines more complicated
22:17
<Hixie>
probably just a matter of putting the rows in an array, sorting the array using a custom predicate, and then reappending the rows in the order you care about
22:18
<Hixie>
when you have your own function you don't have to worry about all the edge cases the algorithm normally worries about
22:18
<Hixie>
like multiple <tbody>s, or row-spanning cells, or all that nonsense
22:26
<roc>
some ability to run pure JS functions cheaply would be useful in various places in the platform
22:26
<roc>
vertex shaders for CSS filters
22:27
<roc>
various extensible layout things
22:27
<roc>
audio processing
22:45
<JonathanNeal>
I think people will demand the ability to sort tables by their own method.
22:50
<caitp>
they certainly will
23:02
<caitp>
anyway JonathanNeal I'm probably not going to get a green light on that intent-to-implement thread, but I'm happy to help work on the polyfill if you want some assistance with it
23:44
<JonathanNeal>
caitp: yes, once it’s in slightly better shape I need to put it on GitHub.