00:23
<MikeSmith>
Domenic: hmm yeah
00:24
<MikeSmith>
and I'm just trusting that tool of Chris's is testing what it seems to be testing (because too lazy to try testing it directly right now)
00:25
<MikeSmith>
I'm busy reading through the now-finally-not-a-private-repo-any-more wasm stuff
00:26
<MikeSmith>
after weeks of luring on the #webassembly IRC channel and trying to reverse-engineer from check-in comments that commit bot posts there
00:26
<MikeSmith>
*lurking
00:29
<jsbell>
MikeSmith: re: encodings - https://encoding.spec.whatwg.org/#names-and-labels
00:31
MikeSmith
looks
00:33
<jsbell>
depends on context, but if it's not in the table the result is "failure" - which means depends on context
00:33
<MikeSmith>
jsbell: I guess annevk needs to update the "utf-8" row in that table to say "also any string that's not a name or label for any other encoding listed in this table"
00:33
<jsbell>
for html, it's still good ol' windows-1252
00:33
<MikeSmith>
oh
00:34
<MikeSmith>
jsbell: so is http://www.lookout.net/test/charsets/iana-charset-support/ not showing what it seems to be showing?
00:34
<MikeSmith>
or are all UAs just not conforming on this?
00:35
<MikeSmith>
hmm now that I say that out loud I realize already know that HTML parsers fall back to windows-1252
00:35
<jsbell>
ugh, contentDocument.charset/characterSet is a whole nother set of aliases/logic...
00:36
MikeSmith
drafts reply to Chris Weber
00:37
MikeSmith
sees that jsbell already beat him to it
00:37
<jsbell>
well, I was missing context so I'm not sure I'm helping.
00:38
<jsbell>
hrm, those aliases aren't as bad as I remember apparently: https://github.com/w3c/web-platform-tests/blob/master/encoding/resources/encodings.js
00:41
<karlcow>
TabAtkins: +1 to the meta element in SVG. that would save a lot of insanity.
00:42
<heycam>
that, and link, would make my life easier writing reftests
00:42
<TabAtkins>
Yeah, we want it for a few other things, too, like declaring SVG Params (the ability to set a CSS custom property via a query parameter).
00:42
<TabAtkins>
Yeah, <link> too.
00:42
<heycam>
but it would need us to finally come up with a plan for supporting these from-HTML elements without duplicating them in the SVG namespace
00:43
<heycam>
(a plan that everyone can live with, that is; we've had a couple of plans...)
00:43
<TabAtkins>
The only "issue" is that in an XML syntax you need to specify the HTML namespace for them.
00:43
<TabAtkins>
(Note that we're now finally actively looking into just using the foreign-content parser for standalone SVG. More than just talk, finally.)
00:44
<TabAtkins>
(That would avoid the need; HTML elements would just be given the HTML namespace automatically.)
00:45
<heycam>
TabAtkins, is that Blink experimenting with using HTML parser for standalone SVG documents?
00:45
<TabAtkins>
Yeah.
00:45
<heycam>
do you have a link for any discussion/announcement?
00:45
<TabAtkins>
Not yet, just internal discussions. We're throwing around some use-counters right now.
00:45
<heycam>
ok. look forward to hearing the outcomes.
00:46
<TabAtkins>
(Attempting to eliminate XML from our engine entirely, if possible. We only need it for SVG, XHR.responseXML, and XSLT.)
00:46
<TabAtkins>
Moving SVG to the HTML parser would be easy and better for authors (consistent syntax), XHR we'll see if we can move to a simplistic parser like XML5 or similar. XSLT, lol I dunno.
00:47
<heycam>
I thought you already were going to just implement XSLT in JS
00:48
<TabAtkins>
That's vaguely the plan, yeah.
00:49
<boogyman>
TabAtkins: re: (the ability to set a CSS custom property via a query parameter). <-- What's the use-case? Why would you want the client to control the aesthetics of your app?
00:50
<TabAtkins>
boogyman: SVG images that can have colors/etc set at time of use. <img src="logo.svg?main-color=blue">
00:50
<TabAtkins>
You have to pre-declare which properties are Params, to prevent arbitrary screwing around.
01:25
<MikeSmith>
ah rektide
01:26
<MikeSmith>
it didn't take you long
01:26
<MikeSmith>
https://groups.google.com/d/msg/emscripten-discuss/k-egXO7AkJY/GLqWCJCPnqwJ
01:26
<MikeSmith>
you need one of those signs man
01:27
<MikeSmith>
"The End is Near"
03:26
<hayato>
Does someone know why DocumentFragment doesn't support innerHTML? It looks only Element support it.
03:33
<hayato>
The context is https://code.google.com/p/chromium/issues/detail?id=497175 . I am wondering how ShadowRoot.innerHTML should behave.
03:57
<Domenic>
hayato: there was a thread a few years back, let me find it
03:57
<Domenic>
hayato: https://lists.w3.org/Archives/Public/public-webapps/2011OctDec/0663
03:57
<hayato>
Domenic: Thank you!
03:59
<Domenic>
lots of support from blink people, here's some from Moz... https://lists.w3.org/Archives/Public/public-webapps/2011OctDec/0670.html
03:59
<Domenic>
Here's some from rniwa... https://lists.w3.org/Archives/Public/public-webapps/2011OctDec/0675.html
04:00
<Domenic>
this one outlines some of the hard cases... https://lists.w3.org/Archives/Public/public-webapps/2011OctDec/0766.html
04:02
<Domenic>
<a> is a problem, SVG vs. HTML https://lists.w3.org/Archives/Public/public-webapps/2011OctDec/0812.html
04:02
<Domenic>
so in conclusion, people like it, but edge cases are probably hard
04:10
<rniwa>
Domenic: I don't think you should consider my comment in 2011 as endorsement from WebKit project
04:10
<rniwa>
Domenic: since i was working for Google at the time.
04:10
<Domenic>
haha right, ok
04:11
<rniwa>
(note that i would necessarily oppose to this particular proposal)
04:11
<boogyman>
rniwa: were you speaking on behalf of yourself or the company?
04:14
<rniwa>
boogyman: in public-webapps, I was a representative of Google until October 2012.
04:14
<rniwa>
boogyman: now I'm a representative of Apple
04:15
<boogyman>
cheers
04:15
<rniwa>
boogyman: do I speak on behalf of my employer? probably not.
04:15
<rniwa>
boogyman: do my opinions change based on goals/needs of my employer? absolutely.
04:17
<boogyman>
Understood. classic User v Business model for making decisions.
04:18
<rniwa>
yeah, each company has its own priority and rightly so.
05:10
<annevk>
MikeSmith: yeah, there's no defaulting to utf-8
05:24
<MikeSmith>
annevk: yeah I don't know what Chris's test is doing
05:30
<annevk>
MikeSmith: when can you review the new URL tests?
05:30
<MikeSmith>
in a few hours
05:30
<annevk>
MikeSmith: thanks!
05:30
<annevk>
MikeSmith: I might create some more today, and I'm super bad with git
05:30
<annevk>
so I'd rather not get a bunch of merge conflicts
05:31
<MikeSmith>
OK
05:31
<MikeSmith>
I reckon I can get to it within 1 hour
06:36
<MikeSmith>
annevk: gedaan
06:48
<annevk>
MikeSmith: sugoi
06:55
<annevk>
https://github.com/w3c/web-platform-tests/commit/2717ed0b937d4612bd18261559bd0dd7800fba96 this looks really bad... SRI fixes shouldn't affect all those files...
06:55
<annevk>
Or is that just how bad merges go? How do we know that did not overwrite any other work? Seems impossible to review...
06:58
<MikeSmith>
annevk: yeah that stuff is all because Joel flubbed and ended up with a branch merge tin that PR
06:59
<MikeSmith>
unfortunately with git that's a very easy mistake to make
06:59
<MikeSmith>
but it's not hard to unwind
06:59
<annevk>
But how do we know nothing bad happened?
06:59
<MikeSmith>
you can look at the commits separately
07:00
<MikeSmith>
though he should probably re-submit another PR with the same commits minus that merge
07:00
<MikeSmith>
or I can do it for him and he can clone the branch I make
07:01
<annevk>
well all the commits already went into master
07:01
<annevk>
oh well
07:01
<annevk>
perhaps there should be some tighter control on who can do the final commits so the whole thing makes a bit more sense
07:01
<MikeSmith>
ah christ you're right
07:02
<MikeSmith>
I thought he had committed to a branch
07:02
<MikeSmith>
yeah that's .. not so great
07:02
<MikeSmith>
but we can still unwind it
07:03
<annevk>
I've started doing things the Domenic-way for my repos so you don't even get merge commits
07:03
<MikeSmith>
annevk: about tighter control we have a soft policy that nobody should commit any of their own changes to master
07:03
<annevk>
I just attribute authorship through --author although that's rather cumbersome
07:03
<MikeSmith>
yeah that works too
07:03
<annevk>
I want GitHub to support an author field in their client
07:04
<annevk>
So I can just use @domenic or some such
07:04
<MikeSmith>
I've never used their client
07:04
<MikeSmith>
I just use the git comment
07:04
<MikeSmith>
*command
07:04
<annevk>
I guess I should learn to use that
07:04
<annevk>
Main problem with the command line and me is multiline stuff
07:05
<MikeSmith>
man I hope Joel did all that by mistake. I mean as far as committing it master
07:05
<annevk>
Well and remembering all the silly parameters
07:06
<MikeSmith>
Brad reviewed that PR, but it didn't have those branch merges in it when he reviewed it
07:06
<MikeSmith>
what probably should have better happened is that Brad just pushed the merge button in the github UI after he reviewed it
07:07
<MikeSmith>
what seems to have happened instead is that Joel committed the branch merges after Brad had reviewed it, then Joel pushed that
07:07
<MikeSmith>
anyway, nothing fatal
07:12
<annevk>
MikeSmith: also, web-platform-tests has a lot of old branches
07:12
<annevk>
MikeSmith: I assume not all of those are active anymore
07:12
<annevk>
MikeSmith: could we maybe kill some of them?
07:19
<MikeSmith>
annevk: yeah I will right now
07:19
<MikeSmith>
kill some
07:20
<annevk>
Maybe https://bugzilla.mozilla.org/show_bug.cgi?id=1170774#c4 should be documented on the wiki
07:20
<annevk>
Amount of WHATWG Standards is slowly growing
07:29
<MikeSmith>
may a hundred flowers bloom
07:29
<MikeSmith>
and a hundred schools of thought contend
07:36
<MikeSmith>
annevk: deleted all the old branches I safely could
07:37
<MikeSmith>
many of the remaining ones were either created by jgraham or Ms2ger and have no associated PR so they would need to check and see if it's a branch they're planning to work again on ever or not
07:38
<MikeSmith>
annevk: btw one of those branches-in-limbo is yours, https://github.com/w3c/web-platform-tests/tree/encoding-and-css
07:38
<MikeSmith>
with https://github.com/w3c/web-platform-tests/commit/02ac3d372060393f4c0a6a0f29a95dd9f0ee8fda
08:14
<annevk>
I wonder if we should require owners to setup 2-factor auth... https://github.com/orgs/whatwg/people
08:15
<annevk>
MikeSmith: yeah I know, I was hoping someone would take that over
08:15
<annevk>
MikeSmith: but despite all the talk from the i18n WG, nothing much materialized
08:15
<MikeSmith>
annevk: well it's not doing any harm so you could just leave it
08:16
<MikeSmith>
annevk: no comment šŸ˜†
08:16
<MikeSmith>
we need a template for that sentence
08:16
<MikeSmith>
"despite all the talk from [aaa], nothing much materialized"
08:46
<annevk>
heh
08:47
<annevk>
There's that skeleton meme...
10:01
<philipj>
annevk: do you have a preference about https://bugzilla.mozilla.org/show_bug.cgi?id=1165851#c41 ? I'm trying to measure something here, but what do you think the most likely end-state of the spec will be?
10:02
<annevk>
philipj: my preference is 1 as well
10:02
<annevk>
philipj: I have good hopes that's fine, smaug____ seems to think otherwise
10:02
<smaug____>
it is just very weird API if it can't be used with SVG
10:03
<philipj>
annevk: ok, so if I'm to measure differences, it's exactly what the spec currently says, plus a "lowercase localName if the document is an HTML document" as part of createAttribute?
10:03
<smaug____>
I'd then make setAttributeNode to throw with SVG elements in html documents
10:03
<philipj>
smaug____: there's still createAttributeNS and setAttribute
10:03
<smaug____>
sure
10:04
<philipj>
smaug____: what end state do you think would be better?
10:04
<smaug____>
but attributes are rarely in any namespace, so I wouldn't expect the API user to remember that *NS should be used with SVG
10:04
<smaug____>
of course no one should use Attr nodes
10:05
<annevk>
smaug____: I don't think making any method throw improves the status quo
10:05
<philipj>
from the code I've found, people are already assuming that case doesn't matter with createAttribute, doing things like createAttribute("frameBorder")
10:05
<annevk>
smaug____: perhaps making setAttribute() more magical to recognize SVG attributes would
10:06
<smaug____>
annevk: well, it would make it clear that if #1 is taken, createAttribute is totally broken with svg
10:06
<smaug____>
(and MathML too)
10:06
<smaug____>
philipj: but case does matter on SVG
10:06
<annevk>
we could rather than lowercase all the things also recognize https://html.spec.whatwg.org/multipage/syntax.html#adjust-svg-attributes
10:06
<annevk>
smaug____: it's not totally broken with SVG
10:06
<philipj>
smaug____: yes, but I found zero cases of SVG attribute together with createAttribute
10:06
<annevk>
smaug____: it works just fine for most SVG usage
10:07
<annevk>
smaug____: even more so for MathML
10:07
<philipj>
annevk, smaug____, how about modifying Attr.localName inside setAttributNode?
10:07
<annevk>
smaug____: there's only one MathML attribute that's not lowercase
10:08
<annevk>
philipj: that's prolly only feasible if Attr's are not really nodes
10:08
<philipj>
at least there's an element around in that case
10:09
<philipj>
annevk: why?
10:09
<smaug____>
the number of non-lowercase doesn't matter, if there is just one. Either the API should work with SVG and MathML, or it shouldn't. The buggy "works occasionally, if you remember to not use it with camelCase attributes" sounds rather horrible to me
10:09
<annevk>
philipj: never mind, might be okay either way, but also somewhat magical
10:09
<smaug____>
philipj: that might work
10:10
<smaug____>
so, preserve case, but when setAttributeNode is called on html element in html document, lowercase?
10:10
<philipj>
annevk: sure, it's a bit odd, but technically it doesn't seem very hard
10:10
<philipj>
smaug____: yes
10:10
<annevk>
I wouldn't really use setAttribute() most of the time anyway... Most attributes have a dedicated getter/setter pair you can use instead
10:13
<philipj>
assuming that no createAttribute() hack works out, would the next best thing be to lowercase input in all other setters and do case-insensitive lookups in getters, removers, and hasAttribute?
10:14
<Ms2ger>
No case-insensitive lookup, please
10:14
<philipj>
Ms2ger: is there another way if a createAttribute() fix doesn't work out? Just trying to figure out what's interesting to measure so we only have to wait for one release cycle
10:15
<annevk>
philipj: yeah, but we should do createAttribute()
10:15
<annevk>
I don't really care if that makes it impractical for some SVG usage, it shouldn't be used anyway
10:15
<philipj>
annevk: well I can get behind that. will it fly in Gecko code review, smaug____?
10:19
<smaug____>
philipj: I don't now understand the proposal
10:20
<philipj>
smaug____: annevk may correct me, but simply to convert localName to lowercase when createAttribute(localName) is called in an HTML document
10:21
<smaug____>
so #1
10:21
<philipj>
yeah
10:21
<smaug____>
which doesn't work with svg
10:21
<philipj>
right
10:21
<philipj>
so will it fly?
10:21
<smaug____>
I'd rather take "preserve case, but when setAttributeNode is called on html element in html document, lowercase"
10:22
<smaug____>
philipj: though, if telemetry shows that Attrs are never used with svg, then #1 would be fine. Horrible, broken API, but easy to implement
10:22
<annevk>
smaug____: that seems like a weirder API since currently names are immutable for all nodes in the DOM
10:24
<philipj>
smaug____: ok, so if the telemtry requested in https://bugzilla.mozilla.org/show_bug.cgi?id=1175031 shows that virtually no calls createAttribute() with non-lowercase input and then attaches that to a non-HTML document, it's OK?
10:24
<philipj>
+one
10:24
<philipj>
non-HTML element I mean
10:24
<smaug____>
at least I'd have hard time to object the behavior then :)
10:25
<smaug____>
philipj: non-HTMLElement
10:25
<smaug____>
not document
10:25
<smaug____>
the worrisome case is svg in HTML document
10:25
<smaug____>
I assume createAttribute won't lowercase ever in non-html documents
10:26
<philipj>
smaug____: right, that would be overkill
10:26
<philipj>
it would be very easy to measure how often setAttribute is called on an SVG element in an HTML document and the attr.localName is not lowercase
10:26
<philipj>
some other ideas would require tracking the source of the Attr object which seems less fun
10:27
<philipj>
or perhaps non-HTMLElement instead of SVGElement to be slightly more conservative
11:11
<annevk>
beverloo: can you tell me where IsIPv4Char is defined within http://src.chromium.org/viewvc/chrome/trunk/src/url/ perhaps?
11:12
<annevk>
Google's URL code is so much more complicated than that of Apple... I wonder what happened
11:12
<beverloo>
https://code.google.com/p/chromium/codesearch#chromium/src/url/url_canon_internal.h&q=IsIPv4Char&sq=package:chromium&type=cs&l=65
11:12
<annevk>
That viewer is so much better!
11:12
<beverloo>
which maps to: https://code.google.com/p/chromium/codesearch#chromium/src/url/url_canon_internal.cc&q=kSharedCharTypeTable&sq=package:chromium&type=cs&l=110
11:13
<beverloo>
It's got a search function too :) https://cs.chromium.org
11:15
<annevk>
So that kind of matches what I expected, but doesn't really explain why 100x.0.0.0 ends up as a normal host name and not as a failure...
11:16
<annevk>
I guess there's some rather arbitrary rulings when something becomes IPv4, "broken", or "not IPv4"...
14:16
<yoav>
Hey! Anyone happens to know why in "<img src width='100*'>" the width is ignored?
14:19
<smaug____>
yoav: because of https://html.spec.whatwg.org/multipage/infrastructure.html#rules-for-parsing-integers
14:21
<yoav>
smaug____: What I see is that * is special cased
14:21
<yoav>
e.g. "200^" is not ignored
14:22
<smaug____>
that is then a bug
14:22
smaug____
checks the implementation
14:22
<yoav>
http://jsbin.com/fonidicohi/edit?html,output
14:22
<yoav>
That's what I see in Blink
14:23
<yoav>
Oh, in Firefox it is not being special cased
14:24
<yoav>
I got a feeling this whole area needs a cleanup
14:24
<yoav>
special cased in WebKit as well
14:27
<smaug____>
yoav: Gecko is doing something against the current spec
14:27
<smaug____>
in your test I see $ and * giving the same behavior
14:27
<smaug____>
both those are ignored
14:28
<smaug____>
yoav: and blink special cases * somehow
14:28
<yoav>
Yeah, in Blink/WebKit, only * drops the entire int
14:28
<yoav>
anything else after the digits is being ignored
14:29
<smaug____>
yoav: any idea why * is such special case?
14:29
smaug____
files a Gecko bug anyhow
14:29
<Ms2ger>
Don't old-style frames use that?
14:29
<yoav>
My question exactly :D
14:30
<yoav>
I think I'll add use counters and try to rip that out if possible
14:30
<yoav>
and '%' too, for that matter (I doubt that would work though)
14:47
<yoav>
Ms2ger: Frames where using it inside their `width` attributes?
15:22
<JonathanNeal>
In the CSS nesting spec, I can use a cartesian product to tie together the parent and child selectors. However, once parent references are involved, I’m lost. Any ideas?
15:31
<annevk>
yoav: you might want to compare with parseInt()
15:35
<yoav>
annevk: I see an explicit special casing for '*' before parsing the integer
15:37
<annevk>
heh
15:41
<yoav>
annevk: It's totally weird. I'll try counting occurrences and kill it with fire if it's not common (and it shouldn't be since it'd have been a compat issue with Firefox)
15:45
<annevk>
I guess it's frame parsing code reused because someone was lazy
15:45
<annevk>
wanderview: so we'll restrict the Cache API on time?
15:45
<annevk>
wanderview: wondering whether or not to ask that on the intent to ship thread...
15:46
annevk
finds https://bugzilla.mozilla.org/show_bug.cgi?id=1175138
15:46
<wanderview>
annevk: I am currently planning to do a partial restriction and then implement the full restriction later... what do you think?
15:46
<annevk>
wanderview: yeah that sounds fine
15:46
<annevk>
wanderview: we need to see if we can restrict SharedWorker somehow
15:46
<wanderview>
chrome is already shipping caches on http I think... although maybe its just in beta
15:46
<wanderview>
annevk: my partial restriction will just be a scheme check to start
15:47
<wanderview>
https:// or localhost (or app: for fxos I guess)
15:48
<annevk>
wanderview: what if SharedWorker created from a secure context could never be accessed from an insecure context?
15:48
<annevk>
wanderview: and only such a SharedWorker gets the goodness?
15:48
<wanderview>
annevk: not sure... seems like we need a spec issue on whatever repo defines SharedWorker
15:48
<annevk>
wanderview: that is, the secure context would just add to the scope of a SharedWorker
15:49
<annevk>
HTML defines SharedWorker, but this needs to be monkey patched first
15:49
<annevk>
I'll file a bug on "POWER" I guess
15:49
<annevk>
Oh, I can add to your issue: https://github.com/w3c/webappsec/issues/406
15:50
<wanderview>
annevk: yea, chrome shipping caches on window in 43... so I believe it allows caches to be used in non-trusted already
15:50
<wanderview>
^shipping^shipped
15:50
<annevk>
bah
15:50
<annevk>
Was sleevi sleeping?
15:50
<annevk>
And Joel? And Chris?
15:51
<wanderview>
annevk: yea... just tried it in the console in chrome release and it worked
15:53
<annevk>
https://github.com/w3c/webappsec/issues/406#issuecomment-113199509 has my proposal for truly secure shared workers
15:56
<yoav>
annevk: Is the bit about frame parsing needing to special case '*' defined somewhere?
15:56
<annevk>
should be
15:56
<annevk>
https://html.spec.whatwg.org/multipage/rendering.html#frames-and-framesets
15:57
<yoav>
Thanks!
15:57
<annevk>
https://html.spec.whatwg.org/multipage/infrastructure.html#rules-for-parsing-a-list-of-dimensions
15:57
<annevk>
That's only used for frames
17:43
<annevk>
Whoa whoa whoa
17:43
<annevk>
"Let mathInt be the mathematical integer value that is represented by Z in radix-R notation, using the letters A-Z and a-z for digits with values 10 through 35."
17:43
<annevk>
That's not very Allen-style...
17:43
<annevk>
But I'm going to steal that as it seems wonderfully short
17:49
<jgraham>
Strictly speaking it doesn't define which digit maps to which symbol
17:51
<annevk>
I would be happy with better language. Was just looking for some inspiration for the IPv4 thingie
18:03
<MikeSmith>
I wonder where TC39 is with progress on value objects http://wiki.ecmascript.org/doku.php?id=strawman:value_objects
18:03
<annevk>
MikeSmith: https://github.com/tc39/ecma262 has a better overview
18:04
<annevk>
MikeSmith: I think you want https://github.com/dslomov-chromium/typed-objects-es7
18:04
<annevk>
MikeSmith: and maybe the SIMD stuff
18:08
<annevk>
Whoa... (0x4334234).toString(33) is undefined. JavaScript is cutting a lot of corners
18:20
<caitp>
to be fair, I'm not sure we actually use base 33 very often in practice
18:22
<caitp>
but wait, no, any radix between 2 and 36 is supposed to do the right thing and not be undefined >_-
18:30
<deltab>
(26231474015353).toString(31)
19:18
<MikeSmith>
annevk: thanks
20:23
<annevk>
https://github.com/WebAssembly/design/pull/195 "You'll have to join the W3C Community Group before we can [correct the link]."
20:23
<annevk>
That's a new level of absurdity right there
20:33
<caitp->
maybe you could just nudge someone whose already in the group?
20:34
<wanderview>
annevk: wow
20:46
<Domenic>
wait it's worse https://github.com/WebAssembly/design/pull/199
20:50
<caitp->
azakai makes a good point though?
20:56
<JoWie>
they might as well lock all issues then