01:42
<TabAtkins>
rektide: Somewhat. Some new APIs will return an Elements object, which is an array subclass, instead of NodeList.
06:05
<hsivonen>
MikeSmith: It's nice that ap considers the idea worthwhile.
06:06
<MikeSmith>
hsivonen: yeah
06:07
<MikeSmith>
it's a good sign. he's pretty careful
06:07
<MikeSmith>
exacting
06:08
<MikeSmith>
fyi for anybody who's waske already (or still), it seems like there's some problem with the w3c network at the moment
06:10
<hsivonen>
annevk-cloud_: why doesn't the big5 encoder in the Encoding Standard have a special case for LATIN CAPITAL LETTER E WITH CIRCUMFLEX AND MACRON etc.? Are those from HKSCS?
06:11
<hsivonen>
I have to wonder what the motivating idea behind Gecko's multibyte decoder abstraction is
06:12
<hsivonen>
is it more performant to avoid the code point math at run time?
06:15
<hsivonen>
annevk-cloud_: why does big5 multiply by 157? that seems weirdly inefficient
06:15
<hsivonen>
annevk-cloud_: is that math representable as shifts but I'm just not seeing it?
06:16
<hsivonen>
(in general, it's a bit annoying that the Encoding Standard doesn't make it obvious, what parts of the math could be done on the ALU)
06:17
<MikeSmith>
hsivonen: btw https://groups.google.com/a/chromium.org/d/msg/blink-dev/zIg2KC7PyH0/Ym28fD2ZIAsJ -- the part where abarth says:
06:17
<MikeSmith>
[[
06:17
<MikeSmith>
My personal view is that XML on the client is going to evolve over the coming years. I expect that browsers will move to a simplified version of XML that doesn't support namespaces or internal DTD subsets because those features introduce vastly more complexity than they're worth. Instead, I expect we'll process XML content using the "foreign content" HTML parsing algorithm, which, for example, is how we parse SVG content in HTML.
06:17
<MikeSmith>
]]
06:18
<hsivonen>
MikeSmith: kinda scary to jus take the HTML parsing algorithm instead of taking XML5/XML-ER.
06:18
<hsivonen>
*just
06:18
<MikeSmith>
yeah
06:18
<hsivonen>
maybe I should subscribe to blink-dev
06:19
<MikeSmith>
hsivonen: subscribe and just read the threads that adam posts to :-) those are usually the ones most worth reading
10:08
<SimonSapin>
gsnedders: It’s not quite ready yet, but https://github.com/SimonSapin/cssselect2 does not depend on lxml anymore
10:42
<hsivonen>
does Python really not have a format code for printing numbers in base 2?
10:44
<hsivonen>
oh, there is bin()
12:15
<jgraham>
Ms2ger: r? https://critic.hoppipolla.co.uk/r/364
12:17
<Ms2ger>
jgraham, which warnings are those?
12:25
<jgraham>
Ms2ger: Some broken cross-references
12:25
<jgraham>
Some minor things
12:26
<jgraham>
See https://readthedocs.org/builds/wptserve/1058788/ if you can
12:26
<Ms2ger>
Hrm, is :ref:`handlers.Python` supposed to be a link?
12:27
<jgraham>
I think readthedocs hasn't updated
12:28
<Ms2ger>
Ah, now it has
12:29
<jgraham>
Yeah, apparently I can't enable the auto-update hook because I'm not special enough
12:29
<jgraham>
(at least according to the W3C)
12:29
<Ms2ger>
So https://critic.hoppipolla.co.uk/showcomment?chain=975
12:30
<Ms2ger>
Do we want to keep null?
12:30
<jgraham>
MikeSmith: I hear you are special
12:30
<MikeSmith>
um
12:31
<jgraham>
Ms2ger: Well. I'm not sure.
12:31
<Ms2ger>
MikeSmith is very special
12:33
<MikeSmith>
special missions subsection junior interim floor manager?
12:33
<jgraham>
MikeSmith: Possibly you can enable the ReadTheDocs webhook for https://github.com/w3c/wptserve. I assume you have to do it through the gh UI, although the rtd documentation isn't super-clear
12:33
<MikeSmith>
ah
12:33
<jgraham>
https://readthedocs.org/projects/wptserve/ since I assume you need it
12:34
<MikeSmith>
github ACLs are ass
12:34
<MikeSmith>
I hope those slackers fix that
12:35
MikeSmith
pusshes the button
12:36
<jgraham>
MikeSmith: Thanks
12:36
<MikeSmith>
jgraham: hmm it doesn't prompt me for a URL though
12:36
<MikeSmith>
*didn't
12:36
MikeSmith
re-checks
12:36
<MikeSmith>
yeah it doesn't let me specify a URL
12:36
<MikeSmith>
"Automatically build documentation hosted on readthedocs.org."
12:36
<MikeSmith>
is all that it says
12:37
<MikeSmith>
I pushed the "Test Hook" button nwo
12:38
<MikeSmith>
..which for most hooks doesn't seem to actually do anything -- it just succeeds or fails, silently
12:38
<MikeSmith>
anyway I've set all that it lets me set from the GH side, afaict
12:38
<jgraham>
MikeSmith: OK. Maybe rtd does magic to work out which docs to rebuild
12:39
<MikeSmith>
yeah it must
12:39
<Ms2ger>
jgraham, I guess keep null for now, then
13:19
<jgraham>
Ms2ger: Thanks
13:19
<Ms2ger>
No, thank you!
13:27
<odinho>
Ms2ger: Why do you want cuddled else? I find it less legible.
13:32
<jgraham>
odinho: Don't argue, I just fixed it :p
13:33
<odinho>
I also did :/
13:33
<odinho>
But you pushed already. But you have one error.
13:34
<jgraham>
(fwiw I think d(global legibility)/d(vertical lines) < 0 even though sometimes d(local legibility)/d(specific line-saving-constructs) > 0)
13:34
<odinho>
1 issue left.
13:35
<odinho>
And review my changes. They are good.
13:35
<odinho>
Only those missing now.
13:52
<jgraham>
odinho: Where does the spec say that * in A-C-A-O doesn't work with preflight?
13:52
<jgraham>
(your change is fine either way)
13:56
<jgraham>
odinho: (and if you can OK my final push I can start on the rebase :|)
13:57
<gsnedders>
hsivonen: %b should give base 2?
13:59
<odinho>
jgraham: It says that somewhere. I've read that spec thousand times. I'm going out now, so can't look up the ref :) Did the ok-ing.
13:59
<hsivonen>
gsnedders: ValueError: unsupported format character 'b' (0x62) at index 1
14:05
<gsnedders>
hsivonen: Oh, so str.format supports b but the % operator doesn't
14:06
<jgraham>
You can do it the nice way but only if you don't also want nice things?
14:23
<jgraham>
So, that was surprisingly easy
14:23
<jgraham>
Oh, and critic did the rebase just like it's supposed to
14:23
<jgraham>
(probably means I missed something)
14:34
<jgraham>
Well the review is accepted
14:35
<jgraham>
I wonder what to do now :)
14:37
<jgraham>
(that is more or less a serious wonder, btw. I could squash and merge, but I don't really trust that everything is going to be working OK and it will break w3c-test.org)
14:48
<Ms2ger>
Post to p-t-infra?
14:51
<Ms2ger>
And land if nobody complains somewhere next week
15:58
<jgraham>
Ms2ger: One more for you
15:59
<Ms2ger>
Which review?
16:00
<jgraham>
r/440
16:02
<Ms2ger>
Ship it
16:04
<jgraham>
Two down, one to go :)
16:43
<SimonSapin>
does the html parser ever emits attributes in a namespace?
16:44
<jgraham>
Yes
16:45
<jgraham>
<svg><a xlink:href="foo">something
16:45
<SimonSapin>
thanks
16:48
<tantek>
because SVG couldn't just re-use <a href> from HTML.
16:49
<tantek>
is anyone working on fixing that so we can deprecate xlink completely?
16:49
<jgraham>
To be fair to SVG they were sold this vision of the glorious XML future
16:49
<Ms2ger>
You? :)
16:49
<jgraham>
Or, possibly, they helped propogate the vision
16:50
<tantek>
jgraham - they were definitely pushing it.
16:51
<tantek>
Ms2ger - I try not to touch svg related things. Usually there's a big undefined mess behind them.
17:07
<Ms2ger>
https://twitter.com/horse_js/status/419144086528868352 narrows down @horse_js's identity quite a bit
17:10
<MikeSmith>
Ms2ger: horse_js is running as a write-in candidate
17:11
<jgraham>
Ah, I assumed that was "running for" as in "running for my dog that died in the spring"
17:13
<Ms2ger>
That's not an expression I've heard before
17:16
<jgraham>
Imagine someone literally running in support or memory of a cause
17:16
<jgraham>
In other news, no innerHTML tests, really?
17:16
<Ms2ger>
A few, in the DOMPS repo
17:17
<jgraham>
Can we migrate them to wpt?
17:17
<Ms2ger>
We certainly can
19:42
<bholley>
Hixie_: see https://bugzilla.mozilla.org/show_bug.cgi?id=956382
19:44
<TabAtkins>
SVG2 is dropping *all* of the attribute namespaces.
19:45
<TabAtkins>
href/etc are just going to be in the null namespace like normal.
20:02
<Hixie_>
bholley: thanks
20:05
<Domenic_>
I'm vaguely curious what APIs use DOMTokenList and not DOMSettableTokenList
20:06
<Hixie_>
whatwg.org/C find-in-page "TokenList"
20:15
<Ms2ger>
Domenic_, relList and classList, for example
20:15
<Ms2ger>
Because rel and className already existed
20:23
<Domenic_>
Ms2ger ah that makes sense, thanks.
21:49
<Ms2ger>
Hrm, why didn't I get jgraham's email.
21:49
<Ms2ger>
?
21:51
<Ms2ger>
"Why is this message in Spam? You previously marked messages from james⊙hcu as spam."
21:51
<Ms2ger>
Oops
21:56
<Hixie_>
heh
21:58
<Ms2ger>
So now I'm reading a month's backlog of jgraham's email
21:59
<Ms2ger>
Gmail kindly deleted everything older than that, of course
22:35
<Hixie_>
Ms2ger: i live in fear of that happening to me
22:35
<Hixie_>
Ms2ger: every now and then i find an e-mail i'm missing, find it in spam, look for others...
22:36
<Hixie_>
Ms2ger: rarely do i find many, but i live in fear of those that i'm missing
22:36
<Hixie_>
Ms2ger: given how much spam i get there's not much i can do about it
22:36
<Ms2ger>
Yeah :/
22:37
<Ms2ger>
At least it was only jgraham here :)
22:37
<Hixie_>
anyone know who sof⊙oc is? (so i can add them to the acks)
22:38
<Ms2ger>
Sigbj�rn Finne, says google
23:03
<Hixie_>
so... where are Map and Set defined. ho dee hum.
23:07
<Hixie_>
what does " Subclass constructors that intend to inherit the specified Map behaviour must include a super call to Map." mean, in practice?
23:07
<Hixie_>
is that a real conformance criteria of some sort?
23:07
<Domenic_>
that if you do class X extends Map { constructor() { blah blah } }
23:07
<Hixie_>
(http://people.mozilla.org/~jorendorff/es6-draft.html#sec-map-constructor)
23:07
<Domenic_>
but blah blah doesn't include super()
23:07
<Domenic_>
then none of your methods will work right
23:07
<Hixie_>
so it's just a statement of fact, not a conformance criteria?
23:08
<Domenic_>
Yeah
23:08
<Hixie_>
i guess the JS spec doesn't mention RFC2119
23:08
<Domenic_>
Huh yeah
23:08
<Domenic_>
Never noticed that
23:08
<Hixie_>
that's... confusing. but ok.
23:09
<Hixie_>
ah, i see.
23:09
Hixie_
reads the conformance section
23:09
<Hixie_>
it's a CSS style spec. "conforming implementations do what the spec says"
23:09
<Hixie_>
ok
23:09
<jgraham>
Google is right about sof btw
23:10
<Hixie_>
thanks
23:23
<Hixie_>
lordy, the js spec's definition of Map is non-trivial to grok
23:32
<odinho>
sof is awesome
23:32
<TabAtkins>
Hixie_: Yeah, it is. What are you trying to grok?
23:36
<Hixie_>
TabAtkins: how to write teh structured clone spec for them. i think i have it, though.
23:49
<Hixie_>
ok, here's a question
23:49
<Hixie_>
are the keys in a Map own properties of the Map?
23:50
<Domenic_>
no
23:50
<Domenic_>
because they can be objects
23:50
<Domenic_>
they are not properties at all
23:50
<Domenic_>
they are stored in the internal [[MapData]]
23:50
<Hixie_>
ah ok excellent
23:50
<Hixie_>
that means i can copy the own properties with impunity
23:52
<Hixie_>
i wonder what alex thinks of Map and Set, given how they are essentially black magic that can't be explained by the platform primitives
23:52
<Domenic_>
How so?
23:53
<Hixie_>
there's no way to use an object as a value other than in these features
23:53
<Hixie_>
since JS has no pointers
23:53
<Hixie_>
or pointer analogue
23:53
<Domenic_>
Um
23:53
<Hixie_>
no?
23:53
<TabAtkins>
???
23:54
<TabAtkins>
Really don't know what you're talking about, Hixie.
23:54
<Hixie_>
like, you couldn't write a polyfill that implements Map or Set without using Map or Set
23:54
<Domenic_>
There's a polyfill for Map on MDN I believe
23:54
<TabAtkins>
Yes you can. It doesn't have the same performance characteristics, though.
23:54
<TabAtkins>
(Linear time gets rather than constant time.)
23:54
<Hixie_>
if it's not the same perf characteristics, it's not compliant :-)
23:54
<Domenic_>
You can do constant time gets using trickery
23:54
<TabAtkins>
Domenic_: Really?
23:55
<Domenic_>
TabAtkins: yeah, the old "put a hidden property on it" trick.
23:55
<Domenic_>
where you override Object.getOwnPropertyNames etc. to not show that property
23:55
<Domenic_>
Doesn't really count though
23:55
<TabAtkins>
I don't understand how it would work.
23:55
<TabAtkins>
With an arbitrary object key.
23:56
<Domenic_>
You put a property on that key
23:56
<Hixie_>
Domenic_: interesting idea
23:56
<TabAtkins>
Oh!
23:56
<TabAtkins>
Tag the object with a ref to the map.
23:56
<TabAtkins>
Or a ref to the value.
23:56
<Hixie_>
or just an ID
23:56
<Hixie_>
that you can then use as a key
23:56
<TabAtkins>
Hixie_: Yeah, you can gensym a key for it.
23:57
<TabAtkins>
Attaching the value to the key directly gives you the GC characteristics you want, I think.
23:57
<Hixie_>
anyway, seems weird that we'd be adding these somewhat elaborate features, rather than just features that would let you implement them simply enough
23:58
<Hixie_>
if we had some sort of reference type, a pointer analogue, that could stringify, etc, it seems like it would make all this less magic
23:58
<TabAtkins>
Note that Extensible Web is never about "rather than". We want the primitives, but we *also* want the sugar. Relying on libraries to do all the sugaring is a failure mode to the economics of the network.
23:58
<TabAtkins>
But yes, there's a hidden primitive in Map.