00:00
<zewt>
heh, nothing's quite as much fun as pages that disable submit buttons, then reloading it and having it stay disables
00:00
<zewt>
i use GM scripts to disable that nonsense on a page or two
00:03
<smaug____>
Hixie: are you using any script libraries (which may do random stuff)
00:43
<Hixie>
bummer, he left
00:43
<Hixie>
no script libraries
04:40
<kennyluck>
hmm… didn't notice that IE and FF aren't doing the UTF8 with error handling fully, e.g. for error happening at the end of streams.
10:58
<annevk>
kennyluck: end of stream is a drama all over the place
12:34
<Ms2ger>
http://webbeep.it/spec.html
12:36
<erlehmann>
Ms2ger, better https://github.com/erlehmann/libglitch/blob/master/FORMAT-draft-erlehmann
12:38
<erlehmann>
Ms2ger, web beeps is probably very useful for 90ies hacker movies
12:39
<erlehmann>
>By restricting these to a pentatonic scale, arbitrary combinations of notes could be used without major dissonance.
12:39
<erlehmann>
clever
13:19
Philip`
always likes diagrams with spellchecker underlines in them
15:05
<annevk>
what's a good name for the encoding data table / mapping table / thingie?
15:05
<annevk>
index? mapping?
15:06
<annevk>
naming sucks
15:08
<Philip`>
Lookup table?
15:09
<annevk>
and URL lookup-table-iso-8859-2.txt ?
15:15
<zewt>
simply "map"?
15:17
<annevk>
"an encoding can have map"
15:18
<zewt>
?
15:18
<zewt>
"can have a map"
15:19
<annevk>
euh yeah
15:19
<annevk>
although some can have more than one I suppose
15:36
<Ms2ger>
Almost a year since Google announced they were dropping support for H.264
15:38
<izhak>
The lack of a good modular html parser implementation on c/c++ is really really oppressive.
15:39
<izhak>
As the DOM C/C++ implementaion lack is.
15:39
<annevk>
so fix it :)
15:41
<izhak>
annevk: I'm trying:)
15:41
<izhak>
Lot's of people trying I suppose.
15:42
<annevk>
I haven't seen much actually apart from browser implementations and implementations in JavaScript
15:42
<izhak>
Well, trying never meant success.
15:43
<izhak>
I think the next step in Web is not possible without having bringing modularity to scene
15:50
<Philip`>
izhak: What do you consider "modular" to be?
15:51
<Philip`>
(e.g. is the parser okay if it forces you to use its own DOM implementation?)
15:53
Philip`
isn't about to suggest an implementation, he's just curious as to what's desired
15:54
<izhak>
Philip`: By modular I mean particular parts of browsers to be done as libraries with stable interfaces
15:54
<izhak>
Say, DOM, HTML forgiving parser
15:57
<Philip`>
Parsers in browsers have to support scripting (and document.write etc) which I imagine makes the interfaces far more complex and messy - would it be better to just have a non-scripting parser with a simpler API?
15:57
<izhak>
UI and Object Model diversion
15:58
<Philip`>
(A non-scripting parser would be useless when implementing a parser, but is presumably what's needed for almost every other instance of HTML parsing)
15:58
<Philip`>
s/implementing a parser/implementing a browser/
15:58
<annevk>
Ms2ger: I'm going to retire the unicode- files I generated earlier in favor of these new files
15:58
<Ms2ger>
Alright
15:58
<izhak>
non-script parser becomes more and more useless today
16:00
<izhak>
As I know even crowlers (why even? - neatly) run scripts today
16:02
<smaug____>
and since you need scripting, designing api for modular parser is hard
16:02
<smaug____>
If you need a C++ parser, I'd take hsivonen's parser and translate it to C++
16:03
<smaug____>
(using some automatic tool)
16:09
<webben>
Philip`: scripting parsers are likely better for a lot of scraping applications.
16:09
<Philip`>
izhak: It wouldn't be feasible for them to run scripts without implementing pretty much an entire web browser (in which case they'd presumably automate a WebKit instead of writing their own code)
16:10
<webben>
yes, hence ChromeDriver and Phantom.JS.
16:10
<webben>
*PhantomJS
16:10
<Philip`>
izhak: (In my testing some time ago, Googlebot dealt with scripts just by looking inside the script text for strings that looked like absolute or relative URLs and then crawling those, without attempting to actually run any scripts)
16:11
<Philip`>
(and a non-scripting parser should be sufficient for that)
16:12
<izhak>
Philip`: That's the issue of UI and Object Model mess
16:12
<izhak>
They should be decoupled
16:12
<izhak>
Just parsing out the links from scripts is not sufficient
16:16
<Philip`>
We can't do anything about the layering-violation mess that browsers have to deal with, so I think my hypothetical question is whether there'd be more interest in a necessarily messy API that can deal with that mess, or a nice clean modular API that ignores it but is probably still good enough for any use case that doesn't involve building an entire browser
16:21
<izhak>
The semantics which is expected to be found on Web in an increasing quantity and quality requires an access to the API of different layers today, opposite to the Past when only the resulted web page was an object of interest.
16:22
<izhak>
So the clean API to that modules I think will be demanded more and more
16:54
<annevk>
Ms2ger: http://dvcs.w3.org/hg/encoding/raw-file/tip/map-iso-8859-3.txt
16:55
<Ms2ger>
Nice, thanks
16:55
<annevk>
since most of section 5 is non-normative now maybe that should be nuked, but it's somewhat valuable I guess
16:56
<annevk>
well you can always look at http://dvcs.w3.org/hg/encoding/raw-file/tip/single-octet-research.html I suppose
17:10
<annevk>
any objections to removing 5.1 - 5.28?
17:28
<annevk>
done
17:33
<annevk>
http://dvcs.w3.org/hg/encoding/raw-file/tip/Overview.html
17:44
<annevk>
I never realized the only multi-octet encodings were from CJK
17:52
<annevk>
"""
17:52
<annevk>
And here's the table they used:
17:52
<annevk>
Big5 to Unicode(codepoint):
17:52
<annevk>
http://moztw.org/docs/big5/table/moz18-b2u.txt
17:52
<annevk>
Unicode(codepoint) to Big5:
17:52
<annevk>
http://moztw.org/docs/big5/table/moz18-u2b.txt
17:52
<annevk>
"""
17:52
<annevk>
the second file has more lines than the first...
17:59
<annevk>
smaug____: btw, before you get surprised, WebKit does not default whatToShow to SHOW_ALL; since WebKit's default was 0 we decided on a better one
19:04
<annevk>
so Chrome still understands a fair bit of iso-2022-cn (which is btw another hopeless mess), it just doesn't use any data tables for it
19:11
<bga_>
gsnedders are you here?
19:14
<gsnedders>
bga_: yes
19:14
<bga_>
. /([@!])(.*?)$/.exec('12@1@!12@3')
19:14
<gsnedders>
Known bug.
19:15
<bga_>
i want non gredy backward search
19:15
<gsnedders>
I think.
19:15
<gsnedders>
If that's what I think.
19:15
<gsnedders>
Oh, wait, no it isn't.
19:15
<bga_>
but its greedy
19:15
<gsnedders>
bga_: So the last @ or ! and everything that follows?
19:16
<bga_>
yeah
19:16
<bga_>
last, not first
19:17
<gsnedders>
/(@[^@]*|![^!]*)$/ will be quicker than anything more complex.
19:17
<bga_>
hm
19:18
<bga_>
. /(.)([^@!]*?)$/.exec('12@1@!12@3')
19:18
<bga_>
thanks
19:18
<bga_>
not general solution but works
19:19
<gsnedders>
Not sure why that isn't greedy off hand.
19:19
<gsnedders>
Oh, duh
19:19
<bga_>
i reversed string before and it worked too, but full string reverse is slow
19:20
<gsnedders>
It matches the [@!] first. Then it backtracks when it has to for the .*
19:20
<gsnedders>
But it never has to backtrack for that, as everything matches it and the following $.
19:23
<erlehmann>
i wonder why people do that https://plus.google.com/105786194532858868645/posts
19:24
<erlehmann>
after all the chrome team has repeatedly shown they do hate the open web
19:24
<erlehmann>
err
19:24
<erlehmann>
the google+ team
19:24
<erlehmann>
i am of sleepy today
19:24
<erlehmann>
sorry to the chrome team ;)
19:24
<gsnedders>
Yeah, I should go get a more modern browser for Google+.
19:25
<erlehmann>
gsnedders, when you use opera, do they show an upgrade message?
20:08
<gsnedders>
erlehmann: Certainly I've seen it several times, yeah
20:08
<erlehmann>
gsnedders, you better do no evil … or else!
20:09
<erlehmann>
„evil“ meaning running a non-approved browser ^_^
20:09
<gsnedders>
Well, Opera is finally slowly getting into Google's automated testing
20:12
<smaug____>
(It feels to me that Google is testing stuff more and more only using Chrome)
22:01
<GPHemsley>
annevk: Are you aware that the '**' and 'Issue:' overlap in your Encoding files (and others?)?
22:02
<zewt>
what the
22:02
<zewt>
cool, chrome is crashing on initial load, heh
22:03
<zewt>
wonder if they pushed a bad update or something
22:07
<annevk>
GPHemsley: ooh, I guess that's because of the WHATWG style sheet
22:10
<annevk>
GPHemsley: fixed for http://dvcs.w3.org/hg/encoding/raw-file/tip/Overview.html
22:10
<GPHemsley>
cool
22:11
<GPHemsley>
annevk: Oh, you just removed the 'Issue' part?
22:11
<annevk>
euh yeah?
22:11
<GPHemsley>
because apparently they were marking two different things
22:11
<GPHemsley>
(not everything had both)
22:12
<annevk>
oh that's because I only applied Issue: to <p>
22:12
<annevk>
and not <ol>
22:12
<annevk>
but it's all the same
22:12
<GPHemsley>
ah, OK
22:12
<annevk>
something that needs better text (or text at all)
22:12
<GPHemsley>
right
22:33
<annevk>
so are we back to no longer discussing new features on some list? https://bugzilla.mozilla.org/show_bug.cgi?id=697132
22:39
<smaug____>
annevk: that is mainly about the b2g thing
22:40
<smaug____>
so, very experimental stuff atm
22:40
<bga_>
hm
22:40
<bga_>
document.elementFromPoint({x: 0, y: 0})
22:40
<bga_>
is it standard feature?
22:41
<smaug____>
elementFromPoint is, but the parameter isn't
22:41
<bga_>
but chrome support object :)
22:41
<bga_>
ok
22:41
<bga_>
thanks
22:42
<smaug____>
huh
22:45
<bga_>
also
22:46
<bga_>
style.width = '1m'
22:46
<bga_>
is it meter or mile?
22:46
<zewt>
one light minute
22:47
<bga_>
chome gives me 968px in offsetWidth
22:47
<bga_>
but if i calc 100cm - its 3700px
22:48
<bga_>
hm