00:01
<jacobolus>
also, max(R, G, B) - min(R, G, B) is an extremely bad proxy for the perceived attribute chroma (which is similar to saturation, but the two have different technical definitions)
00:02
<zewt>
(if these don't match PS then I'd probably never use them, because they'd be incompatible with every art asset our artist hands me)
00:03
<jacobolus>
zewt: it's a sad world, right? Adobe can never change their software, because it wouldn't be backward compatible. everyone else has to copy photoshop, or it wouldn't be interoperable
00:03
<jacobolus>
zewt: and so technical decisions that make no sense in 2012 are entirely impossible to change, ever
00:03
<cabanier>
jacobolus: people don't want us to change it. There would be an all-out revolt.
00:03
<zewt>
it makes perfect technical sense, in the context of the real world
00:04
<jacobolus>
it's just sad, because color is a place where these decisions actually negatively impact every user who tries to create images with them
00:04
<cabanier>
jacobolus: we can add things
00:04
<jacobolus>
a lot of technical decisions can be papered over and never seen by the user
00:05
<cabanier>
jacobolus: like I said, with color, with pulled most of the features because everyone hated it and turned it off
00:05
<zewt>
(yet there's been no revolt when photoshop changed to stop allowing the navigator to navigate outside the border of the image, which screwed up a bunch of my usage habits; maybe I can pay a mob to revolt for me)
00:05
<cabanier>
jacobolus: turning it off was harder than the math. A lot of special case code
00:05
<jacobolus>
zewt: oh really? that was one of my favorite photoshop changes of all time :)
00:06
<zewt>
it's horrible; I used it all the time to edit around the edge, now it's impossible to even see the edge of the image that way (since it falls underneat the toolbars)
00:06
<jacobolus>
cabanier: I don't know what specifically you're talking about here
00:06
<zewt>
also underneath
00:06
<jacobolus>
zewt: wait, what?
00:06
<jacobolus>
zewt: dude, press the "F" key
00:07
<zewt>
i work in fullscreen 100% of the time
00:07
<jacobolus>
zewt: sorry, I thought you meant the change the other direction
00:07
<zewt>
did they fix that in cs6 or something?
00:07
<jacobolus>
could you move past the corners in a document in a window, ever?
00:08
<zewt>
definitely ... up until something like cs2
00:08
<jacobolus>
oh, bummer
00:08
<zewt>
(and you still can by manually dragging around, the navigator just clips at the edge)
00:08
<jacobolus>
yeah, that's terrible
00:09
<cabanier>
jacobolus: I'm technically still on the photoshop team. I can ask them why that decision was made.
00:09
<jacobolus>
zewt: for some reason I remember that one of the two fullscreen modes used to not to past the edge. but I could be inventing that in my head
00:09
<jacobolus>
cabanier: doesn't really affect me, but apparently zewt didn't like it :)
00:10
<jacobolus>
cabanier: if you can bug them about exclusion mode in CIELAB images though... :)
00:11
<jacobolus>
cabanier: also, if you're working on this css compositing spec, you should add 'linear light' mode if it can be easily done
00:11
<cabanier>
jacobolus: I will do so. there must be a reason that they grayed it out
00:11
<cabanier>
jacobolus: ask for it on www-style.
00:11
<jacobolus>
cabanier: the reason is that it's not thought to be meaningful for A/B channels, since they never get close to the extremes
00:11
<jacobolus>
and so combining two arbitrary images ends up with uninteresting looking results
00:12
<cabanier>
have to go...
00:12
<zewt>
run, run, run
00:12
<jacobolus>
cabanier: but exclusion mode is a building block, not a tool to be used alone
00:12
<jacobolus>
cheers
00:12
<jacobolus>
cabanier: anyway, linear light mode. pretty much my favorite blend mode after "normal"
00:15
<jacobolus>
gavinc: you're right, it's the Rec 601 primaries (NTSC) used for this formula, now that I think back about it
00:15
<gavinc>
yay, so it wasn't a choice made in the 90s! It's a choice made in the 70s!
00:16
<jacobolus>
I don't think there was any software image compositing in the 70s
00:17
<jacobolus>
still a choice made in the 90s :)
00:17
<jacobolus>
well, actually of course there was software image compositing in the 70s
00:17
<jacobolus>
but I don't think with these particular "blend modes" anyhow
00:17
<jacobolus>
anyway, I gotta run too
00:18
<jacobolus>
Hixie: sorry to crud up your channel there for a bit
00:18
<jacobolus>
:)
04:44
<MikeSmith>
Hixie: I think there was at least one person other than hsivonen who wasn't happy with the proposed <template> parsing
04:45
MikeSmith
goes to look back at some threads
04:45
<Hixie>
if it's me it doesn't count :-)
04:45
<MikeSmith>
hahaha
04:58
<MikeSmith>
Hixie: not exactly an answer to your question, but I think Scott Gonzalez questioned whether we need a <template> element at all, or need to be trying to do this through markup
04:58
<Hixie>
url?
04:58
<Hixie>
if he has an alternative solution that is certainly a good thing to look at
05:00
<MikeSmith>
Hixie: http://lists.w3.org/Archives/Public/public-webapps/2012AprJun/0294.html
05:01
<MikeSmith>
he's scott_gonzalez on IRC
05:01
<MikeSmith>
dunno what timezone he's in
05:01
<MikeSmith>
maybe US/West or Central
05:01
<Hixie>
i think those points have been pretty fully explained by dimitry and company
05:01
<MikeSmith>
OK
05:02
<MikeSmith>
yeah I remember ojan replying
05:02
<MikeSmith>
anyway I guess Scott is the other person I was thinking of
05:03
<Hixie>
k
05:04
<Hixie>
<template> has bubbled its way to the top of my queue again but i'm not sure what to do since it seems a bit bad to go behind hsivonen's back on this
05:04
<Hixie>
especially after just having gone a way hsivonen didn't really want with the alt attribute thread
05:04
<Hixie>
though at least in that case it was just a naming thing
05:05
<Hixie>
this one is reather more... fundamental
05:07
<MikeSmith>
Hixie: I expect hsivonen will be back soon
05:08
<MikeSmith>
seems like he's been away for 3 weeks or so already
05:09
<Hixie>
that's what i thought a few days ago, which is why i had waited on the alt thing :-)
05:30
<zcorpan>
work on this until he's back :-) https://www.w3.org/Bugs/Public/buglist.cgi?query_format=advanced&short_desc_type=anywordssubstr&short_desc=%3Ctrack%3E+webvtt&longdesc_type=allwordssubstr&longdesc=&bug_file_loc_type=allwordssubstr&bug_file_loc=&status_whiteboard_type=allwordssubstr&status_whiteboard=&keywords_type=allwords&keywords=&bug_status=NEW&bug_status=ASSIGNED&bug_status=REOPENED&emailtype1=substring&email1=&emailtype2=substring&email2
05:30
<zcorpan>
=&bug_id_type=anyexact&bug_id=&votes=&chfieldfrom=&chfieldto=Now&chfieldvalue=&cmdtype=doit&order=Reuse+same+sort+as+last+time&field0-0-0=noop&type0-0-0=noop&value0-0-0=
05:54
<jacobolus>
cabanier: hey Rik. By the way, I wanted to say that even though I got a bit ranty earlier, I do appreciate the addition of more kinds of image compositing to CSS. It should enable some very cool stuff. :)
05:55
<jacobolus>
my annoyance at the "luminosity", etc. blend modes is with their technical details (which fall short of what they could be), not with their concept (which is a great one)
05:56
<jacobolus>
and even just directly copying photoshop's behavior, though perhaps not ideal, is definitely better than not having such features
06:24
<jgraham>
Hixie: It you are looking for things to do I certianly have reported bugs that I would like fixes for ;)
06:26
<jgraham>
In related news, I got an apparently working script last night so I will finish it up when I get to the office. Let me know whatever I need to interact with your end (if you didn't already, I didn't check email)
06:27
<Hixie>
jgraham: i sent you mail
06:27
<jgraham>
I was thinking of providing a URL that would respond to GETs that would cause an immediate update of the data (with some rate limiting to protect bugzilla)
06:28
<jgraham>
For long values of immediate (i.e. it would actually do the update async and call back to update your end)
06:28
<Hixie>
when would i call it?
06:28
<jgraham>
After someone submits a bug
06:28
<Hixie>
sure, i can do that
06:28
<Hixie>
not all bugs go through my script though
06:29
<jgraham>
Sure, but it would allow a slower update frequency whilst still getting mostly-fresh data
06:29
<jgraham>
Anyway, need to get ready to leave for the office now if I want to take the bus
06:30
<Hixie>
k
06:30
<Hixie>
if you want to do that (which is fine by me) reply to that e-mail and i'll hook in tomorrow at work
06:30
<jgraham>
Sure
06:41
<rniwa>
jezz... people are still talking about longdesc :(
06:42
<zcorpan>
longdesc! longdesc for img, longdesc for iframe, longdesc for video! everyone gets a longdesc!
06:42
<rniwa>
zcorpan: i propose we add longdesc element and add longdesc content attribute on that.
06:43
<jgraham>
rniwa: In web standards, the definition of "n00b" is "has endured less than half a decade of longdesc flamewars"
06:44
<rniwa>
jgraham: i had been following the w3c standards for a while but i had never been aware of longdesc flamewars :(
06:44
<rniwa>
jgraham: probably because i had avoided joining mailing lists
06:44
<jgraham>
rniwa: n00b
06:45
<jgraham>
:)
06:45
<rniwa>
jgraham: i must say i'm quite amazed that people participating in that discussion can make living...
06:47
<rniwa>
jgraham: although... on the other hand, the definition of standards n00b might be to consider W3C as too bureaucratic.
06:48
<rniwa>
jgraham: W3C is nothing like IETF or ISO.
06:59
<zcorpan>
https://www.w3.org/Bugs/Public/show_bug.cgi?id=17273#c6 so use case A, i'm pondering about what an api would look like and have trouble with naming
07:00
<zcorpan>
my idea is: video.XYZ.add(0, 90, 100, 100); would occupy the bottom 10% of the video (so cues are pushed upwards) and video.XYZ.clear(); would clear the areas
07:01
<zcorpan>
having a property XYZ instead of putting the methods on video directly my thinking is that it would be easier to extend in the future and maybe easier to work with if you save the object as a variable
07:02
<zcorpan>
anyone have suggestions for the name?
07:12
<zcorpan>
video.viewport.occupy(...) maybe?
07:33
<odinho>
occupy video movement? :P
07:34
<jgraham>
video.viewport.occupy(99%)?
07:34
<odinho>
jgraham: Oh man, too good :D :D
08:14
<kennyluck>
So there's no longer WHATWG weekly….
08:15
<kennyluck>
For what it's worth, I do appreciate Anne's work. I can see how boring writing such summaries is…
08:18
<Ms2ger>
Do I hear a volunteer? :)
08:18
<jgraham>
Pretty sure I heard one
08:22
<odinho>
Cool, nice that it'll continue kennyluck!
08:23
<odinho>
Props to you
08:23
<kennyluck>
Noooooooo
09:07
<zcorpan>
kennyluck++
10:32
<jgraham>
Did someone say status markers on bugs?
10:32
<jgraham>
(seems to be broken in Opera though :( )
10:33
<zcorpan>
:(
10:38
<jgraham>
It is broken for other people, right?
10:38
<jgraham>
I mean the whole status thing is missing?
10:40
<zcorpan>
i see status boxes in 12.01
10:40
<jgraham>
Hmm I have -next
10:41
<jgraham>
But it worked on the spec I had loaded before
10:46
<Ms2ger>
jgraham, there's no way to change the spec section after the fact, I guess?
10:46
<jgraham>
Ms2ger: It just uses the URL field at the moment
10:46
<jgraham>
So editing that ought to work
10:46
<jgraham>
I didn't try though
10:46
<Ms2ger>
How often does it update?
10:47
<jgraham>
At the moment? Never
10:47
<Ms2ger>
That's not a lot
10:47
<jgraham>
But I will set up a cron job and give Hixie a hook to trigger an update after a bug is posted
10:49
<Ms2ger>
Oh, heh, the bugs filed on #head end up attached to the TOC on the multipage parts
10:49
<zcorpan>
jgraham: this is awesome. thanks
10:49
<Ms2ger>
jgraham++
10:51
<gsnedders>
Why is there no clear documentation of what is stable in ES6 drafts and what is likely to change? >_>
10:52
<scott_gonzalez>
Hixie MikeSmith: I can read through the <template> discussions again (and my timezone is ET)
10:53
<MikeSmith>
scott_gonzalez: I remember that Ojan at least had specifically responded to you with some rationale about why markup was needed
10:55
<scott_gonzalez>
MikeSmith: Is the main benefit having a native solution for Web Components?
10:55
<MikeSmith>
yes
10:55
<MikeSmith>
I guess that's not clear from the rest of the thread
10:55
<MikeSmith>
but that in fact is what this is needed for
10:56
<MikeSmith>
I don't know that anybody has any other use in mind for it other than that
10:57
<MikeSmith>
I remember an hsivonen question in that thread that indicated he wasn't aware of that either
10:57
<scott_gonzalez>
I guess that seems fine. I'm not sure it's really a big win to have a declarative way to define loops and the such.
10:57
<scott_gonzalez>
Do we expect components that need logic and don't need JS?
10:58
<scott_gonzalez>
I should probably read the Web Components spec again.
10:58
<scott_gonzalez>
I only read it once and it was a while ago.
11:00
<MikeSmith>
scott_gonzalez: I don't think we'd expect components that need logic but don't need JS, no.
11:00
<MikeSmith>
But I'm not expert on those specs either
11:00
<MikeSmith>
dglazkov would be the person to ask
11:00
<MikeSmith>
he's in US/West I think
11:00
<scott_gonzalez>
ok
11:01
<scott_gonzalez>
I'm juts having a hard time thinking about how a component would make use of nested templates and looping without having to do a bunch of logic in JS anyway.
11:01
<scott_gonzalez>
s/juts/just/
11:01
<MikeSmith>
right
11:02
<MikeSmith>
I would think it still need the JS logic too but I could well be wrong about that
11:02
<MikeSmith>
there is no purely declarative way to do this as far as I can see
11:02
<scott_gonzalez>
I have an item on my todo list to convert a jQuery UI widget into a web component.
11:03
<scott_gonzalez>
But I won't be able to work on that for at least 1-3 weeks.
11:03
<MikeSmith>
that would be a good exercise for sure
11:03
<scott_gonzalez>
We're in crunch time for a release.
11:03
<MikeSmith>
OK
11:03
<MikeSmith>
we should really try to have another face-to-face meeting somewhere early next year to talk about Web Components
11:04
<MikeSmith>
I'm sure we will be having a discussion at the WebApps WG meeting in November
11:04
<MikeSmith>
but that will be in France and I think a lot of people will not be traveling to it
11:05
<scott_gonzalez>
Do you know where the next one after that will be?
11:06
<MikeSmith>
scott_gonzalez: if we do in fact end up doing it, most likely in Mountain View or nearby I guess
11:07
<scott_gonzalez>
I can probably attend that.
11:07
<MikeSmith>
it will just be a matter of seeing if there's enough interest and then somebody taking responsibility for planning it
11:08
<MikeSmith>
I guess it's probably something we'll talk about at the WebApps meeting in November
11:08
<scott_gonzalez>
ok
11:11
<deane>
MikeSmith: Hi, what mailing list do bugs from this component go to? https://www.w3.org/Bugs/Public/enter_bug.cgi?product=Validator%20%28Nu%29
11:11
<MikeSmith>
hey deane
11:12
<MikeSmith>
no mailing list, maybe
11:12
<MikeSmith>
wait no, www-validator-cvs⊙wo
11:12
<MikeSmith>
see the QA Contact field
11:12
<deane>
I see
11:12
<MikeSmith>
deane: they also go to mike+validator⊙wo
11:13
<MikeSmith>
so you can just set a watch on that address if you want
11:13
<deane>
I see
11:14
<deane>
so there will be double ups with .nu's bugzilla then? Two bugzillas for one validator
11:15
<deane>
How come it doesn't have the text field and file upload functionality?
11:15
<MikeSmith>
it does have those
11:16
<MikeSmith>
and yeah I guess there could be redundant bugs
11:16
<MikeSmith>
but those a easy to deal with
11:17
<MikeSmith>
deane: the select menu at http://validator.w3.org/nu/ has "Address", "File Upload", and "Text Field"
11:17
<MikeSmith>
it requires that you have JS enabled
11:18
<deane>
Yeah, just noticed that :)
11:20
<deane>
I wish hsivonen had a mailing list for .nu's bugzilla. I think it's only you and him that get the notifications.
11:30
<zcorpan>
deane: add hsivonen to your watch list. http://bugzilla.validator.nu/userprefs.cgi?tab=email
11:31
<deane>
zcorpan: thanks, I'll set that up.
11:50
<zcorpan>
jgraham: would it be hard to put the bug summary in the link title=""?
11:52
<jgraham>
zcorpan: I was thinking that too
11:52
<jgraham>
Easy on my end at least
11:53
<jgraham>
It would require a different wire format and stuff, but nothing that would be difficult to change I think
12:05
<zcorpan>
would be nice for sure
12:14
<jgraham>
I am thinking we should be able to do something similar for tests (maybe a link per section to tests for that section, genertated from whatever manifest data people have added) and that the other stuff that's currently in the status markers isn't that useful
12:14
<jgraham>
Would be nice if the implementation status could be outsourced to caniuse.com
12:15
<zcorpan>
yeah
12:16
<zcorpan>
and finally, it would be nice to have all this for other specs, too (and i want a pony) :-)
12:17
<jgraham>
I might fix up some code to scrape which tests apply to each section in the next few days
12:19
<jgraham>
Could map caniuse.com data to section ids (for at least top level sections) and get some implementation status data from there
12:19
<jgraham>
https://github.com/Fyrd/caniuse/blob/master/data.json
12:38
<zcorpan>
jgraham: so what happens when a bug is no longer open, does it leave the section box lying around or does it remove it if there's no other data in it?
12:38
<jgraham>
zcorpan: The box will just end up in the state it would have been in if there had never been a bug
12:38
<jgraham>
But Hixie did that whole end
12:39
<zcorpan>
ok
12:39
<jgraham>
My role in the enterprise is just scraper of data
12:39
<jgraham>
(for small values of "scrape" since it is XML and CSV rather than HTML)
12:40
<odinho>
jgraham: That would be ace, in fact. Using caniuse api.
12:40
<odinho>
s/api/datha
13:56
<smaug____>
no rniwa
13:56
<smaug____>
I wonder in which time zone he is in?
13:57
<Ms2ger>
jp?
14:06
<jgraham>
Indeed, I was under the impression he works in the Googleplex in MV
14:07
<smaug____>
Ms2ger: hey, another thing. Since anne is apparently away, do you happen to know how stable prepend()/append() etc are in DOM4
14:07
<smaug____>
after/before will change sure
14:07
<jarek>
is there somewhere a JSON version of this table? http://www.w3.org/TR/SVG/attindex.html
14:07
<Ms2ger>
I haven't seen feedback on it for ages, so I assume either stable or ignored :)
14:08
<jgraham>
Not ignored
14:08
<Ms2ger>
Must be stable, then
14:08
<Ms2ger>
jgraham, you're denying to comment on whether or not you're implementing? :)
14:09
<jgraham>
Ms2ger: No comment on whether or not I am commenting :p
14:09
<Ms2ger>
Dammit :)
14:10
<smaug____>
somewhere between ignored and stable then..
14:51
<gsnedders>
Ms2ger: Oh come on, you know we're implementing it, along with everything else in HTML5/DOM4/CSS3/XSLT2/$otherSpecHere, because how else would we do it first!?
14:52
<gsnedders>
We implement stuff within days of it getting specced!
14:52
<gsnedders>
We just have rather long times to market. :P
14:52
<gsnedders>
s/./ at times./
15:33
<Hixie>
jgraham: yeah adding titles seems like a great thing to do
15:33
<Hixie>
jgraham: i can set something up when i get to the office
15:33
<Hixie>
would be a slightly bigger pain on my end but nothing unmanageable
15:34
<Hixie>
jgraham: doing the tests too would be great, that's actually already supported
15:34
<Hixie>
jgraham: you'd just have to plug into the existing API for updating section markers
15:35
<Hixie>
i looked into doing implementation status from caniuse at some point but that was gonna be more than trivial so i didn't bother
15:35
<Hixie>
i'd love to have that automatic too though
16:51
<odinho>
Hixie: What is needed? A mapping from section to the data? Maybe the upstream data could even get that in.
17:22
<Hixie>
odinho: yeah
17:26
<Hixie>
jgraham: yt?
17:33
<Hixie>
jgraham: i've changed the wire format and database
17:35
<smaug____>
still no rniwa
17:36
<Hixie>
jgraham: updated the front-end, too
17:40
<TabAtkins>
smaug____: rniwa works in our SF office.
17:49
<Ms2ger>
I've seen rniwa on the list
17:52
<TabAtkins>
zcorpan: Someone just suggested that inline CAS + parser-inserted elements could do the adjustments synchronously. I don't immediately see any problems with this - it's just adjusting the set of attributes attached to the element during building, essentially.
17:55
<TabAtkins>
Well, minor problem I suppose - if the inline CAS is late in the document, it'll only apply synchronously to elements *after* it in the stream, I guess. This might be confusing.
17:55
<TabAtkins>
However, there's no reason at all to put your CAS in late - linked CAS is automatically async, and inlined CAS doesn't need to wait for any elements to load, like JS might.
18:30
smaug____
wonders how to deal with this webkit limitation "can't implement this and that because that would leak"
18:30
<smaug____>
that affects heavily to APIs
18:30
<smaug____>
and it is very odd limitation
18:31
<smaug____>
rniwa: FYI, I doubt Gecko would implement undomanager per page approach
18:35
<Hixie>
man i hate how you can't 'transition' from max-height:0 to max-height:auto
18:42
<Hixie>
Ms2ger: are there specific urgent bugs you need me to look at?
18:42
<Hixie>
Ms2ger: (if so, mark them "critical")
18:42
<Ms2ger>
Don't think so
18:42
<Hixie>
k
18:42
<Ms2ger>
Actually, you fixed one of them yesterday :)
18:43
<Hixie>
the only bugs i remember fixing yesterday were typos that i was fixing while watching tv :-)
18:43
<Ms2ger>
The Attr bug
18:43
<Hixie>
oh right
18:43
<Hixie>
close enough to a typo
18:43
<Ms2ger>
You should watch more tv, it seems to help you get work done :)
18:44
<Hixie>
heh
18:44
<Hixie>
i've been doing lots of stuff recently
18:44
<Hixie>
rewrote the whoel ruby section :-)
18:44
<Hixie>
that was like days' worth of work
18:44
<rniwa>
smaug____: i'm not suggesting that either.
18:45
<rniwa>
smaug____: i specifically avoided spec'ing how undo inside a text field works
18:45
<rniwa>
smaug____: given that there are different needs from different vendors
18:45
<Ms2ger>
Hixie, well, I don't care about ruby ;)
18:45
<rniwa>
smaug____: we probably need to make undoscope content attribute optional
18:46
<rniwa>
smaug____: alternatively, we can get rid of "undo()" and "redo()" from undo manager API
18:46
<smaug____>
rniwa: optional in which sense?
18:46
<rniwa>
smaug____: and the definition of active undo manager platform dependent
18:46
<rniwa>
smaug____: that some browsers won't support it
18:46
<smaug____>
rniwa: no optional features in APIs
18:47
<rniwa>
smaug____: then, we need to get rid of undo() and redo() from undoManager.
18:47
<rniwa>
smaug____: i mean... i don't have to use the term optional
18:47
<rniwa>
smaug____: i can just make it not do anything on browsers that don't support multiple undo managers per document.
18:47
<rniwa>
smaug____: the thing is... we can have multiple undo managers per document, and that's final.
18:47
<rniwa>
smaug____: there's nothing we can do about it.
18:47
<Hixie>
ok i poked at the status boxes' styles a bit
18:47
<Hixie>
hopefully y'all think they look prettier now
18:47
<smaug____>
we isn't the idea to give web developers to make whatever kind undo handling they want
18:48
<smaug____>
page level or field level
18:48
<rniwa>
smaug____: no.
18:48
<hober>
undo behavior should match the local platform convention
18:48
<rniwa>
smaug____: the idea of undo manager is to let browsers know the existence of undo stack in the page
18:49
<rniwa>
smaug____: if they just wanted to do whatever the heck they want, just add a random entries to undo manager
18:49
<rniwa>
smaug____: and then mantain your own undo manager
18:49
<rniwa>
smaug____: then you can do whatever the hell you want.
18:49
<Hixie>
btw are there any opera people around who have an opinion on <template>?
18:49
<rniwa>
smaug____: and i don't intend stop you from doing that.
18:50
<rniwa>
smaug____: all we're asking is to let us not violate platform conventions and let us make our decision as to what can be implemented and what cannot be implemented in our engine.
18:50
<Hixie>
also i plan to post http://wiki.whatwg.org/wiki/What_you_can_do to alistapart pretty soon so if anyone sees anything wrong with it, please let me know asap
18:51
<rniwa>
smaug____: i'm totally fine and respectful of the fact gecko (and opera) chose to have a separate undo manager per text field
18:51
<smaug____>
rniwa: yeah, in practice it is "this is hard to implement in webkit, so lets do a dummy API"
18:51
<rniwa>
smaug____: and i don't intend to comprose that either.
18:51
<rniwa>
smaug____: no.
18:51
<Ms2ger>
Hixie, hmm, I don't see bug annotations, am I just missing them?
18:51
<rniwa>
smaug____: it's not just hard. it's impossible.
18:51
<smaug____>
rniwa: what I'm even more worried that since webkit can't handle certain kinds of APIs, that will lead to worse APIs also elsewhere
18:52
<Hixie>
Ms2ger: they're all gone except in the #introduction box for now
18:52
<smaug____>
rniwa: it is hard, not impossible
18:52
<Hixie>
Ms2ger: i'm waiting for jgraham to update his end to send bug titles
18:52
<rniwa>
smaug____: it is impossible in practice
18:52
<smaug____>
rniwa: Gecko had similar problems
18:52
<rniwa>
smaug____: we're not going to adopt Gecko's approach
18:52
<smaug____>
but then we implemented cycle collector
18:52
<rniwa>
smaug____: nor are we willing to fix that problem.
18:52
<smaug____>
other options are also possible
18:52
<Ms2ger>
Hixie, oh, right
18:52
<smaug____>
like to gc for everything
18:52
<rniwa>
smaug____: but we're not going to do that.
18:52
<rniwa>
smaug____: we have discussed this.
18:52
<smaug____>
that is your choice
18:53
<rniwa>
smaug____: and that's our conclusion.
18:53
<rniwa>
smaug____: yes
18:53
<rniwa>
smaug____: and i'm asking you to respect that.
18:53
<rniwa>
smaug____: the matter of fact is that we're going to veto the spec anyway if we kept the spec as is.
18:53
<smaug____>
rniwa: well, I'm worried that web APIs will be less than optimal because one major browser engine can't handle certain basic things
18:54
<rniwa>
smaug____: i don't consider this as "basic things"
18:54
<smaug____>
rniwa: at least would be great to have some documentation what all webkit can't handle
18:54
<smaug____>
so that we could try to avoid such constructs in APIs
18:54
<rniwa>
smaug____: maybe.
18:55
<rniwa>
smaug____: it was my fault. i should have chekced our how our object model works earlier
18:55
<smaug____>
rniwa: if you have a GCed language like JS, and you design APIs for it, it is quite natural to expect that the underlying implementation can handle cycles
18:55
<rniwa>
smaug____: we can handle cycles in javascript objects
18:55
<Hixie>
fwiw, when there's a disagreement like this, at the end of the day, if you can't come to agreement, the way to solve it is to implement what you think is best and then ship it early enough that your implementation gets more traction than the other
18:56
<Hixie>
and then the "losing" side gets to implement the other API or lose web compat
18:56
<rniwa>
Hixie: are you talking to us?
18:56
<Hixie>
yes :-)
18:56
<rniwa>
Hixie: yeah.
18:56
<rniwa>
Hixie: my current plan is create a custom build of chromium with this feature
18:56
<smaug____>
Hixie: well, so far webkit devs have made pretty clear they won't accept any certain kinds of APIs, which cause cycles in C++
18:57
<rniwa>
Hixie: and let develpoers play with it.
18:57
<Hixie>
smaug____: if you just implement that kind of API and get it widely adopted, you'll force the webkit devs to implement such an API anyway, and then your problem is solved :-)
18:57
<Hixie>
(presumably at great cost to the webkit project)
18:58
<smaug____>
(I don't know why it is so great cost when all the other engines have the solution)
18:58
<Hixie>
(disclaimer: i have no idea what exactly we're talking about here in terms of the undomanager api, so i have no opinion on the actual issue)
18:58
<rniwa>
Hixie: i mean... we're not going to implement it anyway.
18:58
<rniwa>
Hixie: it's not a matter of whether we work hard or not.
18:58
<Hixie>
rniwa: if all the other browsers implement it, you'd end up implementing it
18:58
<rniwa>
Hixie: not really.
18:58
<rniwa>
Hixie: we can choose not to implement it :)
18:59
<Hixie>
you can chose to lose lots of market share :-)
18:59
<rniwa>
Hixie: just like WebGL isn't implemented by IE
18:59
<Hixie>
yeah, and we'll see how long they manage to hold out
18:59
<smaug____>
Hixie: as far as I've understood webkit devs say they won't implement certain kinds of APIs. It is not quite clear to me what all cause problems for them
18:59
<smaug____>
apparently even MutationObserver is hard (and leaky atm)
19:00
<rniwa>
smaug____: the problem is that our JS engine uses garbage collection but all C++ objects are ref-counted
19:00
<smaug____>
yes
19:00
<smaug____>
Gecko works the same way
19:00
<smaug____>
JS is GCed and C++ refcounted
19:01
<smaug____>
(but we have cycle collector to kill the cycles )
19:02
<rniwa>
smaug____: https://docs.google.com/document/d/1uYHpq7u5Sslj54UgzXjA7pYR53XjidpBcrCa-neOGQs/edit?pli=1
19:02
<rniwa>
smaug____: this explains how nodes are managed in webkit
19:03
<Hixie>
rniwa: for the record, "we can't implement that" is a bit of a weak argument given that we're talking about software. i mean, you can implement it. not wanting to is a different matter.
19:03
<Hixie>
rniwa: (again, i've no knowledge of the precise issue here, i'm just talking in general terms)
19:03
<rniwa>
Hixie: well, if we were to implement this, we might as well as write our engine from scratch
19:03
<Hixie>
mozilla did do that once
19:04
<rniwa>
Hixie: and we're not going to do that.
19:04
<rniwa>
Hixie: so it's impossible in practice
19:04
<smaug____>
cycle collector was added to gc+refcounted engine
19:04
<smaug____>
it was a bit painful yes
19:04
<smaug____>
but it is not impossible in practice
19:05
<Hixie>
"we don't want to do that" is a different argument than "it's impossible". i'm just saying you'll get much better reactions from other vendors if you just say "we don't want to" than if you claim that something is impossible, especially if they have done it.
19:05
<rniwa>
smaug____: the problem is that cycle collector will regress the performance will introduce a significant complexity to the code base.
19:05
<rniwa>
Hixie: sure. i guess it's a wording issue :/
19:07
<Hixie>
(from a competitive point of view, it seems gecko would be well positioned to introduce a widely-used api that forced you to take that hit to remain relevant, so it seems wise for webkit to get the alternative API used widely before gecko does theirs :-) )
19:07
<rniwa>
Hixie: that's why we're already implementing it :)
19:08
smaug____
needs to design some awesome new API which is all about cycles :)
19:09
<Hixie>
rniwa: then you run the opposite risk, namely pissing off other vendors because you're forcing what they consider a bad api down their throat... it seems you're screwed either way :-)
19:09
<rniwa>
Hixie: it's okay :)
19:09
<rniwa>
Hixie: i'm used to pissing other ppl off
19:09
<rniwa>
Hixie: that's my way life :P
19:09
<rniwa>
way of*
19:13
<rniwa>
Hixie: at the end of the day, there are things we don't do. like we'll never implement XBL2.0
19:14
<Hixie>
if firefox and IE both used it and Amazon, CNN, and eBay all depended on it, you would.
19:14
<Hixie>
s/used/implemented/
19:14
<rniwa>
Hixie: maybe.
19:15
<Hixie>
come now
19:15
<Hixie>
there's no "maybe" there
19:15
<rniwa>
Hixie: or maybe we'll just lose the market share because we decide not to implement it.
19:15
<Hixie>
uh huh
19:16
<Hixie>
hober: aren't bugs supposed to get some boilerplate when they're closed? (https://www.w3.org/Bugs/Public/show_bug.cgi?id=16793)
19:30
<cabanier>
jacobolus: I got some info from the photoshop engineers.
19:31
<cabanier>
jacobolus: exclusion doesn't give intuitive or useful results in Lab mode. The blending modes are meant to be visually pleasing, not mathematically correct
19:32
<hober>
Hixie: yes
19:32
<hober>
Hixie: looks like jay doesn't know that
19:32
<hober>
Hixie: i will bug him
19:32
<cabanier>
jacobolus: people who want to emulate the math behind exclusion can do it themselves
19:53
<Hixie>
hober: k
20:19
<jacobolus>
cabanier: the reason that's unsatisfying as answers is (1) exclusion mode never gives visually pleasing results, in any color space; it's purely for tricky special effects or as an intermediate step in something else, and really only useful to someone who knows what's going on, and the “usefulness” is identical in RGB or Lab, (2) it's not possible to "emulate" the result in any obvious way, except by through a bunch
20:19
<jacobolus>
of explicit manual steps, or by doing some math in an external tool, or similar. There's no way to do it that is anywhere near so useful when you're actively working in photoshop and exclusion mode would enable some truly awesome possibilities
20:21
<jacobolus>
also (3) no one but an expert is using Lab mode anyhow, because none of the tools are very well optimized for it. This is yet another limitation that makes it less pleasant than RGB, and this time an entirely arbitrary one
20:23
<jacobolus>
but anyway, oh well. I came to accept that it wouldn't happen several years ago. not worth worrying too much about
21:35
<Hixie>
who's css3 ui's editor currently?
21:35
<Hixie>
or selectors, i guess
21:35
<Hixie>
let me rephrase
21:36
<Hixie>
who is in charge of the :read-only and :read-write selectors?
21:37
<Hixie>
no tab, no tantek, no ms2ger
21:37
Hixie
checks his calendar to make sure he's not missing some event or something
21:42
<hober>
tantek's editing css3 ui, and fantasai is editing selectors 4
21:47
<Hixie>
k
21:53
<Hixie>
heycam: i got some [AllowAny] DOMString arguments, do I just drop [AllowAny] ?
21:54
<heycam>
Hixie, yes DOMString arguments should always be preferred when an argument doesn't match the other overload types
21:54
<Hixie>
ok
22:03
<Hixie>
heycam: so there's no difference between an attribute that's "attribute any foo;" and one where it's "attribute (DOMString | SpecialFoo) foo;" except that in the latter case things like coaxing to DOMString if the input is null or { valueOf: ... } etc are handled by Web IDL?
22:04
<Hixie>
er, "or", not "|"
22:06
<Hixie>
heycam: and if so, is there a way for me to define that i want an attribute that on setting always coaxes the input to DOMString but on getting could return "any"?
22:26
<heycam>
Hixie, for the first question, yes that's right; for the second, unfortunately there isn't
22:27
<Hixie>
k
22:27
<Hixie>
is there an equivalent for [AllowAny] for "long" ?
22:27
<heycam>
no, but if there is no DOMString argument overloaded with it it will select the long
22:27
<Hixie>
i have a method where UAs treat integers one way, and everything else 0
22:28
<Hixie>
do i have to make it take any?
22:28
<heycam>
I don't think so, let me check though
22:28
<Hixie>
(HTMLOptionCollection.remove())
22:28
<heycam>
so you want f("abc") and f(node) to be like f(0)?
22:29
<Hixie>
yeah
22:29
<Hixie>
and f(NaN)
22:29
<heycam>
ok so I *think* it might be like that currently but I'll just confirm]
22:29
<Hixie>
and f(0.5)
22:29
heycam
always forgets what the latest state of all the overload resolution stuff is
22:29
<Hixie>
but f(60.5) treated as f(60) apparently
22:29
<heycam>
always truncating?
22:30
<Hixie>
i've tested four numbers so far
22:30
<Hixie>
they all seemed to truncate
22:30
<Hixie>
but what do i know
22:30
<heycam>
ha
22:30
<Hixie>
seems like it truncates for all the numbers i care about
22:31
<Hixie>
-0.1 and -0.9 => 0
22:31
<heycam>
ok so that should be right, if you don't have an exact match then it'll prefer a DOMString argument, and if there's no DOMString argument it'll prefer a primitive argument
22:31
<heycam>
so it's then just down to the normal rules for converting to long
22:31
<Hixie>
ah cool
22:31
<heycam>
if they match what you need..
22:31
<Hixie>
so if i have a method that takes a long it'll never throw TYPE_MISMATCH?
22:31
<Hixie>
or whatever hte exception is
22:31
<heycam>
(TypeError)
22:31
<heycam>
right
22:32
<Hixie>
how convenient
22:32
<Hixie>
ok
22:32
<Hixie>
cool
22:32
<Hixie>
thanks
22:32
<heycam>
cool
22:35
<Hixie>
it astounds me how good a job we are doing of speccing the web platform these days, compared to where we were 13 years ago
22:35
<Hixie>
(web idl being a huge part of that)
22:36
<heycam>
hooray
22:44
<Yuhong>
http://www.w3.org/2001/tag/2011/12/evolution/
22:44
<Yuhong>
http://www.w3.org/wiki/Evolution
22:44
<Yuhong>
I wonder what happened
22:47
<Hixie>
hober: your copy/paste resolutions are cute :-P
22:54
<Hixie>
hober: is there any documentation anywhere that tracks which revisions the w3c spec has adopted and which it hasn't?
22:59
<Hixie>
hober: also, is there something somewhere i can use to keep track of the decisions that were applied? I'm trying to update the list of ways the specs are forked
23:11
<Yuhong>
DirectX vs OpenGL again, now against CSS Shaders: http://codeflow.org/entries/2012/aug/22/css-shaders-w3c-microsoft-and-broken-standards/
23:12
<jamesr>
the rant is strong with this one
23:13
<zewt>
yeah i closed the window as soon as i saw every other word bolded with some randomly in weird colors
23:15
<zewt>
(the distracting animations and mugshot didn't help on the "worth reading" scale, either)
23:26
<Yuhong>
https://news.ycombinator.com/item?id=4422022
23:38
<Yuhong>
jamesr: But considering it is MS fragmenting the web yet again, I am not surprised.
23:38
<jamesr>
i think the article is really misinformed