00:03
<Philip`>
Hmm, silly Gentoo
00:03
<Philip`>
I upgraded Opera to 9.25, and now when I click on the Opera icon it launches Opera 9.50 Alpha inside Wine
00:28
<kig>
http://glimr.rubyforge.org/cake/canvasmage.html hihih
00:30
<Philip`>
kig: Doesn't work in Opera 9.2 :-(
00:30
<Philip`>
"Syntax error while loading: line 55 of linked script at http://glimr.rubyforge.org/cake/canvasmage.js : }"
00:31
<Philip`>
(maybe due to a trailing comma in a { ... } list?)
00:32
<kig>
thanks, took it out
00:32
<Philip`>
kig: Works better now - thanks :-)
00:32
<Philip`>
The creature looks pretty broken in Opera, but that's probably just arc bugs
00:33
<kig>
circle strokes with gaps and visible polygon segments?
00:34
<Philip`>
kig: http://tinyurl.com/3y9kp3
00:35
<kig>
dfsdg tinyurl data urls don't quite work in firefox
00:35
<Philip`>
Use Opera ;-)
00:36
<inimino>
with its buggy <canvas>? ;-)
00:36
Philip`
ought to find a better way of uploading small files
00:36
<kig>
wow, that is messed up
00:36
<Philip`>
inimino: It's not really more buggy than anyone else's <canvas> :-)
00:37
<inimino>
Philip`: Perhaps, but when I tried to use <canvas> last the only bugs I hit were in Opera
00:38
<jwalden>
kig: dunno if you have time, but is tinyurl trying to 302 redirect to them or something?
00:38
<kig>
jwalden: it gives 301 with the data: url
00:38
<jwalden>
I seem to recall that intentionally not being supported, sec
00:38
<kig>
which looks funny when trying to wget
00:38
<inimino>
(specifically just arc() stuff, that was about as far as I got)
00:40
<kig>
Philip`: does it look any better now?
00:40
<Philip`>
kig: It looks one pixel high
00:41
<kig>
eh?
00:42
<Philip`>
The example images are about 640x1 and 320x1 and 1200x1, like lines with some yellow/black/white gradient
00:42
<kig>
ghhh
00:42
<Philip`>
(By the way, your .json files aren't really JSON because you'd need to quote the key strings)
00:43
<jwalden>
kig: https://bugzilla.mozilla.org/show_bug.cgi?id=211999#c19 seems to say it's fixed in 3.0, doesn't sound like it'll be fixed for 2.0 unless somebody complains
00:44
<kig>
jwalden: alright
00:44
<Philip`>
jwalden: Can I complain? I like having a tinyurlfs - but then again I like telling people to use Opera so maybe it's good that FF2 doesn't support it ;-)
00:45
<jwalden>
Philip`: complain in the bug if you want :-)
00:45
<jwalden>
redirecting to data URLs isn't especially mainstream, so I'm not super-worried if it doesn't work, to be honest
00:46
<Philip`>
kig: By the way, you could perhaps use ctx.__defineSetter__('fillStyle', ...) in Firefox and Opera 9.5 (and maybe others?) to catch people saying 'ctx.fillStyle = ...', instead of requiring 'ctx.setFillStyle(...)'
00:46
<jwalden>
think Safari added getter/setter support recently, dunno if it's in 3 or not
00:48
<kig>
Philip`: yeah, that'd be nice
00:50
<Philip`>
kig: It's no longer one pixel tall - it looks about the same as it did before
00:50
<Philip`>
s/about/exactly/
00:50
<kig>
one pixel tall was using "draw, insert canvas, remove image"
00:50
<Philip`>
s/exactly/exactly except for a little bit around the beak/
00:51
<kig>
now it's using "draw, img.src = canvas.toDataURL()"
00:51
<Hixie>
acid2 is apparently fixed
00:51
<Philip`>
kig: Hmm, odd
00:51
<kig>
but i'm doing hackery with copying the image style to the canvas style, so that might screw things up
00:54
<kig>
Philip`: try now
00:54
<Philip`>
kig: By the way, the CanvasMage thing is quite neat :-)
00:56
<Philip`>
It looks slightly more broken in Opera, since the red rectangle is now a wide triangle
00:57
<kig>
alright, that's screwed up
00:57
<kig>
i'll have to try and debug on opera tomorrow
00:57
<Philip`>
Oh
00:57
<Philip`>
It's random
00:57
<Philip`>
(It tends to change each time I reload the page)
00:58
<Philip`>
Probably same bug as http://canvex.lazyilluminati.com/misc/operanondet.html which gets rendered as http://canvex.lazyilluminati.com/misc/operanondet.png (but differently each time you reload)
00:59
<kig>
okay, opera gets the wtf award of the day
01:01
<Philip`>
That's fixed in 9.50
01:01
<Philip`>
The whole image looks correct in 9.50, except for some gaps where curves don't join up neatly
01:02
Philip`
would probably suggest giving up on Opera 9.2 because it's far too broken
01:03
<kig>
yeah :/
01:03
<Philip`>
(or emulate the curve-drawing functions with JS)
01:04
<kig>
i thought of breaking circles into four beziers but then i found another shiny bauble
01:04
<Philip`>
Hmm, I thought it was impossible to represent circles as Beziers, but maybe I'm misremembering
01:05
<Philip`>
http://pages.sbcglobal.net/tymmott-smith/Math/Bezier.html - "we cannot get a circle with a Bezier curve" - ah, I wasn't wrong
01:05
<Philip`>
though maybe you can approximate something good enough
01:06
<kig>
flash (iirc) uses quadratic curves for everything
01:06
<Philip`>
Flash doesn't have to worry about interoperability or about passing annoying test cases that try to ensure curves are exactly the right shape ;-)
01:08
<kig>
did some tests with canvasmage animations, but they tend to be a bit big. 300k for 50 frames uncompressed, 30k gzipped
01:09
<kig>
1k procedural
01:10
<Philip`>
kig: What is the advantage of saving images/animations rather than just re-executing the original canvas script every time someone views it?
01:11
<kig>
faster to execute, more evil
01:13
<Philip`>
Maybe you could make the output much smaller by limiting the number of decimal places in all the floats, since those won't compress well
01:13
<kig>
e.g. image/animation based on heavy physics simulation, might take hours to generate, but the command array runs at 60fps
01:15
<kig>
yeah, let's see
01:19
Philip`
doesn't like how he keeps writing <script src=...> and forgetting the </script>
01:21
<kig>
stripping decimals might save 10-20%, but circles get gaps
01:22
<Philip`>
Can you strip fewer decimals?
01:31
<kig>
no visible gaps with four decimals, 10% smaller gzipped
01:32
<kig>
(599b vs 664b, i need a bigger test file)
01:32
<Philip`>
Try recording the output of Canvex ;-)
01:33
<kig>
no image support yet :<
01:43
<kig>
but hmm, maybe it'll work. let's see
01:45
<kig>
would have to replace render_ctx with a RecordingContext and add frame markers to the command stream
01:45
<jruderman>
Philip`: i forget the </script> even for inline scripts
01:45
<jruderman>
Philip`: i wish firefox would warn me instead of just showing a blank page
01:46
<jruderman>
Philip`: and i'm the one who insisted on that behavior :P
09:21
<annevk>
hmm, I didn't even get that bug about version= and baseProfile=
12:01
<salty-horse>
hi. seen this? http://stephenfry.com/blog/?p=32
12:02
<annevk>
"due out soon"
12:02
<annevk>
heh
12:03
<annevk>
seems also that WHATWG is now a synonym for W3C...
12:04
<salty-horse>
you're welcome to comment with corrections
12:06
<Philip`>
"all subscribe to the Working Hypertext Application Technology Working Group mailing list (whatwg.org/mailing-list#specs) and exert pressure" - hmm, that probably won't have the desired effect
12:08
<hsivonen>
ouch. I'm a bit disappointed. I expected more accurate posts on stephenfry.com
12:09
<Lachy>
omg, that's the most innacurate crap I've read in a while
12:09
<Lachy>
since when did Nokia claim any ownership of ogg?
12:10
<Lachy>
and Apple have explicitly stated that they're not pushing their own formats
12:11
<Philip`>
Lachy: What someone explicitly states is not necessarily what they are really doing
12:11
<Philip`>
so saying "Apple doesn't like ..." is not clearly wrong (whereas saying "Apple says it doesn't like ..." would be, since they don't say that)
12:13
<Philip`>
(and saying "Nokia claims ownership of some elements of Ogg" is wrong since they haven't claimed that (as far as I'm aware))
12:45
<kig>
heh, svgtext.svg is 64k, 27k gzipped. svgtext.json is 79k, 22k gzipped
12:55
<hdh>
heh, that stephenfry post also conflates open source and open standard, not a good move
13:03
<Dashiva>
If it's not open source, it's not open at all!
13:05
<hdh>
some business people twitch at open source (see "cancer"); but none can dispute the words "open standard"
13:28
<tndH>
okay, so now guardian readers think html 5 is evil?
13:29
<gsnedders>
tndH: what? in today's? where?
13:29
<tndH>
http://www.guardian.co.uk/technology/2007/dec/22/internet.digitalmedia
13:30
<tndH>
(same as the stephenfry.com link before)
13:31
<hsivonen>
well, the silver lining is that now the WHATWG list is noted on the Guardian
13:31
<gsnedders>
We're in print!
13:32
<hsivonen>
(if you believe that all publicity is eventually good)
13:35
<gsnedders>
Hey, more uninformed emails!
13:44
<tndH>
all i can imagine while reading through the stephen fry article is the QI siren going off and deducting him 10 points
14:01
<salty-horse>
are you going to email the guardian/post a comment?
14:05
<maikmerten>
err....
14:05
<maikmerten>
"The problem is that Nokia claims ownership of some elements of Ogg, which is essentially the file format for the streaming and delivery part of the codec"
14:05
<maikmerten>
that is... not correct
14:06
<maikmerten>
Nokia to my knowledge never made such a claim
14:13
<gsnedders>
salty-horse: the issue is most of the copies of that article will be in print, and already sold.
14:14
<maikmerten>
actually here in Germany HTML5 is in print, too
14:14
<maikmerten>
in Magazin für Computertechnik (c't)
14:14
<maikmerten>
again mostly about <video> and codecs
14:16
<Philip`>
There will eventually be a point at which we have enough bandwidth and processing power to store video losslessly at the maximum resolution the human eye can perceive, so this codec thing is only a temporary problem
14:17
<gsnedders>
Philip`: the total bandwidth needed once we get more and more online would be huge, though
14:18
<maikmerten>
and even then there's the question why anyone would want to be so wasteful with resources
14:18
<salty-horse>
gsnedders, he could issue a clarification in the next column -- it's not the first time a print article was wrong
14:18
<gsnedders>
salty-horse: that's true
14:18
<maikmerten>
video can be compressed by a factor > 10 without anyone noticing anything
14:18
<maikmerten>
(usual factor 100 and bigger)
14:18
gsnedders
is half asleep doing too much
14:19
gsnedders
sigha
14:19
<gsnedders>
*sighs
14:19
<gsnedders>
IE6 is still the most common UA to my site :\
14:19
gsnedders
blames search referrers such as "why does life get complicated when you think you gay"
14:20
<takkaria>
hah
14:20
<gsnedders>
(OK, that only got one, but you get the ideA)
14:20
<gsnedders>
*idea
14:24
<jwalden>
hsivonen: too bad Hixie actually didn't think to make it a task force instead of a working group ;-)
14:25
<gsnedders>
WTF?
14:25
<gsnedders>
(as in WHATTF?)
14:27
<jwalden>
yes, exactly
14:27
<jwalden>
I remember him saying one time that if he'd thought of it he'd have done it :-)
14:30
<hsivonen>
jwalden: well, fantasai has whattf.org
14:30
<takkaria>
I dislike Christmas; people stop posting so often to mailing lists and I find myself reading slashdot instead
14:30
<jwalden>
well-played!
14:44
<hsivonen>
so I'm thinking that I should provide a setting for filtering out foreign namespaces
14:44
<hsivonen>
should it filter out tags or the entire subtree rooted at a dropped element?
14:45
<hsivonen>
I'm leaning towards dropping the entire subtree
14:45
<hsivonen>
(consider foreign metadata elements in <head> or SVG <metadata> and what would happen if only text content were left)
15:54
gsnedders
is amazed that Lachy is actually twittering about snow in Norway
16:35
<annevk>
html5.org got over a 100.000 page requests on 11 December
16:36
<annevk>
normal amount is about 800-1000 a day
16:36
gsnedders
blinks, then realises annevk is doing the European norm of . being a thousand separator
16:39
<Dashiva>
gsnedders: Lachy is new in town :)
16:40
<gsnedders>
Dashiva: I know, it still seems odd.
16:40
gsnedders
wants snow here (where here = east coast of Scotland)
16:40
takkaria
wants snow in the middle of Manchester, UK, but it's not going to happen. :)
16:41
annevk
would like 20 degrees in NL
16:41
<gsnedders>
haven't had proper snow here since 1997
16:41
<annevk>
(Celcius)
16:42
<gsnedders>
we have 275K here
16:43
<gsnedders>
(i.e., 2°C)
22:56
<gsnedders>
oh I do love PHP. randomly using a 1-based array for unpack() :\
22:58
<hdh>
on PHP, I just came across this, http://zestyping.livejournal.com/233348.html
22:58
<gsnedders>
hdh: not taught? it shouldn't exist!
22:59
<hdh>
http://zestyping.livejournal.com/124503.html?thread=690519#t690519 ← is that you ?-P
22:59
<gsnedders>
bo
22:59
<gsnedders>
*no
22:59
<hdh>
:)
23:01
gsnedders
wonders whether to rely on PHP have this bizarre behaviour forever or not
23:03
<Philip`>
Regardless of that particular bizarre behaviour, it's probably quite safe to rely on PHP having some bizarre behaviour forever, and thus use Perl or Python or Ruby or anything else :-)
23:03
<gsnedders>
no, Philip`, the PHP devs claim to not break b/c then break it in some bizarre way that breaks everything :)
23:03
<gsnedders>
(and I don't have choice about programming language, sadly)
23:04
hdh
is still temped towards touching some PHP, the HTML embedness feels nice, like XSLT
23:06
<gsnedders>
hdh: don't.
23:06
<gsnedders>
hdh: As someone who's used PHP for years, I can wholeheartedly say you don't want to use PHP.
23:06
<Philip`>
hdh: There are about three million different HTML templating systems in pretty much every language, so that's no reason to choose PHP :-)
23:07
<hdh>
uhm, ye, I haven't checked any out yet
23:08
hdh
spent some moments looking for a nice [Tt] and [Ff], and found "WƮƑ?", to avoid spiders making his page as non-safe
23:09
<hdh>
s/making/marking/
23:10
gsnedders
wonders how his UTF-32 decoder is more expensive than his UTF-16 one
23:11
<gsnedders>
re-running it makes UTF-32 < UTF-8 < UTF-16 which is still odd
23:11
<gsnedders>
third time I get UTF-32 < UTF-16 < UTF-8 which is the expectation
23:13
Philip`
is currently using Template Toolkit in Perl, but doesn't really like it because it has seemingly-arbitrary syntactic restrictions so e.g. you can't use expressions inside [...]-style list literals
23:15
hdh
got a flash like "hook cheetah to pyblosxom"; a name to hunt down
23:40
Philip`
must remember not to set up web pages that do stuff like <a href="edit/[%message_id]">[% message_title %]</a> when message_title might be an empty string
23:40
<Philip`>
Uh, s/[%message_id]/[% message_id %]/