02:20
<Hixie>
aa: yt?
02:21
<Hixie>
i'm changing executeSql() to be async
02:22
<Hixie>
so you do executeSql('update scores set score=? where user=? and score < ?', score, user, score, function (result) { ... });
02:22
<Hixie>
but what happens when an error occurs?
02:23
<Hixie>
should i call the callback anyway, just with an object representing an error?
02:23
<Hixie>
or should i dispatch an 'error' event to the database?
02:23
<Hixie>
i guess the former is better...
02:23
<om_sleep>
aa: did you see my email suggestion of how to change ResultSet, which proposes adding errorCode and errror fields?
02:24
<othermaciej>
Hixie: sorry, meant that for you
02:24
<Hixie>
assuming you're talking to me, yes, that's why i'm asking
02:24
<othermaciej>
letting the callback handle the error seems easier to work with
02:24
<Hixie>
j
02:24
<Hixie>
k, even
02:24
<othermaciej>
and you'd also want that for the sync version
02:25
<Hixie>
and i guess we just say that if you call executeSql() on the database while in a callback that went wrong, you raise an exception?
02:25
<Hixie>
since that probably indicates you didn't check for errors
02:25
<Hixie>
or do we just start a new transaction in that case
02:26
<othermaciej>
I'm not sure
02:26
<othermaciej>
are there any errors where it ever makes sense to continue the same transaction?
02:26
<othermaciej>
I'd guess probably not
02:26
<othermaciej>
so exception seems ok
02:26
<Hixie>
well you can't continue the same transaction
02:26
<Hixie>
since you've rolled back by then
02:27
<othermaciej>
that's true
02:27
<Hixie>
but you could start a new one
02:27
<Hixie>
if we raise an exception, which i think is the better thing to do for catching errors, it means that the only way to then do a new transaction as part of error handling is to have a timeout
02:27
<Hixie>
to start a new "context"
02:27
<Hixie>
which is ugly
02:27
<othermaciej>
continuing blindly seems like it could give bad results, on the other hand, the timeout thing to retry is ugly
02:27
<Hixie>
yeah
02:27
<Hixie>
exactly
02:28
<Hixie>
hard to know which is better
02:28
<othermaciej>
you could have a closeTransaction() method on Database
02:28
<Hixie>
could do
02:28
<othermaciej>
which isn't needed in the normal case, but could be used for retry in a callback
02:28
<Hixie>
yeah
02:28
<Hixie>
ok
02:28
<Hixie>
will do that
02:59
<Hixie>
i suppose we really want the database version to be set as part of a transaction
02:59
<Hixie>
oh well
03:00
<Hixie>
i guess it doesn't matter
03:01
Hixie
thinks to himself... . o O ( the upgrading app can set it to a secret value while upgrading, then to the new version when done )
03:02
<Hixie>
grr this is going to introduce all kinds of race conditions
03:11
<Hixie>
hm
03:11
<Hixie>
how do we handle a commit failing?
03:12
<Hixie>
we have no way to notify the app
05:51
<Lachy>
access keys don't seem to work in Opera 9.5 alpha, but they do in 9.2
05:53
<Lachy>
pressign Shift+Esc on this page lists all the keys defined, but labels them poorly. http://www.cs.tut.fi/~jkorpela/forms/accesskey.html#ex
05:54
<Lachy>
<input ... accesskey="N"> shows up as: "(N) (Null)" in the menu
05:54
<Lachy>
ah, that's probably because there's no <label> for those controls
05:57
<Lachy>
no, makes no difference whether you define the accesskey on the input or label, it still lists (Null)
06:40
<om_sleep>
Lachy: seems like it will always use the "value"
06:41
<othermaciej>
Lachy: or title
06:43
<Lachy>
I see. That's not very useful
06:44
<othermaciej>
Lachy: it does seem like neither of those is suitable for a text field
06:44
<othermaciej>
but a keyboard shortcut to focus a specific text field seems a little silly in itself
06:45
<othermaciej>
anyway Opera is not really selling me on the potential of accesskey not sucking
06:46
<Lachy>
it's not that silly. It's common for Alt+(key) to focus controls in dialog boxes on Windows
06:46
<Lachy>
the key is typically underlined in the controls label
06:48
<othermaciej>
ok, "silly to me as a mac user"
06:49
<othermaciej>
I think even on windows you don't see that on text fields that aren't part of a dialog box
06:49
<Lachy>
yeah, Macs have a lot of bad UI design ;-)
06:49
<othermaciej>
that's what I've always heard
06:50
<othermaciej>
on mac it's pretty rare for a dialog to have more than a handful of items in the focus cycle so you just use tab
06:51
<Lachy>
well, Alt+D typically focuses the address bar in IE and Firefox, so it doesn't only occur in dialog boxes
06:52
<othermaciej>
on mac Cmd-L is the usual shortcut for that in browsers, but it's kind of an exception
06:52
<Lachy>
Ctrl+L also works in Firefox
06:52
<othermaciej>
(and it has a menu item, so it's a menu shortcut, not a focus shorcut; there's certainly no underlined letter)
06:55
<Lachy>
where's the menu item located for it?
09:07
<Lachy>
one of the problems with the accesskey discussion is that some people don't seem to understand how <command> and <menu> work in HTML5. Perhaps if someone clearly explained that it can be used for toolbar menus or context menus, much like native apps use.
09:59
<othermaciej>
Lachy: yeah, my original proposal assumed too much
10:00
<othermaciej>
Lachy: I mainly just wanted to jot down my idea somewhere that it wouldn't get lost
10:01
<othermaciej>
Lachy: btw Open Location... in the File menu is the menu item with Cmd-L as a shortcut, at least in Safari and Firefox on Mac
10:03
<Lachy>
ok
10:53
<hsivonen>
hmm. dev.w3.org is taking too long to respond. I hope my list reply wasn't silly given the current edit state of the principle doc (which I was unable to check)
11:21
<billyjack>
hsivonen - fwiw, your reply didn't seem silly at all to me
11:21
<billyjack>
if you're referring to the "HTML5 does *not* aim to be backwards compatible with HTML 4" one
11:26
<annevk>
hmm, why is the callback the last argument in executeSql...
11:26
<annevk>
it looks weird because of the arguments... in the middle
11:32
<annevk>
yeah, made sense to me too fwiw
11:33
<hsivonen>
MikeSmith: yeah, that was what I was referring to
11:33
<hsivonen>
MikeSmith: did you see my list of GNU error format issues?
11:35
<hsivonen>
MikeSmith: http://wiki.whatwg.org/wiki/Validator.nu_GNU_Output
11:35
<MikeSmith>
hsivonen - no, haven't seen it yet
11:35
MikeSmith
is looking now
11:41
<MikeSmith>
hsivonen - I agree about the tab issue (should count as one column); as far as UTF-8 in the message, I think there are likely many users in console envrionments who don't have UTF-8 support; as far as how to distinguish classes of messages, I think "Error; " and "Warning; " and "Note; " as a prefix in the message field
11:41
<MikeSmith>
The other questions I don't have any brilliant insights one
11:41
<MikeSmith>
on
11:44
<hsivonen>
MikeSmith: Do Emacs, etc., recognize Error: and Warning; labels?
11:45
<hsivonen>
MikeSmith: that is, if the aim is to make Emacs tell you that you don't have errors, will a Note: or Warning: line be counted as error?
11:46
annevk
reads posts on the alt attribute such as http://whatnitishknows.blogspot.com/2007/09/get-traffic-from-image-serach-engines.html
11:46
<annevk>
"Its the very sweet ‘alt’ attribute which most of us do not use in our Image tags because we want to save our time or mainly because we think its useless. The ‘alt’ attribute is very important to place while writing the tag code for your image as it servers another keyword targeting purpose for image search bot."
11:46
<hsivonen>
MikeSmith: my own use of emacs is *very* superficial, so the exact integration screnarios aren't at all clear to me
11:46
<annevk>
seems that SEO, not authoring tools, might help authors care :)
11:47
<hsivonen>
hendry: any insights on http://wiki.whatwg.org/wiki/Validator.nu_GNU_Output from the vim perspective?
11:49
<MikeSmith>
hsivonen - I think the expectation is that users need to do a certain about of Emacs customization for each time of compiler/checker/whatever they want to be able to have Emacs parse error messages for
11:49
<hsivonen>
MikeSmith: ok
11:50
<MikeSmith>
some other apps, like Vim, have similar features; e.g.:
11:50
<MikeSmith>
http://mail-archives.apache.org/mod_mbox/ant-user/200101.mbox/%3C20010129173134.A24005⊙zo%3E
11:50
<hsivonen>
MikeSmith: should I try to keep the forward-compatible two-level taxonomy that I intend to use for XML and JSON?
11:51
<hsivonen>
MikeSmith: that is, should I do something like Error, Fatal:
11:51
<hsivonen>
so that forward-compatible script can extract "Error" when they see Error, Foobar:
11:51
<MikeSmith>
I would personally prefer that
11:51
<hsivonen>
MikeSmith: ok
11:51
<MikeSmith>
yeah
11:58
<hsivonen>
annevk: it is rather sad how much "blog" content these days is written by humans only to bait users to see Google Ads
11:59
<hsivonen>
annevk: check out tasks on mturk.com and you'll find that poor people can make a cent by participating in astroturfing schemes
12:02
<hsivonen>
mturk.com is already used for transcribing audio. I wonder if it'll be used for outsourcing alt text writing.
12:09
<annevk>
oh yeah, heard about that
12:10
<annevk>
I think marp suggested that something like that might work, but it doesn't really seem to scale to the amount of images that are hosted on Flickr for instance
12:10
<annevk>
or maybe it does...
12:11
annevk
wonders why role= needs namespaced extensibility
12:15
<hsivonen>
annevk: I'm not ethically comfortable with mturk. It feels exploitative.
12:22
annevk
reads an old Gecko bug where canvas.showFallback() is suggested
12:23
<annevk>
see https://bugzilla.mozilla.org/show_bug.cgi?id=291216#c18
12:23
<annevk>
other options: <canvas require-context="3d funky">
13:40
<jwalden>
question: are there any URLs which can be used to link to specific versions of the HTML5 spec?
13:40
jwalden
has a feeling now may not be the best time to ask, PDT-wise
13:40
<krijnh>
jwalden: http://www.whatwg.org/specs/web-apps/current-work/multipage/
13:41
<jwalden>
krijnh: isn't that just the latest version?
13:41
<jwalden>
I think I really meant s/version/revision/ in my question, actually
13:42
<krijnh>
Ah, versions, my eyes told me you said sections
13:42
<krijnh>
Sorry :)
13:42
<jwalden>
fragment identifiers work for that anyway
13:46
<gsnedders>
jwalden: http://dev.w3.org/cvsweb/ (wherever HTML5 is under that, but it's down at the moment)
13:46
<jwalden>
heh
14:45
<hsivonen>
annevk: btw, regarding firehose photostreams and metadata, tantek's photostream is an interesting case study. no sets, no human-entered titles, but, yet, elaborate tagging
14:46
<hsivonen>
jgraham's Flickrs stream is very different
14:46
<hsivonen>
people use these tools very differently
16:36
<Lachy_>
does anyone know where the command attribute is defined? It's mentioned once, but doesn't seem to be defined anywhere http://www.whatwg.org/specs/web-apps/current-work/#the-icon
16:43
<Dashiva>
That paragraph could probably do with significant rewriting
16:46
<Lachy_>
yeah, probably. But I believe the intention is that one can do, e.g. <input type=button command=foo> pointing to <command id=foo ...>, and then the input (and any other elements that reference it) get included in that command elements .triggers DOM attribute.
16:46
<Lachy_>
and when one activates that input, the associated command is triggered.
16:47
<Dashiva>
Sounds sensible
17:14
<jgraham_>
Lachy: As far as I can tell, it isn't
18:33
<Hixie>
othermaciej: i love that you commented only on the part of hte Database APIs that i hadn't changed at all yet
18:36
<Hixie>
and apparently the spec is confuding, but right now you only get a transaction if you call executeSql from within a callback
18:36
<Hixie>
exactly as originaly suggested :-
19:29
<annevk>
hsivonen, "firehose photostreams"? :)
19:41
<hsivonen>
annevk: the kind of photostreams that contain lots and lots of photos. more than what you'd want to type in titles or alternative text for
20:22
<hsivonen>
are there any video sharing sites that output Ogg these days?
20:26
<Philip`>
There's http://blog.blip.tv/blog/2006/05/31/introduction-ogg-theora-support/
20:29
<Philip`>
Looks like they only offer FLV and the original source version for most videos, though
20:31
<Philip`>
but there is e.g. http://blip.tv/file/340551/ with http://blip.tv/file/get/Anvazher-PruebaVideo318.ogg
20:31
<Philip`>
(but the applet doesn't work for me - first it crashed the browser, now it just shows a black box)
20:33
<gsnedders>
hsivonen: Wikipedia uses ogg, but that isn't really a video sharing site. probably the biggest, though
20:51
<Dashiva>
Are the people arguing for merging ul, ol, nl, etc into a single element really serious?
20:59
<gsnedders>
Dashiva: I expect so
21:12
<hsivonen>
Philip`: I just registered to blip.tv and uploaded a file. the only format they autogenerate is .flv
21:12
<hsivonen>
which isn't cool. Apparantly, if I want .ogg or .mp4 I need to generate those myself and upload
21:13
<hsivonen>
which means that blip.tv doesn't really solve the boring and hard part of Web video
21:52
jgraham
notes that even though his flickr stream is not a firehose writing good alternative text has an unfavourable opportunity cost
21:53
<jgraham>
Because the two groups of people I care about viewing the content are: my friends and family (who are all presently sighted) for the purposes of documentary and random strangers for any aesthetic value the pictures might have
21:55
<jgraham>
So the only reason to write descriptions now would be so that the descriptions were available should they be needed by any member of the first group in the future
21:56
<jgraham>
However I don't think the content is valuable enough to spend the time writing descriptions rather than, say, processing more photos
21:58
jgraham
wonders if he will be called anti-accessibility now
22:00
<Dashiva>
There are two groups of those, the "lazy" ones and the "evil" ones
22:00
<Dashiva>
Most of us fall into the lazy group
22:05
<jgraham>
But part of "laziness" is opportunity cost; e.g. the opportunity cost of writing descriptions for photos is that I have less time to spend on producing more photos. The second activity has a small, identifiable, audience (my friends and family), the first has a, presumably smaller, unknown audience
22:06
<jgraham>
Of course the situation would be entirely different if I were producing content with a different audience. A government site should have useful alternative text for all non-decorative images
22:06
<jgraham>
(for example)
22:08
<Dashiva>
Yeah, that's why it's not evil. You aren't doing it to hurt accessibility, it's just a side effect
22:09
<Dashiva>
But objectively, accessibility suffers
22:10
<jgraham>
Sure.
23:32
<Philip`>
http://software.hixie.ch/utilities/js/live-dom-viewer/?%3C!DOCTYPE%20html%3E%0D%0A%3Cbody%20onload%3Dw(document.getElementById('i').width)%3E%0D%0A%3Cimg%20id%3Di%20src%3Dimage%20style%3Dmax-width%3A1em%20width%3D100%3E
23:32
<Philip`>
IE7 makes the DOM controlled by CSS
23:33
<Philip`>
(Opera and Firefox return different things from .width, too)
23:39
Philip`
changes his code so it calculates image sizes on the server then outputs <img _width=...> so the client side can recover the real width, which seems totally horrible