15:41
<mpt>
Could someone kindly give me an example of something that was in the (WhatWG-era) HTML spec for a long time before implementors decided it wasn’t a good idea?
15:43
<werle>
mpt cc mathiasbynens
15:44
<SteveF>
mpt: command element was around for a while
15:47
<mpt>
That would do nicely, thanks SteveF
15:57
<Domenic_>
yeah the whole <menu> thing has been sad... I wish implementers would implement at least *something* like it.
16:02
<ondras>
I find XUL's <command> and <key> elements really useful
16:10
<SteveF>
Domenic_: firfox has context menus...
16:10
<Domenic_>
SteveF: right, that's true, I guess I meant "everyone" not "implementers"
16:18
<SteveF>
Domenic_:would like to see it widely implemented, but think that styling of menus needs to be very flexible for it to be useful
16:19
<Domenic_>
SteveF: yeah, that's probably true, as much as I would like our designers to suck it up and use platform conventions :P
16:21
<Domenic_>
We did get some start on discussing styling of <select> on, I think it was the whatwg list? Or webapps? But that trailed off... Someone just needs to do the work of writing something down for implementers to comment on, I think. Maybe now that promises is trailing off I can try to make that my next project alongside streams.
16:21
<SteveF>
Domenic_:right like that's gonna happen...
16:22
<Domenic_>
?
16:22
<Domenic_>
oh the designers
16:22
Domenic_
shakes fist
16:27
<wilhelm>
mpt: http://platform.html5.org/history/
16:27
<wilhelm>
Grep for 'dropped'.
16:28
<mpt>
wilhelm, I started doing exactly that when you provided the link. :) Thanks.
16:43
<MikeSmith>
mpt: datagrid is another examplse I guess
16:43
<MikeSmith>
hsivonen: annevk-cloud: fyi https://bugs.webkit.org/show_bug.cgi?id=126375 "Consider using top-level domain name for encoding sniffing"
16:44
<SteveF>
Domenic_:yeah have been following threads on various lists on the subject
16:44
<MikeSmith>
mpt: oh, or what wilhelm said, and what you said after that :-)
16:46
<SteveF>
Domenic_: be great if you could get some momentum happening
16:46
<MikeSmith>
hsivonen: (I see know that ap already had Cc'ed you on that bug)
16:48
<Domenic_>
SteveF: I'll certainly try!
16:59
<Hixie_>
mpt: <command> mostly became <menuitem>, <datagrid> was never really fully fleshed out. a better example might be the repeating model from web forms 2, though i don't recall if that was dropped due to implementation disinterest or if it was dropped due to my realising it sucked.
17:00
<Hixie_>
a lot of the ones labeled as "dropped" in the history page were either replaced by something else or dropped before implementors really looked at it in detail
17:01
<Ms2ger>
Oh, that
17:01
<Ms2ger>
I've written code that used it
20:51
<Hixie_>
jgraham: ping
20:52
<Ms2ger>
You haven't pushed the a.ping change yet?
21:02
<jgraham>
Hixie_: hej hej
21:03
<Hixie_>
jgraham: anolis is 504ing.
21:03
<Hixie_>
Ms2ger: ^
21:04
<Ms2ger>
I see
21:08
<jgraham>
Hixie_: Not sure what I can do about that
21:08
<Hixie_>
me either
21:08
<Hixie_>
do you see anything on your end?
21:08
<Hixie_>
like, in the logs?
21:10
<rektide>
querySelectors not being able to extract attribute values is a :( v.s. XPath
21:12
<rektide>
what is changing in a.ping? i'm writing some code that uses ping right now. i'm pretty sure it's not going to work either.
21:12
<Hixie_>
extract attribute values?
21:12
<rektide>
//a/@href
21:12
<Hixie_>
a[href]
21:12
<rektide>
won't that return a list of nodes with an href attr?
21:13
<Hixie_>
selectors return a boolean
21:13
<Hixie_>
"does this node match this selector"
21:13
<Hixie_>
same as xpath, as far as i'm aware?
21:13
<Hixie_>
but i'm no xpath expert
21:14
<rektide>
querySelector() returns a node list. document.evaluate does too.
21:14
<Hixie_>
querySelector() returns a node, but sure
21:14
<Hixie_>
querySelectorAll() returns a list
21:14
<Hixie_>
the node is the first on the list
21:15
<rektide>
yes, pardon. i'm new to this upstart new guy on the scene.
21:15
<Hixie_>
and the list is just the result of filtering the complete list of elements against the expression
21:15
<Hixie_>
and only including the ones that return true
21:15
<rektide>
xpath let me do more complex aaddressing of things that i wanted to pull out, to select
21:15
<Hixie_>
can you elaborate? what are you trying to do?
21:16
<rektide>
//a/@href. which is different from //a[@href] which is just a predicate
21:16
<jgraham>
Hixie_: Based on previous experience I am rather sure it is part of the server setup on that host and that I can't really configure it. Last time I think we managed to make the process faster
21:16
<Hixie_>
rektide: assume i know nothing about xpath, and tell me what your higher-level problem is
21:16
<Hixie_>
jgraham: :-(
21:16
<Hixie_>
jgraham: k
21:17
<rektide>
i would like to select attribute values
21:17
<Hixie_>
jgraham: i'll wait a bit and see if the load lightens, then :-)
21:17
<Hixie_>
i don't understand waht you mean by "select attribute values"
21:17
<rektide>
i want to write an expression taht queries a document finding things that i want
21:17
<Hixie_>
why do you want to do it? what's your higher-level problem?
21:17
<Hixie_>
e.g. are you looking for a list of URLs that are a document's links point to?
21:18
<rektide>
yes
21:18
<Hixie_>
s/are//
21:18
<Hixie_>
ok you can do that much easier than using selectors, then
21:18
<Hixie_>
just iterate through document.links
21:18
<rektide>
i may not have told you my real problem in the full detail
21:18
<rektide>
a.foo's hrefs
21:19
<Hixie_>
iterate through querySelectorAll('a.foo[href]')
21:19
<Hixie_>
(though what you probably really want is ':link.foo, :visited.foo')
21:20
<Hixie_>
(which xpath can't do :-) )
21:21
<Hixie_>
(':any-link.foo' if you're using newer Selectors)
21:22
<rektide>
i have a feature request
21:23
<rektide>
I'd like ping to be viable for onclick
21:24
<rektide>
so many webapps that i'd like to userscript some auditing into have onclick handlers instead of links, and it kills my ability to use ping there.
21:25
<rektide>
i'll probably build a 1/3 complete hackey barely-working polyfill that let's me ignore this gap in ping
21:25
<rektide>
"If a hyperlink created by an a or area element has a ping attribute, and the user follows the hyperlink, and the value of the element's href attribute can be resolved,..." is overspecific a use case, I'd say, for what kinds of things ought be auditable in the web environment
21:26
<Hixie_>
you could just use a real link with href="javascript:..."
21:26
<Hixie_>
instead of onclick=""
21:26
<rektide>
userscript. it's not my own website i'm adding auditing to.
21:27
<rektide>
and i don't see why there ought to be an audit gap between those two otherwise near-equivalent features
21:27
<rektide>
if they are so interchangable as that, then why is one in-auditable?
21:27
<Hixie_>
i don't understand the use case. can you elaborate? you want to be able to inject script into a page you're visiting to know what link you click on?
21:27
<rektide>
yes
21:28
<Hixie_>
why don't you just tell yourself what link you're clicking on...?
21:28
<rektide>
you are an artful dodger of real problems, lord of the web
21:28
<rektide>
:/
21:29
<rektide>
i thought ping would be a convenient means to audit my activity in an automated fashion
21:29
<rektide>
i thought ping was there for that purpose
21:29
<rektide>
it seems amazing, i love the idea that i can add a tag to report back
21:29
<rektide>
but the things that i can audit are very very very limited
21:29
<rektide>
and i do not understand why auditing wouldn't go further
21:30
<rektide>
i also do not really understand your proposition. here's one example i'd like to instrument with auditing: <div class="arrow down login-required" tabindex="0" aria-label="downvote" role="button" onclick="$(this).vote(r.config.vote_hash, null, event)">
21:31
<rektide>
what do you propose for "tell myself what link i'm clikcing on" for this case, where someone has authored this HTML?
21:32
<Hixie_>
ping="" is primarily there so that sites can work out what link you clicked when leaving the site, since normal auditing methods fail when you click a link to leave the site
21:33
<Hixie_>
rektide: i guess i'd open a spreadsheet where i'm recording this stuff, and write "i clicked on the down arrow" or something? :-)
21:33
<rektide>
that's one very narrow specific conception of a feature
21:33
<Hixie_>
rektide: i don't understand your high-level problem here, so there's very little for me to go on.
21:33
<rektide>
Hixie_: that gets boring after the first 200 mintues spent doing that
21:34
<rektide>
and i'd have to keep opening hte spread sheet where i record stuff, and that takes effort
21:34
<rektide>
are we not human? the tool making species?
21:34
<rektide>
you have my use case! your alternative suggestion is the exact idea of what i want to do!
21:34
<rektide>
i just don't want to have to do it
21:34
<rektide>
i want links to be auditable
21:34
<Hixie_>
why
21:34
<Hixie_>
your use case is at least three levels too specific for a use case :-)
21:35
<rektide>
because it's work and i am better than work and i am human and i can make better things
21:35
<Hixie_>
yikes, it's later than i realised. gotta go get lunch. bbiab.
21:35
<rektide>
what if it's a div.onclick that triggers a move offsite? is that for some reason not auditable?
21:35
<rektide>
why wouldn't ping be allowed to work there?
21:36
<rektide>
you're in the code by definition there, you could instrument your audit yourself in that codebase, but isn't HTML's nice declarative composable structure supposed to clean up so much of this DIY?
21:36
<rektide>
enjoy lunch
21:36
<rektide>
chao
21:44
<Hixie_>
rektide: if you have a <div onclick=""> that triggeres a move off-site, turn it into an <a href=""> :-P
21:44
<Hixie_>
it's true that we can't audit e.g. form submissions
21:44
<Hixie_>
but it's not clear why you'd want to really
21:44
<rektide>
i just think auditing is super valauble, i love the declarative clean ismple form, and i want more of the web to be able to report on itself [doing whatever verb it is doing].
21:45
<Hixie_>
we could develop an API that's just for sending one-off pings, certainly
21:45
<Hixie_>
should probably wait for ping="" to be more widely implemented first, though
21:45
<rektide>
i see absolutely zero reason for embracing a very narrow specific use for what ping is
21:45
<rektide>
i love love love "add your URL to the list, get notified" with a common packet
21:45
<Hixie_>
the general approach is to come up with high-level use cases, then address them. if the use cases are specific, the solution ends up specific.
21:46
<Hixie_>
nobody had a use case for notifying in the cases you're mentioning, so we didn't even consider them.
21:46
<rektide>
i'm probably 1/2 done with my crappy incredibly shitty polyfill
21:47
<rektide>
that i will never maintain or like or enjy, but which will serve this huge gaping need i have
21:47
<rektide>
and it will let me add auditing that seems really basic and sensible to reddit
21:47
<rektide>
without re-engineering the entire web site to support ping's requriements
21:48
<rektide>
i totally understand that i'm adding a new scope of concern, and i have enormous sympathy for being late to the show
21:48
<rektide>
but i also think the spec as it stands is cutting off it's own limbs unnecesarrily
21:48
<rektide>
thanks for talking with me ian
21:56
<rektide>
does anyone know if https://developer.mozilla.org/en-US/docs/Web/API/NodeList#Why_can%27t_I_use_forEach_or_map_on_a_NodeList.3F will ever get better? "Why can't I use forEach or map on a NodeList"?
23:48
gsnedders
wonders how fast Anolis would be on PyPy, even with more stuff having to be done in Python (than in C, in lxml).
23:49
<gsnedders>
(The reason why lxml does that stuff in C is essentially the fact you end up spending ~100% of computation time in dispatch overhead, so PyPy shouldn't suffer)