05:47
<zcorpan>
dsheets could maybe use <svg><svript> although that has a different set of quirks
05:47
<zcorpan>
i didn't follow what his use case was
05:48
<SamB>
zcorpan: he seemed intent on doing the "polyglot" thing without any explanation
05:49
<zcorpan>
he doesn't need different parsing of script to do polyglot...
05:50
<SamB>
anyway he wasn't terribly coherent :-(
05:52
<zcorpan>
if his problem was that <script><![CDATA[ doesn't work then the fix is js comments
05:53
<zcorpan>
or stop using polyglot of course
07:40
<annevk>
zcorpan: I think I used http://www.fileformat.info/convert/image/svg2raster.htm or some such
07:40
<annevk>
zcorpan: might not give the best results
07:40
<annevk>
SamB was working on something better
08:20
<annevk>
JakeA: note that get() takes a dictionary argument for filtering, defaulting to all
08:21
<JakeA>
annevk: I guess you can only have one notification per tag?
08:22
<annevk>
JakeA: there can be multiple
08:23
<JakeA>
annevk: Still feels like it should be getAll then, but it's not a big deal. Eg we have cache.match and cache.matchAll, both methods take a filtering arg
08:24
<annevk>
fair, querySelector/querySelectorAll
08:24
<annevk>
but I don't think there's really a case here where you'd want the first so from that perspective...
08:26
<JakeA>
annevk: also map.get
08:26
<JakeA>
But yeah, if Firefox already has an impl, that seems like a stronger reason to leave it as .get
08:41
<annevk>
I wonder why Mark Watson of Netflix is pushing back on requiring HTTPS for new APIs
08:41
<annevk>
http://lists.w3.org/Archives/Public/public-webappsec/2014Aug/0062.html
08:42
<annevk>
I guess it's mostly because of https://www.w3.org/Bugs/Public/show_bug.cgi?id=26332
08:42
<annevk>
But Netflix goes over TLS so what is the big deal?
08:42
<annevk>
Perhaps the TV environment where they care less about security?
08:42
<annevk>
hsivonen: ^^
11:16
<annevk>
foolip: Firefox throws if you try to modify an attribute's child list
11:17
<annevk>
foolip: and in fact it pretends there are no children
11:18
<annevk>
foolip: e.g. attr.firstChild returns null, childList is an empty NodeList
11:26
<foolip>
annevk: that sounds so sweet
11:27
<foolip>
I bet there's a FIXME that fortunately never got fixed :)
11:36
<zcorpan>
does anyone use http://resources.whatwg.org/status-warning.js ?
11:39
<annevk>
foolip: I put a summary in https://www.w3.org/Bugs/Public/show_bug.cgi?id=26609
11:40
<annevk>
Domenic: could we deploy HSTS on resources.whatwg.org?
11:46
<annevk>
jgraham: "In addition, wildcards themselves can have subjectAltName extensions, including other wildcards. For example: The wildcard certificate *.wikipedia.org has *.m.wikimedia.org as an Subject Alternative Name." seems to be what we want?
11:46
<annevk>
jgraham: that's 360 USD a year?
11:49
<jgraham>
annevk: Yeah, I think a SAN cert would work
11:49
<jgraham>
But IANASSLE
11:49
<annevk>
jgraham: well wildcard + SAN right?
11:50
<jgraham>
Yes
12:09
<Ms2ger>
foolip, actually, not a FIXME... We used to support that, kinda, but I killed it
12:28
<gsnedders>
what's the correct behaviour of vh on paged media?
12:33
<JakeA>
Bothers me that the quota API doesn't really specify the units of the quota
12:34
<JakeA>
I'm guessing bytes. The 'usage' property is in bytes.
12:45
<foolip>
Ms2ger: in that case you're the hero of the day!
12:47
<Ms2ger>
\o/
13:02
<foolip>
Ms2ger: was that change simply making Attr not a container Node like Element, Document and DocumentFragment?
13:05
<Ms2ger>
https://bugzilla.mozilla.org/show_bug.cgi?id=737122
13:08
<annevk>
jgraham: per https://www.startssl.com/ it seems we can do this for USD 60
13:08
<annevk>
if Hixie_ had an account with them already it might not be too bad
13:09
<annevk>
EV is getting cheaper too, I thought someone recently told me it was 10k or so
13:10
<foolip>
Ms2ger: thanks
13:45
<jgraham>
annevk: EV doesn't work with wildcard certs anyway
13:45
<annevk>
jgraham: yeah, wasn't really considering EV for us, it's mostly a scam
13:58
<zcorpan>
https://github.com/whatwg/html-mirror/commit/7c23a0789467f3c3d9659b75db963e31e21c9d76 is gold
14:06
<foolip>
zcorpan: I like that one too :)
15:00
<annevk>
JakeA: perhaps we should float the Fetch part on the WHATWG list?
15:01
<annevk>
JakeA: I can do that tomorrow
15:02
<JakeA>
annevk: I'm happy to do it, will post a summary tomorrow morning
15:02
<annevk>
sgtm
15:09
<annevk>
Domenic: if only we had someone to maintain IDL :p
15:09
<Domenic>
annevk: hehehe...
15:10
<heycam>
Domenic, send more PRs! :)
15:11
<Domenic>
yeah, if people like this observable idea I'd be happy to PR it in
15:13
<Ms2ger>
heycam, accept the PRs you get first!
15:14
heycam
looks at Ms2ger's PR
15:14
<heycam>
sorry i haven't looked at your CSSRule rejigging patch yet either
15:14
<Ms2ger>
heycam, and ArtB's that's been sitting for 5 months?
15:16
<zcorpan_>
MikeSmith: can you make a whatwg html-differences bugzilla component please?
15:17
<MikeSmith>
yup
15:17
<MikeSmith>
gimme a minute
15:18
<annevk>
heycam: in Europe already?
15:18
<heycam>
annevk, yep, just arrived
15:18
<heycam>
annevk, did you move to Switzerland yet?
15:18
<annevk>
yeah I have
15:18
<heycam>
ah cool
15:19
<annevk>
it is a bit, very rainy summer
15:23
<MikeSmith>
zcorpan_: https://www.w3.org/Bugs/Public/buglist.cgi?component=HTML%20Differences&list_id=42473&product=WHATWG&resolution=---
15:23
<zcorpan_>
MikeSmith: thanks!!
15:29
<annevk>
I wonder if https://twitter.com/hfiguiere/status/498885668937170945 is still true. Really seems like someone ought to reinvent the CA business
15:32
<JakeA>
It's in dire need of "disrupting" as them business folks would say
15:35
<Ms2ger>
"Orca screen reader now supports longdesc (in Gecko)"
15:35
<Ms2ger>
Progress!
15:36
<foolip>
why is everyone talking about https for whatwg.org these days?
15:40
<jgraham>
foolip: https jusst became fashionable
15:59
<annevk>
foolip: well, seems like a good place to start with TLS
16:00
<annevk>
Also, if that's what is required to have a normal site these days, it would be nice to set a good example
16:07
<zcorpan>
please review http://html-differences.whatwg.org/
16:08
<Ms2ger>
zcorpan, weird underline for " and stopSorting(). "
16:08
<Ms2ger>
Also " and border attribute"
16:11
<zcorpan>
fixed thx
16:11
<estellevw>
there are a few areas matching p > span:not([title=""]):not([class~="XXX"]):not([class~="impl"]):not([class~="note"]), li > span:not([title=""]):not([class~="XXX"]):not([class~="impl"]):not([class~="note"]) that get a green underline.
16:11
<zcorpan>
estellevw: reload?
16:11
<estellevw>
fixed
16:12
<estellevw>
took me longer to write that line than for you to fix it ;)
16:12
<zcorpan>
:-)
16:35
<Ms2ger>
TabAtkins, so, about -goog-... What's the advantage over doing the same thing without -goog-?
16:42
<Hixie_>
marcosc: yt?
16:50
<annevk>
Ah, answer is at http://lists.w3.org/Archives/Public/public-webappsec/2014Aug/0067.html
16:50
<annevk>
Netflix streams encrypted content over HTTP
16:50
<Hixie_>
yay unencrypted signalling
16:50
<annevk>
And Netflix does not want to upgrade their infrastructure so instead they argue that new features should be enabled over HTTP
16:51
<annevk>
I guess Mixed Content allows that kind of thing :-(
16:57
<zcorpan>
is https the new utf-8?
17:01
<TabAtkins>
Ms2ger: Ah, easy to answer.
17:02
<TabAtkins>
Ms2ger: If you ship to 1% *without* prefixing, then the page feature-tests for the unprefixed feature. When we eventually turn on the feature for real, possibly with changes, it triggers that codepath, and possibly causes errors.
17:03
<TabAtkins>
Ms2ger: So instead, you use a prefix, so the eventual final thing won't be hit by the "experimental code" branch. And by only shipping it to 1% or so of people, you ensure that it really *is* a branch, not the program just relying on the prefix all the time.
17:03
<TabAtkins>
1% is nice, because it still gives plenty of hits and data-collection opportunity, but takes far too long to restart-cycle yourself into having it working.
17:04
<TabAtkins>
We discussed this as an option back when we were planning the Blink fork, but decided to just follow Firefox's lead instead. We're back to experimenting with it now.
17:10
<Hixie_>
prefixing doesn't work because people just assume the non-prefixed version will be the same and preemptively give both
17:12
<TabAtkins>
Hixie_: It's possible that might still happen. We'll see.
17:13
<TabAtkins>
We're optimistic that the fact that you have to provide fallbacks and explicitly feature-test will mean that people don't just auto-provide unprefixed paths as well.
17:15
<Ms2ger>
I'm doubtful
17:17
<TabAtkins>
Welp, we'll see. Not a huge deal for this thing if people still screw it up.
17:22
<annevk>
zcorpan: yeah something like that
17:23
<annevk>
zcorpan: the new http://www.webstandards.org/
17:24
<annevk>
zcorpan: the new XHTML 1.1
17:24
<annevk>
oh wait
17:24
<annevk>
that didn't go so well
17:26
<Ms2ger>
XHTML2?
17:29
<SamB>
annevk: what, SSL certs aren't already a scam ;-P?
17:30
<SamB>
(those signed by the CArtel, I mean)
17:30
<annevk>
SamB: via startssl you can get one for free
17:30
<SamB>
so it's not a very-well-run scam
17:30
<annevk>
but yes, replacing the CA system would be great if that was feasible
17:31
<annevk>
bit of a shame that it took us this long to realize everyone needs TLS
17:31
<SamB>
heck, even restricting the national CAs to their own TLDs would be an improvement
17:49
<annevk>
If we could make it part of DNS somehow and just let the registrars deal with it
17:49
<Hixie_>
do we have a spec for real-world content-disposition?
17:49
<annevk>
Everyone that has a domain has figured out how to deal with registrars
17:50
<annevk>
Hixie_: jreschke has tests for that at least
17:50
<annevk>
Hixie_: http://greenbytes.de/tech/tc2231/
17:50
<Hixie_>
yeah but i'm sure his tests just show that browsers don't follow the spec
17:58
<SamB>
annevk: well, obviously that's only sane if you have dnssec working
18:03
<SamB>
and, say, the glibc people are understandably not too enthusiastic about pulling in the requisite crypto code to check dnssec results in-process
18:12
<SamB>
annevk: though they seem open to adding support for declaring external resolvers (e.g. dnsmasq) as trusted to check dnssec results, and adding flags to treat failure to obtain a dnssec-verified result as an error ...
18:13
<SamB>
annevk: so were you thinking that thing where you can put your SSL key fingerprint or whatever into a DNS record, or something else?
18:37
<Hixie_>
annevk: so... for load-settings=""... do we want to do a JSON blob?
18:37
<Hixie_>
annevk: or something more user-friendly?
18:37
<Hixie_>
annevk: and do we want to reflect that as a DOMString? or something more usable?
18:37
<Hixie_>
annevk: it's just a flat name-value pair dictionary, no? nothing complex?
18:38
<Hixie_>
annevk: maybe it should just be a semi-colon separated set of colon-separated name/value pairs?
18:53
<annevk>
Hixie_: if you want both headers and settings, it'll be more complex than name-value
18:54
<Hixie_>
ah, right
18:54
<annevk>
Hixie_: well, it's name-value, but not string-string
18:56
<annevk>
SamB: I haven't really considered it in detail, it's mostly a thought
18:57
<annevk>
SamB: would require quite a bit more effort to turn this into something tangible
19:17
<annevk>
Array.prototype.slice has a fast path for NodeList https://twitter.com/bz_moz/status/500121412523532288
19:17
<annevk>
Oh DOM, you so bad
20:53
<miketaylr>
is there a specified behavior for https://miketaylr.com/bzla/docwrite.html?
20:53
<miketaylr>
firefox/presto and blink/webkit disagree
20:55
<caitp>
typeof undefined is usually not "function"
20:55
<Ms2ger>
Probably, yes
20:56
<Ms2ger>
I'll bet that hsivonen even knows which one is right
20:57
<Ms2ger>
I suspect it's Fx/Presto
20:57
<Ms2ger>
Because the inner window changes
20:57
<caitp>
I get the same results in FF nightly and chrome stable
20:57
<caitp>
so... go figure?
21:00
<miketaylr>
caitp: really
21:01
<miketaylr>
i get different results in chrome stable and canary... and chrome mobile
21:01
<miketaylr>
naturally there are sites depending on this... https://bugzilla.mozilla.org/show_bug.cgi?id=1015725#c2
21:02
<caitp>
seems to be same results in canary (39.0.2129.0 (Official Build 290688) canary) too
21:02
<miketaylr>
let me take out document.googWrite
21:02
<miketaylr>
you might be hitting that
21:02
<miketaylr>
:P
21:03
<miketaylr>
caitp: so once you click the button, Chrome and Firefox both say ____?
21:03
<miketaylr>
for me, chrome says true, fx false
21:03
<caitp>
ah I see
21:04
<caitp>
yes, different results there
21:04
<miketaylr>
ok, phew
21:20
<caitp>
well, the applicable spec that matters here is probably http://www.whatwg.org/specs/web-apps/current-work/#dom-document-open
21:24
<caitp>
so you get a new "window" global, effectively an entirely new script isolate really
21:24
<caitp>
so I guess we're doing the wrong thing in blink/v8
21:43
<smaug____>
miketaylr: Gecko does the right thing
21:45
<miketaylr>
thx caitp smaug____
21:46
<caitp>
wonder if there's a crbug for that
21:47
<miketaylr>
dunno, will probably file one if not
21:47
<miketaylr>
but every site that uses mobify.js relies on their current behavior
21:47
<caitp>
i dunno if you can really call it a change that would "break the web" if other popular browsers are doing the right thing
21:49
<miketaylr>
heh, doing the right thing (and have broken mobile sites as a result)
21:49
<caitp>
https://crbug.com/149785 looks like there's a bug
21:49
<miketaylr>
yep, looks like it
23:34
<Domenic>
annevk: sorry missed HSTS question. I dunno really anything about HSTS. I think the cert was $15 so.…