05:17
<annevk>
igrigorik: GitHub was not interested in fixing the redirect issue because HTTPS support is not supported
05:18
<annevk>
igrigorik: I pointed out it clearly works, just not the redirects, no reply yet...
05:22
<annevk>
igrigorik: you might be right about CSP, I guess once the processing model is clear that should be easy enough to evaluate
05:23
<annevk>
igrigorik: I'm still worried that you can download something using img-src and if the cache doesn't store that you can then use it for script-src, but who knows
05:35
<MikeSmith>
annevk: did you take a look at https://github.com/zaach/jsonlint
05:36
<annevk>
MikeSmith: briefly
05:36
<annevk>
MikeSmith: did you listen to http://mozaic.fm/post/108439721723/14-whatwg?
05:36
<MikeSmith>
eh
05:36
<MikeSmith>
what is this
05:36
<annevk>
MikeSmith: I don't have npm I think
05:36
<MikeSmith>
oh it's Yakura-san, and Hiro
05:37
<MikeSmith>
annevk: well you should have npm...
05:37
<MikeSmith>
this seems like a good use case for setting it up
05:38
<annevk>
I really dislike software with tons of dependencies that infects all kinds of directories and I can never be sure whether it's gone from the OS or how it actually works
05:38
<annevk>
But I guess...
05:39
<MikeSmith>
yeah sadly that sorta describes npm pretty well
05:39
<MikeSmith>
but if you have to work with python you probably have had to use pip or whatever
05:39
<MikeSmith>
anyway
05:40
<MikeSmith>
wow this podcast is pretty interesting, even just from looking at the agenda, and listening to the first few minutes
05:40
<annevk>
MikeSmith: yes, I have the same problem with Python :-)
09:48
<annevk>
Yay, down to two open issues for https://encoding.spec.whatwg.org/
09:48
<annevk>
And neither is something I can address
09:50
<MikeSmith>
"Filing this issue so hopefully one day a browser can experiment and figure out if this is doable."
09:50
<MikeSmith>
kind like putting a message in a bottle and throwing it into the ocean
09:52
<annevk>
hah
09:52
<annevk>
So hsivonen expressed interest in doing it at some point, that's why it's not completely dead
09:52
<annevk>
But it wasn't worth keeping the original bug open for that and therefore keep having the "legacy open bugs" link at the top of the Standard
09:55
<MikeSmith>
what blocks the other bug?
09:55
<MikeSmith>
needs further implementor feedback?
10:08
<annevk>
MikeSmith: yeah, need to figure out if it's actually feasible
10:09
<annevk>
I'm pretty happy to have resolved all bugs around encoders/decoders though
10:09
<annevk>
Was less work than expected
11:17
<annevk>
JakeA: how is progress on promise.abort() going?
11:18
<JakeA>
annevk: halted :( been given tight deadlines for other things, and I think Domenic is busy too
11:18
<JakeA>
Domenic: you got any spare time coming up?
11:18
<annevk>
If he does, perhaps IDL can finally get some love
11:22
<annevk>
JakeA: do you know balfanz from https://github.com/whatwg/fetch/issues/30?
11:22
<annevk>
JakeA: I'd like to help them out, but communication seems to be going rather poorly
11:24
<JakeA>
annevk: I don't know them. Not familiar with token binding at all, but will try to go through that doc
11:31
<annevk>
JakeA: don't worry about it if you have deadlines, this has been going on for a while now
11:32
<JakeA>
Maybe in a couple of weeks I'll feel more in control and wonder what I was panicking about. That's usually how it goes.
11:34
<annevk>
wanderview: where did the fetch() tests go from Chrome?
12:32
<annevk>
MikeSmith: https://www.w3.org/Bugs/Public/show_activity.cgi?id=27502 last activity looks suspect
13:20
<wanderview>
annevk: we have not imported them yet https://bugzilla.mozilla.org/show_bug.cgi?id=1186106
13:28
<Huvet>
can I ask questions about html5lib here?
13:29
<Huvet>
(python implementation)
13:29
<Ms2ger>
Sure
13:30
<Huvet>
I would like to write a small script that parses a html snippet, removes all style and script tags, and then serializes it out as a string with HTML again. This is how far I've come: http://dpaste.com/3CD5V55
13:31
<Huvet>
but when serializing I get the error: AttributeError: 'TreeWalker' object has no attribute 'getroot'"
13:31
<Ms2ger>
Ah, the mess of python tree apis :/
13:32
<Huvet>
yeah, if anyone has better ideas of how to get this done I'm all ears
13:33
<Ms2ger>
I wish
13:34
<Huvet>
the question becomes: how do i serialize a lxml tree?
13:35
<Ms2ger>
I wish I could say something useful, but I always have similar issues
13:37
<MikeSmith>
annevk: Thanks, disabled that bugzilla account
13:38
<Huvet>
I did it! http://dpaste.com/0PN6E5F
14:51
<wanderview>
annevk: is there any way for the body of an opaqueredirect to be actually used? its not exposed to script and it seems the UA won't look at it
15:18
<Domenic>
JakeA: no real spare time, but I've been trying to figure out Q4 goals and wondering if cancelable promises should be one of them...
15:51
<annevk>
wanderview: nope
16:04
<igrigorik>
annevk: re, Github.. wow, really? That's not like them. I might have to resort to public shaming.
16:06
<annevk>
igrigorik: I got another follow up
16:06
<annevk>
igrigorik: the problem is their CDN
16:07
<annevk>
igrigorik: which they have said before, so basically HTTPS is only partially supported...
16:07
<igrigorik>
when in doubt, always blame the CDN! Fastly? We can bug them too :)
16:07
<annevk>
igrigorik: but yeah, shaming might work, at some point they have to start offering HSTS et al
16:07
<annevk>
igrigorik: they didn't mention which CDN, could be
16:08
<igrigorik>
github.io maps to github.map.fastly.net, so it is Fastly. did they say what the issue actually is?
16:09
<JakeA>
Domenic: might be an idea, my end of year is looking messy :(
16:09
<annevk>
igrigorik: just that they're not in control
16:09
<annevk>
igrigorik: could shame Fastly asking when they'll finally properly support github.io
16:09
<igrigorik>
who are you talking to at Github? Any chance you can cc me and I can loop in the fastly crew?
16:10
<annevk>
was an exchange with someone from support
16:10
<igrigorik>
hmm, ok, I'll kick off a separate thread then
16:11
<annevk>
I forwarded you an email
16:11
<igrigorik>
cool, thanks.
16:41
<terinjokes>
anyone know whom i should talk to at w3 about upgrading the OpenSSL version used for their Feed Validation Service?
16:42
<Ms2ger>
systeam@?
16:43
<annevk>
and MikeSmith most likely
16:43
<terinjokes>
people have been reporting the bug to the list since at least last October
16:43
<annevk>
there's also #sysreq on irc.w3.org
16:44
<ccardona-work>
Good morning and happy friday WHATWG crew 😎✌🏽
16:44
<terinjokes>
ccardona-work: morning
16:44
<terinjokes>
annevk: thanks, i'll try there
16:45
<annevk>
ccardona-work: good evening!
16:45
<ccardona-work>
annevk: :) Did I at least get the Friday bit correct?
16:45
<annevk>
ccardona-work: yes!
16:45
<ccardona-work>
score
16:46
<Ms2ger>
Not in Tokyo ;)
16:46
<annevk>
Yeah, seems MikeSmith is ahead of us
16:46
<ccardona-work>
:)
16:46
<annevk>
and hopefully asleep
16:46
<ccardona-work>
MikeSmith: is always ahead of us :)
16:46
<ccardona-work>
Thoughtleader && time leader
16:46
<igrigorik>
annevk: based on our chat yesterday: https://rawgit.com/w3c/preload/processing/index.html#processing - see 'obtain' section.. wdyt?
16:48
<annevk>
igrigorik: seems to be heading in the right direction
16:48
<igrigorik>
there are other details to work out there, but first want to resolve the as -> type mapping..
16:50
<igrigorik>
as written, <link rel=preload as=image href=thing.jpg> would be initialized with Request.type=image and will be subject to image-src.
16:51
<annevk>
igrigorik: so... say I have CSP img-src: example.com elsewhere.com; script-src: elsewhere.com
16:51
<annevk>
igrigorik: <link rel=preload as=image href=example.com>
16:51
<annevk>
example.com redirects to elsewhere.com
16:51
<annevk>
I then do <script src=elsewhere.com>, does it get the preloaded resource?
16:52
<igrigorik>
annevk: no, it shouldn't.. when you request it via <script> the request.type won't match, since that was set to image.
16:53
<annevk>
igrigorik: so you store the type in the preload cache?
16:54
<igrigorik>
annevk: yeah, we'll need that to get the plumbing right.. afaik, SW caches api does that already?
16:55
<annevk>
the Cache API stores the Request, but it doesn't use everything when matching
16:55
<annevk>
it won't use type for instance
16:56
<igrigorik>
I see, then yeah we might need to sort out that plumbing and include that check. As I noted later in that doc the 'preload/fetch cache' is a bit of a magical thing in the browsers today.. we need to figure out how to spec-explain it :)
16:57
<annevk>
Perhaps you can reuse some stuff from the CORS cache
16:57
<annevk>
It's not that hard to define a cache in prose, you just need to be clear about when writing and reading happens and what happens for misses and such
16:58
<annevk>
And the conditions of course, such as whether type needs to match :-)
17:00
<annevk>
https://twitter.com/jacobrossi/status/634768416088424448 this is great
17:00
<annevk>
igrigorik: why do you have nigiri as favicon? :-)
17:01
<igrigorik>
thanks for the tip, I'll take a look.. On that note though, seems like the lookup logic should live in Fetch? Since, implicitly what we have today is: prefetch cache -> HTTP cache -> network
17:01
<annevk>
ooh, that's from rawgit.com
17:01
<igrigorik>
yeah :)
17:01
<annevk>
igrigorik: yeah, the lookup needs to happen in Fetch someplace
17:02
<igrigorik>
you're looking at a branch.. I haven't landed those updates yet. wanted to get a sanity check first.
17:02
<annevk>
igrigorik: ideally just a simple hook where I invoke something in preload
17:02
<annevk>
igrigorik: and preload returns me a response, and if I have a response I simply don't go to the network (or the service worker?)
17:03
<igrigorik>
annevk: yep, makes sense.. since SW could have served the prefetch response ^_^
17:03
<annevk>
right
17:05
<igrigorik>
annevk: Ok, well, sounds like the 'as' bits are in a reasonably good place. I'll land that as a start and we can iterate on the cache bits to define all the matching + hooks semantics.
17:18
annevk
curses blob URLs
19:03
<annevk>
nox: is https://twitter.com/nokusu/status/634686055548100608 your work?
19:03
<annevk>
nox: sounds pretty cool
19:03
<nox>
annevk: That's me yes.
19:03
<nox>
@RustLang RT'd it, I keep getting notifications now, hah.
19:09
<annevk>
nox: it's kind of cool that even encoding/decoding can use this stuff
21:24
<jamesr___>
convert ASCII to UTF-8
21:24
<jamesr___>
?
21:24
<jamesr___>
is that not memcpy() ?
21:26
<jamesr___>
oh, the input bytes are maybe ASCII but you still have to check?
23:05
<MikeSmith>
terinjokes: I would guess that whatever version of openssl running on that host is whatever security-patched version is currently in debian stable
23:58
<TabAtkins>
annevk: I'm rewriting the serializer to be more opinionated about formatting (and as a result, no longer strew bonus newlines everywhere).
23:58
<TabAtkins>
Since you care about the output, what do you prefer in the situation where a block element contains only inline content:
23:59
<TabAtkins>
(a) the block's start tag, content, and end tag are all on the same line, or
23:59
<TabAtkins>
(b) the block's start tag is on one line, the content is indented on the next line, and the end tag is on a third line