05:31
<igrigorik>
annevk: thanks for the reference, followed up on the thread: https://github.com/whatwg/fetch/issues/47#issuecomment-143948846
06:13
<annevk>
gsnedders: yeah, mkwst is investigating writing a standard for it
06:13
<annevk>
igrigorik: you haven't really addressed the issue
06:52
<mkwst>
gsnedders, annevk: Yeah. I had a spec for this, and the group got cold feet after some challenges popped up. I'd like to bring it back up, but I'm loath to do so until I know how to do the blocking in Chromium so that it's not just an empty document.
06:53
<annevk>
mkwst: why does it have to be an empty document?
06:53
<annevk>
mkwst: is this related to Chromium making error pages same-origin?
06:53
<mkwst>
annevk: Hrm? No, sorry. Maybe I didn't understand the discussion.
06:54
<mkwst>
annevk: I mean that it's tough for our network stack to distinguish a priori an intranet address from an internet address before connection time.
06:54
<mkwst>
annevk: And we don't have enough context at the time that we're making a connection to determine whether we should block the former or not.
06:54
<annevk>
mkwst: oh
06:54
<mkwst>
So I need to do some work (or find someone to do some work).
06:55
<mkwst>
I don't really want to put up a doc that says "DO X!" if I don't know how to do X in the browser I'm responsible for. :)
06:55
<mkwst>
That's what I mean by "empty document". A spec without code is fairly worthless.
07:03
<annevk>
mkwst: oh, I thought you meant the type of error handling
07:04
<mkwst>
Yeah, sorry. I wasn't clear.
07:09
<annevk>
By the way, there's an intern spot to be filled for the HTML Standard: https://wiki.mozilla.org/Outreachy/2016/December_to_March#Contribute_to_the_HTML_Standard.21
07:10
<annevk>
And also somewhat relevant to the WHATWG, folks can work on implementing Fetch in Servo: https://wiki.mozilla.org/Outreachy/2016/December_to_March#Servo:_Complete_implementation_of_Fetch_standard
08:23
<philipj>
Ms2ger: there?
08:45
<philipj>
Ms2ger: It looks like idlharness.js doesn't try to pass null or invalid types to methods in order to check if they'll throw TypeError, is that right or am I doing it wrong?
08:46
<Ms2ger>
That sounds plausible
08:47
<Ms2ger>
In general, "invalid type" is somewhat difficult to construct
08:50
<philipj>
Ms2ger: for interface types that are not nullable, passing in null would be nice
08:50
<Ms2ger>
Agreed
08:51
<Ms2ger>
There's also overloads, unfortunately
08:51
<philipj>
and I guess there are no types that are both instanceof Document and instanceof Window, so one could always use one of those?
08:51
<philipj>
oh, right
08:52
<philipj>
yeah, whatever the implementation of giveMeAnInvalidType() is, it should be possible to write an interface that defeats it
08:56
<philipj>
Ms2ger: is there somewhere I could file an issue for the simple null case that has a non-zero chance of being fixed?
08:56
<philipj>
I'm trying to answer a question for someone else, so I'm not excited about fixing it myself today :)
09:45
<Ms2ger>
philipj, testharness.js repo, I sometimes look at it :0
09:45
<philipj>
Ms2ger: ok :)
10:21
<SteveF_>
Domenic: pinged you on twitter (then realised I am persona non grata [wise move] in your twitterverse) so leaving here Script-Based Web Accessibility proposal https://github.com/cyns/wapa/blob/master/ScriptAccessibility.md#script-based-web-accessibility
11:05
<annevk>
mkwst: this whole modularizing CSP seems like a Process hack and I suspect will only cause you pain due to the shifting dependencies and requirements over time
11:11
<annevk>
mkwst: this is basically what's wrong with the way the W3C works, you start optimizing for producing RECs and as a result you create a disaster for everyone else trying to follow along
11:37
<mkwst>
annevk: Eh. I understand that critique, and I agree to an extent.
11:38
<mkwst>
annevk: However: 1. Beyond W3C process, I think separate documents will give folks like Microsoft the ability to say they implemented a thing without blocking on implementing everything.
11:39
<mkwst>
annevk: 2. Directives are already modular to a great extent. There's no particular reason that `upgrade-insecure-requests` needs to be in the same document as `sandbox`. They do completely divergent things, and are only in the same document because of the delivery mechanism.
11:39
<mkwst>
annevk: 3. Separate documents allow more explanation. I think the cookies feature is significantly clearer, both in justification and scope, in a separate, focused document.
11:45
<mkwst>
(In other words, CSP is a generic delivery mechanism for a bunch of stuff. Some of that stuff is similar to some other bits of that stuff, and grouping them together makes sense to me.)
12:02
<annevk>
mkwst: did Microsoft actually say as much? It has never been a problem for them before
12:02
<annevk>
mkwst: see e.g., HTML or URL
12:02
<mkwst>
annevk: No. They've said they aren't prioritizing "CSP2", and have been unresponsive to my (repeated) suggestions that there are important bits and not so important bits.
12:03
<annevk>
mkwst: right, I don't think splitting will change that
12:05
<mkwst>
annevk: I'm not suggesting that it solves the problem. I think it makes the problem smaller.
12:06
<Ms2ger>
It makes the problem bigger, because now they have a dozen specs they can ignore
12:08
<mkwst>
If the specs are smaller, implementing them is easier, which means they can tick a nice feature checkbox. That's appealing.
12:09
<mkwst>
(also, annevk: if you haven't already, tell me I'm an idiot on the mailing list. :) )
12:11
<jgraham>
I'm unconvinced that smaller specs are easier to implement
12:11
<jgraham>
depending on how they became "smaller"
12:12
<jgraham>
Like, one way to make small specs is to specify features in isolation and conveniently ignore the interaction between that feature and other features
12:12
<jgraham>
So things look small but actually they aren't
12:12
<philipj>
annevk: when merging PRs that need a new ack, do you think we should just change the commit, or do a separate "ack for previous commit" thing?
12:12
<jgraham>
They're just underspecified
12:12
<annevk>
mkwst: I've made the point that the W3C Process has detrimental effects on everyone but lawyers enough times
12:12
<jgraham>
(*cough* web-performance *cough*)
12:12
<annevk>
philipj: I think we should change the commit and override with --author
12:13
<mkwst>
annevk: If you can invent a world in which lawyers don't matter, I'm totally happy to join you there.
12:13
<philipj>
annevk: you mean preserve the original author, or to pretend that we are the authors of the fix too?
12:13
<annevk>
philipj: preserve
12:13
<annevk>
mkwst: you're already here
12:13
<philipj>
annevk: right, that's what git commit --amend does by default
12:13
<mkwst>
...
12:14
<annevk>
:-P
12:14
<annevk>
philipj: so whenever I use git commit --amend I can modify the commit message, but not the commit itself
12:14
<philipj>
annevk: you need to git add -p first
12:15
<philipj>
or any kind of git add, -A would do as well
12:15
<annevk>
philipj: so make a change, use git add -p, then do git commit --amend?
12:15
<philipj>
annevk: yep
12:15
<annevk>
philipj: thanks, that should make things easier
12:15
<philipj>
annevk: :)
12:15
<annevk>
philipj: I usually fiddle around a bit with the GitHub GUI which is not really ideal, but I managed to get the same effect in the end
12:16
<jgraham>
mkwst: So the flip side of the lawyer coin is that if no one had done HTML5 because W3C dropped it and WHATWG weren't lawyer-friendly then, well, that world would be a worse world than this one. So it's empirically true that it is possible to worry too much about the problems with doing work and not enough about the problems of not doing work.
12:17
<mkwst>
jgraham: Put the lawyer stuff to the side. My problem at the moment is that the work I'm doing is helping Chrome users, and Firefox users to an extent, but isn't helping anyone else (and, due to that lag, isn't helping Chrome users as much as it should).
12:17
<mkwst>
jgraham: I'm looking for solutions to that problem.
12:18
<jgraham>
Well, I don't know of a way to force Microsoft to implement something :|
12:18
<mkwst>
Or Apple. Or Mozilla.
12:18
<mkwst>
Forcing is the wrong question, I think.
12:18
<mkwst>
I want to make things appealing to implement.
12:19
<jgraham>
I think things are appealing to implement if they solve real problems that users are percived to have
12:19
<jgraham>
I suspect the way the spec is structured is a correction to that
12:20
<mkwst>
I think splitting things out makes individual improvements clearer, yes.
12:21
<mkwst>
https://w3c.github.io/webappsec/specs/csp-cookies/ is a tiny spec that (tries to) identify a specific problem, and proposes a specific solution. It delivers that solution in terms of CSP. *shrug* That seems like a clear package.
12:21
<mkwst>
(With the usual strawman, rough draft, etc. caveats. Assume I spent time to polish things.)
12:22
<annevk>
I've never know the presentation of a specification to be what convinces Apple and/or Microsoft
12:22
<annevk>
What matters is developers requesting the functionality and the popularity of the features
12:23
<annevk>
That's why Microsoft eventually caved and implemented CORS for XMLHttpRequest, and dropped XDomainRequest
12:27
<Domenic>
SteveF_: oh cool, will take a look... and should probably unblock you, I just remember blocking you after you said something about cocks in response to one of my tweets
13:48
<annevk>
Hmm, more magic in HTML
13:48
<annevk>
Discard tasks. If anything got discarded, do x.
13:57
<SteveF_>
Domenic: sorry can't remember tweeting "about cocks in response to one of my tweets", but will make an effort not to tweet such stuff to you in future.
14:52
<annevk>
Domenic: could you give https://github.com/whatwg/html/pull/203 a final pass?
14:53
<Domenic>
annevk: will do
15:09
<annevk>
Domenic: ah thank you for catching that
15:09
<annevk>
Domenic: it seems the parser does not choke on typos
15:10
<Domenic>
annevk: yeah I figured this one would need a review of the built product
15:54
<annevk>
Domenic: pushed the changes you wanted
15:54
<annevk>
Domenic: I think HTMLAllCollection should just become its own class
15:54
<Domenic>
annevk: reviewing built output now
15:55
<Domenic>
annevk: that seems probably true, although we should investigate in a follow-up bug
15:55
<annevk>
Domenic: although perhaps someone could quickly peek at the Chromium code for HTMLAllCollection whether that is true there too
15:55
<annevk>
Domenic: will file an issue
23:30
<MikeSmith>
anybody else having a problem with gmail currently not loading in Canary (on OS X at least)?
23:31
<MikeSmith>
and if so, know any workaround
23:31
<MikeSmith>
it hasn't been working for me for days now