| 09:08 | <annevk> | In WPT, how do you change "http://{{host}}:{{ports[http][0]}}/" into something that omits the port when it's the default port? |
| 09:09 | <annevk> | gsnedders: why does http://web-platform-tests.org/writing-tests/testharness-api.html host the content in an <iframe>? |
| 09:11 | <annevk> | I can never find documentation for that syntax either |
| 09:12 | <annevk> | Sigh, this is all so broken |
| 09:15 | <annevk> | Put your thoughts in https://github.com/w3c/web-platform-tests/issues/2618 if you're reading this and now how to fix it |
| 09:32 | <zcorpan> | annevk: http://wptserve.readthedocs.io/en/latest/pipes.html |
| 09:33 | <zcorpan> | annevk: omitting default port is something you'd have to do in javascript I believe. There's no substitution feature for that in wptserve |
| 09:33 | <annevk> | Sigh |
| 09:34 | <annevk> | You'd think these kind of things would be common enough to be abstracted by now |
| 09:34 | <annevk> | There's a bunch of logic in fetch/ somewhere for this and I've seen it in various other places too |
| 09:34 | <annevk> | It all looks rather brittle |
| 09:38 | <annevk> | jgraham: https://github.com/w3c/web-platform-tests/pull/3451 has gone for six months without review, can I just land it then? |
| 09:45 | <zcorpan> | annevk: reviewed and merged |
| 09:46 | <annevk> | thanks |
| 09:47 | <annevk> | zcorpan: https://github.com/w3c/web-platform-tests/labels/XMLHttpRequest has three more at the top looking for review |
| 09:54 | <zcorpan> | annevk: is there an issue or PR where URLSearchParams was added to BodyInit? |
| 10:11 | <annevk> | zcorpan: no |
| 10:12 | <annevk> | zcorpan: I think it was there in Fetch from the start and then XHR was refactored |
| 10:12 | <annevk> | zcorpan: but I might be mistaken |
| 10:13 | <annevk> | zcorpan: are you filing browser bugs? |
| 10:13 | <zcorpan> | annevk: yes |
| 10:14 | <zcorpan> | filed one for chromium |
| 10:14 | <zcorpan> | trying to figure out if Edge needs an issue |
| 10:14 | <annevk> | zcorpan: it might be that Chrome fails due to not supporting the constructor I used |
| 10:15 | <zcorpan> | annevk: no, the constructor works afaict |
| 10:15 | <zcorpan> | but URLSearchParams does not exist in Edge 13 |
| 10:21 | <zcorpan> | per http://caniuse.com/#feat=urlsearchparams does not exist in 14/15 either |
| 10:38 | <annevk> | jgraham: I get 'Request' object has no attribute 'raw_headers' and _raw_headers still seem to exist |
| 10:38 | <annevk> | jgraham: no submodule mismatch is reported and when I manually go into tools/wptserve/wptserve/ it does say raw_headers without _ prefix |
| 11:00 | <jgraham> | annevk: You did |git submodule update --recursive| ? |
| 11:00 | <annevk> | jgraham: yes, otherwise I would have outstanding changes |
| 11:03 | <jgraham> | annevk: Well it's working here |
| 11:03 | <jgraham> | def main(request, response): return [("Content-Type", "text/plain")], str(request.raw_headers) |
| 11:05 | <annevk> | jgraham: I get AttributeError: 'Request' object has no attribute 'raw_headers' if I load that file |
| 11:06 | <jgraham> | annevk: Which commit are you on? |
| 11:07 | <annevk> | jgraham: https://github.com/w3c/web-platform-tests/commit/d19fe427d7670a71ba55206208a6515303590b5b |
| 11:07 | <annevk> | jgraham: if I open tools/wptserve/wptserve/request.py it does say raw_headers |
| 11:07 | <annevk> | hmm |
| 11:08 | <annevk> | jgraham: apologies, forgot to restart the server |
| 11:08 | <annevk> | jgraham: I just keep that thing running forever, didn't think about it |
| 11:10 | <jgraham> | OK, np :) |
| 11:24 | <JakeA> | What tools do people use to validate webidl? |
| 11:25 | <annevk> | bz |
| 11:25 | <tobie> | :D |
| 11:25 | <annevk> | Bikeshed too I suppose |
| 11:26 | <tobie> | Yup. Bikeshed won't let you write grammatically incorrect WebIDL. |
| 11:26 | <JakeA> | yeah, I'm at an earlier stage than bikeshed :D |
| 11:26 | <JakeA> | Actually, maybe that's wrong |
| 11:26 | <JakeA> | I should just start specing |
| 11:27 | <tobie> | It will however let you write WebIDL that's grammatically correct but otherwise not. |
| 11:29 | <annevk> | zcorpan: did you ever conclude URL + encoding bugs/tests? |
| 11:29 | <annevk> | zcorpan: I find XMLHttpRequest/open-url-encoding.htm still failing everywhere |
| 11:35 | <zcorpan> | annevk: i think i did not report browser bugs |
| 11:39 | <zcorpan> | annevk: tests in html/infrastructure/urls/resolving-urls/query-encoding |
| 11:42 | <annevk> | zcorpan: running query-encoding/windows-1251.html I get numerous Python errors |
| 11:42 | <zcorpan> | annevk: yeah, i just noticed that as well. :-( |
| 11:42 | <annevk> | zcorpan: it's also probably too big to reliably run on Chrome's testing servers which seem to have weird limits |
| 11:46 | <zcorpan> | annevk: filed https://github.com/w3c/web-platform-tests/issues/4934 |
| 13:33 | <annevk> | zcorpan: note that Chrome does pass XMLHttpRequest/send-usp.html so I think it's the constructor that's the problem |
| 13:33 | <annevk> | (I didn't know that test existed) |
| 14:26 | <gsnedders> | annevk: the reason why testharness docs are in an iframe is because they're in a submodule and we can't get them onto the site on GitHub Pages without duplicating the submodule (with the risk of it ending up at a different revision than at /resources). There's a Jekyll bug on this and GitHub are happy to have something that fixes it, but obviously it needs fixed and then released and then gone ... |
| 14:26 | <gsnedders> | ... live on GitHub Pages before it can be fixed |
| 14:27 | <annevk> | We could also not use GitHub Pages? |
| 14:27 | <annevk> | Write some kind of build script and push out HTML using rsync? |
| 14:27 | <annevk> | Does GitHub Pages have HTTPS for custom domains? |
| 14:30 | <gsnedders> | annevk: I was trying to avoid that, as it gets a bit fragile a bit too easily (push two things seconds after one another, whichever website build completes last on Travis goes live) |
| 14:31 | <gsnedders> | (we can totally come up with ways to avoid that, but it potentially makes the travis website build jobs slow) |
| 14:31 | <annevk> | It hasn't been a problem for any WHATWG Standard |
| 14:32 | gsnedders | guesses none get the same number of pushes per day as wpt |
| 14:33 | <jgraham> | annevk: Only https if you use cloudflare, which isn't really |
| 14:33 | <annevk> | gsnedders: even so, if you count the number of changes to WPT documentation... |
| 14:33 | <annevk> | jgraham: :/ |
| 14:33 | <jgraham> | I would be happy to use real hosting, but that might mean building a deployment pipeline |
| 14:34 | <annevk> | jgraham: https://github.com/whatwg/whatwg.org is what we use for whatwg.org, it's fairly trivial |
| 14:34 | <annevk> | jgraham: just need a server you can talk to over SSH and Travis CI |
| 14:38 | gsnedders | thinks the simplicity of the current setup is a feature, esp. given the two big issues (testharness.js docs and HTTPS) will get fixed |
| 14:38 | <jgraham> | They will? |
| 14:38 | <jgraham> | In any case, for the short term I don't have any time and I would prefer gsnedders work on CSSWG merge |
| 14:38 | <annevk> | HTTPS for GitHub Pages has been a problem forever |
| 14:39 | <jgraham> | So I guess if someone steps up to do it I'm pretty happy to pay for the hosting or whatever |
| 14:39 | <annevk> | Fair, although the current documentation state feels shittier than before |
| 14:39 | <jgraham> | Well it's better in that the documentation is easier to find and keep up to date |
| 14:39 | <jgraham> | But worse in that it isn't https |
| 14:40 | <jgraham> | Overall it seems like a win with a notable regression |
| 14:40 | <annevk> | The <iframe> is pretty bad |
| 14:40 | <jgraham> | Oh yeah that too |
| 14:41 | <jgraham> | But we will monorepo everything, which might help with that |
| 14:41 | <annevk> | Links to wptserve documentation have gone missing |
| 14:41 | <gsnedders> | annevk: HTTPS on github.io happened eventually, and AIUI custom domains will get HTTPS "soon" |
| 14:41 | <annevk> | gsnedders: pointer? |
| 14:41 | <gsnedders> | annevk: sadly can't remember where I saw that |
| 14:43 | <annevk> | gsnedders: https://github.com/isaacs/github/issues/156 has nothing it seems |
| 14:43 | <gsnedders> | annevk: tbf that's an unofficial third-party issue tracker |
| 14:43 | <annevk> | sure, but if you saw something I'd imagine it would at least be mentioned there |
| 15:06 | gsnedders | dislikes that cloning wpt requires cloning the entire history of all of the vendored modules |
| 15:09 | <jgraham> | annevk: How does one (in specese) create a function body from some text and execute it, these days? |
| 15:09 | <jgraham> | Asking for WebDriver, which needs to be updated for recent changes in HTML |
| 15:09 | <jgraham> | (where recent is actually quite a long time( |
| 15:10 | <jgraham> | gsnedders: Why wouldn't it |
| 15:10 | <annevk> | jgraham: best to ask Domenic that question |
| 15:11 | <gsnedders> | jgraham: every other Python project I know just uses `pip install -t _vendor` rather than including the entire git history of the project they're depending on, especially given not everything lives in git |
| 15:15 | <gsnedders> | (which we do actually do for py already) |
| 15:17 | <jgraham> | gsnedders: But we had a no-pip policy and the submodules will go away anyway |
| 15:18 | <jgraham> | Domenic: ^ |
| 15:20 | <gsnedders> | jgraham: no, I mean you commit the output of that |
| 15:24 | <annevk> | Submodules are going? 🎉 |
| 15:24 | <jgraham> | gsnedders: Seems like it would make updating even worse than it already is |
| 15:39 | <gsnedders> | Anyone have preferences on <https://github.com/w3c/wpt-tools/pull/37>;? |
| 15:48 | <Ms2ger> | -1 |
| 16:15 | <Domenic> | jgraham: is your string like `function (x, y) { ... }` or is it like `...` ? (i.e., full function, or just function body)? |
| 16:18 | <jgraham> | Domenic: Just function body |
| 16:19 | <jgraham> | (I guess there is a reasonable question about whether that's a good idea given it's supposed to be possible to pass arguments into the function, but I think it would be hard to change at this point) |
| 16:19 | <Domenic> | jgraham: I think the easiest way will be CreateDynamicFunction(%Function%, undefined, "normal", << bodyText >>) |
| 16:19 | <Domenic> | Using https://tc39.github.io/ecma262/#sec-createdynamicfunction |
| 16:20 | <Domenic> | You could also extract out the parts of that that actually do stuff in your no-parameters normal-function case, which will be like 6 steps |
| 16:20 | <jgraham> | Domenic: OK. And what does one need to do to arrange for such a function to be executed? |
| 16:21 | <Domenic> | Call(func, thisArg, argsList) |
| 16:21 | <Domenic> | Using https://tc39.github.io/ecma262/#sec-call |
| 16:21 | <JakeA> | annevk: wanderview: I lied about getting to fetch cancelation today. Tomorrow… |
| 16:22 | <jgraham> | Domenic: OK. There isn't any HTML-specific part here? It seems like scripts and event handler attributes have some extra boilerplate |
| 16:23 | <wanderview> | np |
| 16:23 | <Domenic> | jgraham: oh, hmm... well if you need to track incumbent and entry settings objects correctly, which I guess you probably do, then yeah, you need to prepare to run script + prepare to run a callback. But you could make Web IDL do that for you by converting the created function to a Web IDL callback type (Web IDL's `Function` type should work) then using Web |
| 16:23 | <Domenic> | IDL's "invoke a callback function" algorithm. |
| 16:24 | <wanderview> | JakeA: I'd just love to have a proposal we can vote on at the April meeting if we can |
| 16:32 | <jgraham> | Domenic: OK, thanks, will suggest that |
| 16:38 | <annevk> | wanderview: would prefer async agreement |
| 16:39 | <wanderview> | annevk: well, I meant async agreement as much as we can, but then with other browser vendors in the room agree we are moving forward if we can |
| 17:40 | <Ms2ger> | For people who are interested in script execution stuffs and hadn't seen it yet: https://bugzilla.mozilla.org/show_bug.cgi?id=1340145 |
| 17:46 | <TabAtkins> | JonathanNeal: The sectioning roots are all elements whose semantics mean that they can reasonably be given a heading. Versus other elements that are all pieces of text or other info *within* a section. |
| 17:47 | <TabAtkins> | JakeA: Yeah, just start writing stuff down right away. `bikeshed template > index.bs`, and start hacking, it's just the easiest way to record stuff. |
| 17:48 | <annevk> | "I understand all that" proceeds with not understanding |
| 17:49 | <JakeA> | wanderview: promise to have it by then. Still aiming for tomorrow |
| 17:49 | <JakeA> | TabAtkins: i did https://rawgit.com/WICG/background-fetch/master/ |
| 17:50 | <TabAtkins> | JakeA: Set your github.io branch to be master, stop using rawgit. ^_^ |
| 17:53 | <JakeA> | TabAtkins: yes, thanks tabsplainer, I don't have permission to do that on the WICG, but I've requested it. |
| 17:53 | <JakeA> | 😀 |
| 17:53 | <TabAtkins> | ur welcome |
| 20:47 | <felixjet> | when i insert "UI text", like for example: "We are hiring, join us." (and a <a> tag after that with a call to action link), should i put that phrase between <p>? Because i'm using <div> or <span> most of the time, and only using <p> when it's not "UI text" |
| 20:47 | <TabAtkins> | Semantically, paragraphs are implicit; <p> just helps mark them up when you have several in succession, and aids in styling. |
| 21:11 | <felixjet> | oh |
| 21:29 | <zcorpan> | annevk: fyi https://github.com/w3c/webauthn/issues/354 |
| 21:32 | gsnedders | wonders if anyone has modelled WebAuthn in ProVerif |
| 21:33 | <gsnedders> | Because I really feel like core security stuff that we're baking into browsers at this point should be proven secure to start with. |
| 21:34 | <gsnedders> | Yes, that doesn't stop side-channel attacks, and no, it will never be complete because people will discover new attacks. |
| 21:52 | <jyasskin> | gsnedders: That seems like a good request to file a github issue about. I tend to trust the security folks who did FIDO, but a formal verification can't hurt. |
| 21:56 | <gsnedders> | jyasskin: INRIA did stuff on FIDO, but ofc the Web Authn isn't exactly that |
| 21:56 | <gsnedders> | AFAIK INRIA are doing stuff on the W3C stuff |
| 21:57 | <gsnedders> | but I'd *love* to see some of this presented as part of the moving to PR stage |
| 21:57 | <jyasskin> | +1. |