| 02:12 | <Domenic> | tobie: I think we have |
| 08:52 | <annevk> | tobie: Domenic: maybe IDL should start requiring tests for any changes too and bugs? |
| 08:52 | <annevk> | tobie: Domenic: it's at such a low-level of the stack that it seems important to keep it all together |
| 08:54 | <tobie> | annevk: I have no idea what the state of the WebIDL test suite is in, nor frankly how to properly test most of what's in the WebIDL spec. |
| 08:54 | <annevk> | tobie: we can test many things indirectly through actual APIs defined through the IDL feature |
| 08:55 | <tobie> | annevk: so we should keep idlharness in sync? |
| 08:55 | <annevk> | tobie: at the very least that seems like a good thing, no? |
| 08:58 | <tobie> | annevk: I was doing TDD when pretty much only Rubyist were into it, so you're preaching to the choir here… |
| 08:59 | <tobie> | annevk: last time I looked, the internals of idlharness were quite a sight |
| 09:00 | <tobie> | annevk: so I'm concerned this could be a pretty significant investment |
| 09:01 | <annevk> | tobie: discuss it with Domenic at least? It's really been paying off for WHATWG work |
| 09:01 | <Ms2ger> | Yeah :/ |
| 09:02 | <annevk> | tobie: the main problem is that by not doing it, we hit these very subtle issues down the road because browsers didn't follow a change |
| 09:02 | <annevk> | tobie: perhaps at the very least we should make sure each change has corresponding bugs in the four major browsers, or are you doing that already? |
| 09:02 | <tobie> | annevk: dude, you're preaching to the choir |
| 09:02 | <tobie> | convince Domenic it's a good use of my time. |
| 09:02 | <annevk> | tobie: and then maybe the browsers can take care of writing the tests, due to limited resources on the standards side |
| 09:03 | <annevk> | tobie: sorry, don't mean to patronize |
| 09:03 | <tobie> | annevk: oh, no, no. I'm just saying, I don't need convincing. |
| 09:04 | <tobie> | annevk: we are filing bugs against the browsers, but not for all changes. |
| 09:05 | <tobie> | annevk: I still have little visibility into what parts of WebIDL are properly implemented by browsers and which parts aren't at this point. |
| 09:11 | <tobie> | annevk: , Ms2ger: comments/thoughts welcomed at https://github.com/heycam/webidl/issues/268 |
| 09:26 | <annevk> | In a single promise test, how can you do the equivalent of three assert_throws? |
| 11:02 | <hsivonen> | annevk: thanks for merging |
| 11:02 | <hsivonen> | annevk: fwiw, the improvement in EUC-KR encode that's attributable to looking at the visualizations is *massive* |
| 11:02 | <annevk> | hsivonen: thanks for writing it all! |
| 11:03 | <hsivonen> | annevk: on raspberry pi 3, it removed about 99.5% of the running time of the benchmark! |
| 11:04 | <hsivonen> | (that's for the EUC-KR benchmark) |
| 11:04 | <hsivonen> | for gb18030 and Japanese encode it was a win, too, but not as massive |
| 11:04 | <hsivonen> | the corresponding decode-site perf degradation was modest |
| 11:06 | <annevk> | hsivonen: wow, guess you need a new benchmark then |
| 11:06 | <annevk> | hsivonen: to find the new hot spots |
| 16:01 | <tobie> | Domenic: https://github.com/heycam/webidl/issues/222#issuecomment-274108164 |
| 16:10 | <annevk> | tobie: I kinda like it |
| 16:10 | <annevk> | tobie: might even drop the parenthetical in the second example |
| 16:11 | <annevk> | tobie: StructuredClone doesn't have those either |
| 16:11 | tobie | looks |
| 16:12 | <annevk> | tobie: you'll soon run into https://github.com/heycam/webidl/issues/97 I think |
| 16:13 | <tobie> | annevk: indeed. :D |
| 16:13 | <tobie> | annevk: didn't know it was a thing, but was wondering about this minutes ago |
| 16:13 | <annevk> | tobie: figures |
| 16:13 | <annevk> | tobie: fixing that would be great |
| 16:14 | <annevk> | tobie: having a way to say whether something is an Element that is actually grounded in primitives would be a huge improvement |
| 16:15 | <tobie> | annevk: yup |
| 16:16 | <tobie> | annevk: would you: Let |T| be Type(|V|) at the beginning of algos or not? |
| 16:17 | <annevk> | tobie: seems okay, I haven't read enough ECMAScript to know if it has precedent |
| 16:17 | <tobie> | Also, any idea why some variables are uppercase and others lowercase? |
| 16:18 | <annevk> | tobie: doesn't seem like it (searching for "be type(") |
| 16:18 | <annevk> | tobie: hmm examples? |
| 16:24 | <tobie> | annevk: https://heycam.github.io/webidl/#dfn-obtain-unicode |
| 16:27 | <annevk> | tobie: no idea, prolly mistake? |
| 16:27 | <annevk> | tobie: longer variable names wouldn't hurt either |
| 16:28 | <tobie> | annevk: heh |
| 16:39 | <annevk> | wanderview: re: https://github.com/whatwg/html/issues/2273 I'm pretty sure Chrome has a design where a window.open() could result in a new process, they have it at least for <iframe> |
| 16:39 | <annevk> | wanderview: and they proxy things back and worth |
| 16:39 | <wanderview> | annevk: has that shipped? also I thought rel=noopener was needed for window.open() |
| 16:40 | <wanderview> | but historically anyway this has prevented separate processes for windows in the same "constellation" |
| 16:41 | <annevk> | wanderview: not sure, but they did find a way that if things cannot reach each other (just reach the global/location object) to use different processes |
| 16:41 | <annevk> | wanderview: which you can observe it through URL fragment manipulation and timeouts, but maybe nobody uses that anymore |
| 16:42 | <wanderview> | dunno... just pointing out in the issue that a multi-process browser can still keep some things in the same process if necessary |
| 16:43 | <wanderview> | anyway, I have to go get lunch |
| 16:45 | <annevk> | bon appetit |
| 17:06 | <annevk> | Domenic: https://github.com/whatwg/url/pull/185 is ready for review again I think |
| 17:06 | <annevk> | Domenic: I updated the patch and tests with the aim to just do the host changes |
| 17:07 | <annevk> | Domenic: I haven't run a modified whatwg-url to check things, since as I'm still looking for advice on how to run it against different tests |
| 17:08 | <annevk> | Domenic: thanks |
| 17:20 | <annevk> | Domenic: nice system |
| 17:20 | <Domenic> | :) |
| 17:21 | <annevk> | Domenic: only remaining failures are setter tests and I'm not really sure what to do there yet |
| 17:21 | <annevk> | Domenic: investigating |
| 17:25 | <annevk> | tobie: can we stop adding wg-webplatform labels to url tests? |
| 17:25 | <annevk> | tobie: although maybe you should first tell me what the label is for before I destroy something |
| 17:26 | <tobie> | tobie: iirc Dom was responsible for that code. |
| 17:27 | <tobie> | annevk: iirc Dom was responsible for that code. |
| 17:28 | <tobie> | annevk: thing it was something to do with letting chairs figure out what PRs needed reviews. |
| 17:30 | <tobie> | annevk: let me see if I can fix that in specref |
| 17:53 | <tobie> | annevk: that looks easily fixable |
| 17:53 | <tobie> | annevk: and also seems to be a proper way to fix aliasing of old W3C specs to WHATWG ones. |
| 17:54 | <annevk> | tobie: that PR looks fine, hopefully one day "URL" can be the authoritative term and WHATWG-URL the alt |
| 17:54 | <tobie> | annevk: that PR makes it so for all practical purposes |
| 17:54 | <annevk> | tobie: coolio |
| 17:55 | <tobie> | annevk: both Respec and Bikeshed resolve aliases for you |
| 17:55 | <tobie> | annevk: so you can now do [URL] |
| 17:55 | <tobie> | annevk: and that'll get you the WhatWG version |
| 17:55 | <annevk> | tobie: I think TabAtkins put in some local overwrites already for us so we can |
| 17:56 | <annevk> | tobie: but maybe those were not exposed to everyone |
| 17:56 | <tobie> | annevk: dunno |
| 17:56 | <tobie> | annevk: might as well do the other ones while I'm at it |
| 17:56 | <tobie> | annevk: what were they? |
| 17:57 | <tobie> | annevk: DOM, FULLSCREEN, what else? |
| 17:57 | <TabAtkins> | No, they're global, but the list is relatively short |
| 17:57 | <TabAtkins> | One sec |
| 17:57 | <annevk> | tobie: https://wiki.whatwg.org/wiki/Fork_tracking has an overview |
| 17:59 | <jyasskin> | tobie: Have you looked at all at whether WASM needs any WebIDL changes? |
| 17:59 | <TabAtkins> | ["notifications", "fullscreen", "dom", "url", "encoding"] |
| 17:59 | <tobie> | TabAtkins: thanks |
| 17:59 | <TabAtkins> | https://github.com/tabatkins/bikeshed/blob/master/bikeshed/ReferenceManager.py#L583 |
| 18:00 | <tobie> | jyasskin: my understanding is it'll need completely new bindings at some point. |
| 18:02 | <annevk> | yup, I actually think they plan on writing that spec |
| 18:02 | <annevk> | though we should probably start coordinating at some point if we're still going to not have distinct IDL and JavaScript types for some things |
| 18:03 | <tobie> | jyasskin: but until then, they just call DOM APIs through JS. More here: http://webassembly.org/docs/gc/ |
| 18:04 | <jyasskin> | tobie: Mhmm, makes sense. |
| 18:05 | <tobie> | jyasskin: and in particular: http://webassembly.org/docs/gc/#webidl-integration |
| 18:07 | <jyasskin> | Yeah. I don't need it for anything I'm doing, but I wanted to make sure that they'd made contact in the right ways. |
| 18:28 | <nox> | Shouldn't the 2 arguments version of document.open be [SameObject]? |
| 19:22 | <annevk> | Turns out you can get DreamHost SSH keys out-of-band, at https://panel.dreamhost.com/index.cgi?tree=users.sshkeys, though even that suggests it's safe to ignore the warning |
| 19:23 | <annevk> | Anyone know if the dangers of ignoring SSH warnings are outlined anywhere? I feel like maybe I should write a blogpost to point out the obvious |
| 19:30 | <Domenic> | I really feel like we must be missing something... maybe writing a blog post will be a good way to find out, as technies descend on us and tell us why we're wrong :) |
| 19:31 | <annevk> | Domenic: pretty sure I'm not wrong |
| 19:32 | <annevk> | Domenic: there's a reason the ssh program says "Warning" and such |
| 19:33 | <annevk> | Domenic: it's just that TOFU used to be acceptable for some reason and no longer is |
| 19:33 | <Domenic> | I dunno, maybe it's just old and bogus or defending against an unrealistic threat model or something. |
| 19:33 | <Domenic> | You're probably right, but I try to at least keep my mind open that so many people aren't wrong. |
| 19:34 | <annevk> | Domenic: for what we do it doesn't matter, we just push public data somewhere, if that somewhere happens to be an attacker, all we lose is that the standard isn't updated |
| 19:34 | <annevk> | Domenic: although setting a bad example is bad of course |
| 19:34 | <annevk> | oh well, yes, I should write a blog post and we'll find out |
| 21:10 | <Domenic> | TabAtkins: ping on https://github.com/tabatkins/bikeshed/issues/900 ? |
| 21:10 | <botie> | will do |
| 21:11 | <Domenic> | botie no. botie why. |
| 21:11 | <botie> | Domenic: sorry... |