| 00:57 | <MikeSmith> | Domenic: about https://github.com/whatwg/html/pull/1958/files/cfb8f34614ec51ac16964d32c9c613e7c2e19b9d..cb322d90333b2362252df305360aa323a94f6e77 we can get the URL for the `origin` remote with just `git config remote.origin.url` |
| 00:58 | <MikeSmith> | is there any reason to avoid that? |
| 00:58 | <Domenic> | MikeSmith: nope that's way better |
| 00:59 | <MikeSmith> | k |
| 01:10 | <MikeSmith> | Domenic: another question: about URLs for remotes for PR branches, is there a reason to favor or GH https URL over GH SSH URLs? |
| 01:11 | <MikeSmith> | I personally do not have my git environment set up right for auth to work with GH https URLs |
| 01:11 | <KiChjang> | MikeSmith, i hear HTTPS is more secure, but i also hear the other way |
| 01:11 | <MikeSmith> | KiChjang: OK |
| 01:11 | <MikeSmith> | well I just wonder which is more commonly used |
| 01:12 | <MikeSmith> | or if we can programatically detect which the user is actually set up for |
| 01:12 | <Domenic> | MikeSmith: https://help.github.com/articles/which-remote-url-should-i-use/ |
| 01:12 | MikeSmith | looks |
| 01:12 | <Domenic> | Also http://stackoverflow.com/questions/11041729/why-does-github-recommend-https-vs-ssh |
| 01:13 | <Domenic> | Shouldn't your Git config trick detect what is currently used? |
| 01:13 | <MikeSmith> | ah “These URLs work everywhere--even if you are behind a firewall or proxy.” |
| 01:13 | <MikeSmith> | Domenic: yeah it will for the origin remote |
| 01:13 | <MikeSmith> | ah but I get your point |
| 01:14 | <MikeSmith> | so yeah I can have it use for other added remotes whatever the protocol is used for the origin remote |
| 01:14 | <MikeSmith> | thanks |
| 01:51 | <MikeSmith> | Domenic: one last question on the bash snippet for PR thing: any reason to avoid adding --track to the `git checkout` command? |
| 01:52 | <MikeSmith> | if I switch the advice to saying the local branch name should be prefixed with teh author name, then that --track will allow git to know how to map the local branch name o the remote branch name |
| 01:52 | <MikeSmith> | for fetching |
| 01:55 | <MikeSmith> | or maybe git does that automatically? |
| 01:55 | MikeSmith | tests |
| 01:56 | <MikeSmith> | yup nm, I see git does it automatically |
| 05:36 | <GPHemsley> | Ms2ger: I'm mostly hands-off on mimesniff now, but I do recall seeing something recently about getting real mime types for fonts. annevk is more up-to-date on things than I am. |
| 07:51 | <zcorpan> | i suppose it's time i retrain my fingers to type => and const |
| 07:52 | <annevk> | I got => down, but const/let not yet |
| 07:53 | <zcorpan> | i guess we need to fork live dom viewer also since a global const is still taken in the next edit, at least in chromium |
| 07:54 | <zcorpan> | or global let |
| 08:02 | <tobie> | So I need to rebase https://github.com/heycam/webidl/pull/205. Won't that totally ruin the reviews? Is there a better strategy? |
| 08:05 | <tobie> | Is merge a better strategy now that we have squash in the GH interface? |
| 08:10 | <annevk> | tobie: not really, no, no |
| 08:10 | <tobie> | no, no to what? |
| 08:10 | <annevk> | tobie: three question marks, three answers |
| 08:11 | <tobie> | ok |
| 08:11 | <tobie> | annevk: thanks |
| 08:11 | <annevk> | tobie: it's worth it though to ask again to make sure the rebase didn't go wrong |
| 08:12 | <hsivonen> | anyone have an idea of what the median length of a URL query string is ignoring zero-length query strings? |
| 08:13 | <tobie> | annevk: yup |
| 08:13 | <annevk> | hsivonen: I don't recall ever seeing those kind of numbers |
| 08:14 | <hsivonen> | annevk: thanks. |
| 08:15 | <annevk> | tobie: you can't really avoid rebase / force push I think if you also want a linear commit history |
| 08:16 | <annevk> | tobie: you can sometimes avoid it due to squashing, if you fix the thing it's breaking on in a new commit, but that's rather hairy |
| 08:17 | <hsivonen> | I think I'm going to give up on trying to find representative encoder workloads and am going to to the simplest thing, which is running the decoder workload in reverse |
| 08:17 | <hsivonen> | if the workload is arbitrary, that workload qualifies |
| 08:18 | <hsivonen> | (the decoder workload being the Wikipedia article for apple, the fruit) |
| 08:18 | <tobie> | annevk: I pretty much only rebase, so I don't really know what the common outcome of merging in such cases is. Is it hairy? |
| 08:20 | <hsivonen> | (apple, the fruit, was the most cross-culturally-relevant Wikipedia topic that I was able to think of) |
| 08:20 | <annevk> | tobie: I'm not super familiar with merging, other than that it always seems to mess up history and create super long sequences of commits |
| 08:20 | <hsivonen> | (to find non-stub articles in various languages, that is) |
| 08:20 | <tobie> | annevk: oh, that, for sure. But with squashing in the UI, it's no longer an issue. |
| 08:21 | <annevk> | tobie: yeah, though it's still a little ugly then as GitHub will create all these links if those other commits that are pulled in linked anywhere |
| 08:21 | <annevk> | hsivonen: isn't the goal simply to be as fast or faster than Gecko's implementation for now? |
| 08:21 | <tobie> | annevk: oh--that's the part I had no experience with. Makes sense. |
| 08:22 | <annevk> | hsivonen: or have you set a higher bar for yourself? |
| 08:22 | <zcorpan> | tobie: http://www.bitsnbites.eu/?p=221 |
| 08:23 | <zcorpan> | hmm maybe that was not relevant to your question |
| 08:23 | <hsivonen> | annevk: I think as fast as uconv for decode (except ISO-2022-JP and UTF-16) and for UTF-16-to-UTF-8 encode is the minimum bar |
| 08:24 | <hsivonen> | annevk: even faster would be nice, though |
| 08:25 | <zcorpan> | tobie: see https://github.com/w3c/webvtt/pull/201 |
| 08:25 | <hsivonen> | annevk: that is, being faster than ICU and Windows would be good |
| 08:25 | <hsivonen> | annevk: since that'd mean being faster than other browsers |
| 08:26 | <annevk> | hsivonen: and another reason for folks to consider Rust |
| 08:26 | <hsivonen> | annoyingly, the Vietnamese article for apple is rather short |
| 08:26 | <hsivonen> | Turkish, too |
| 08:27 | <hsivonen> | but those are relevant languages to test, because they are SSE2-unfriendly |
| 08:27 | <hsivonen> | (a language is SSE2-unfriendly, if it's a Latin-script language where an accented character occurs every 16 characters or more frequently) |
| 08:31 | <zcorpan> | hsivonen: try https://en.wikipedia.org/wiki/World_War_II |
| 08:34 | <hsivonen> | zcorpan: I was trying to go for a non-controversial topic. it's rather sad if war is what's globally relevant |
| 08:35 | <hsivonen> | zcorpan: but yeah, WWII seems to have inspired longer articles than apple in e.g. various languages of India |
| 08:38 | <annevk> | hsivonen: Food is slightly longer in Vietnamese |
| 08:39 | <annevk> | But World War II is indeed far more impressive length-wise |
| 08:39 | <annevk> | Or you could use some API to download a lot more articles |
| 08:40 | <zcorpan> | i tried Love at first but it was short in Vietmanese :-( |
| 08:46 | <tobie> | zcorpan: oh, yeah, that's ugly. |
| 08:49 | <tobie> | zcorpan: hadn't thought about how that would add all the intermediary commits to the PR. |
| 08:57 | <hsivonen> | even basic elements like gold and iron have rather uneven Wikipedia articles |
| 08:58 | <hsivonen> | going with WWII seems very sad, but it might be the most even-length topic without putting too much effort into finding one :-( |
| 08:59 | <MikeSmith> | hsivonen: if you want big files you can get abstract files in all langs from https://dumps.wikimedia.org/backup-index.html |
| 09:00 | <MikeSmith> | the latest ones for each lang are at URLs, e.g., https://dumps.wikimedia.org/zhwiki/latest/zhwiki-latest-abstract.xml |
| 09:00 | <MikeSmith> | but those are massive |
| 09:00 | <MikeSmith> | that one is 905.4 MB |
| 09:01 | <hsivonen> | MikeSmith: I don't want big files. I want a normal Wikipedia article in at least ar, el, en, he, ja, ko, ru, zh-cn, zh-tw, vi, tr, fr, de |
| 09:01 | <hsivonen> | MikeSmith: bonus points for th, cs and pl |
| 09:01 | <MikeSmith> | Krinkle (who works on Mediawiki) might have some ideas |
| 09:01 | <hsivonen> | MikeSmith: in fact, the WWII article may have the downside of being abnormally long |
| 09:01 | <MikeSmith> | ah OK |
| 09:02 | <hsivonen> | what I'm looking for is a Web-representative interleave of ASCII and non-ASCII |
| 09:02 | <MikeSmith> | ah |
| 09:02 | <hsivonen> | Wikipedia is an actual top site, suitably licensed and has content in multiple languages |
| 09:03 | <hsivonen> | generating test cases using Google Translate doesn't work, because it injects its own markup, which messes up the natural interleave of ASCII and non-ASCII |
| 09:03 | <hsivonen> | the declaration of human rights doesn't work, because it has abnormally little markup |
| 09:03 | <MikeSmith> | why not use just the home page of each lang? |
| 09:04 | <MikeSmith> | or some other stock page that the Wikipedia for each lang is guaranteed to have |
| 09:04 | <hsivonen> | MikeSmith: I was thinking an article would have a more representative interleave of ASCII and non-ASCII than the front page |
| 09:04 | <hsivonen> | MikeSmith: but I didn't think of policy stock pages |
| 09:04 | <MikeSmith> | ok |
| 10:38 | <hsivonen> | FWIW, I posted to Wikipedia Village Pump: https://en.wikipedia.org/wiki/Wikipedia:Village_pump_(technical)#Finding_representative_pages_in_various_languages |
| 10:43 | <nox> | hsivonen: https://github.com/whatwg/html/issues/1954 |
| 10:45 | <nox> | hsivonen: What about one of https://en.wikipedia.org/wiki/Wikipedia:Wikipedia_records#Articles |
| 11:16 | <hsivonen> | nox: those seem like outliers instead of typical articles. also, it doesn't help with other languages |
| 11:16 | <hsivonen> | I commented on github regarding recursion |
| 11:21 | <Ms2ger> | I wonder if there's a way to get non-minified sources from google docs |
| 11:37 | <tobie> | annevk: re https://github.com/heycam/webidl/pull/211, what about nullable types? |
| 11:39 | <annevk> | tobie: I haven't seen [NewObject] used for something that returns either an object or null |
| 11:39 | <tobie> | annevk: so (DOMString or DOMException?) wouldn't be an object type |
| 11:40 | <annevk> | tobie: I see, hmm, maybe requiring unions to contain object types is simpler for now |
| 11:40 | <annevk> | tobie: bz prolly has thought this through already so might be easiest to just wait a little for him to wake up |
| 11:40 | <tobie> | annevk: :) |
| 12:25 | <nox> | zcorpan: Nice, thank you! |
| 12:25 | <zcorpan> | nox: np |
| 12:26 | <zcorpan> | i managed to crash the tab while writing that so i had to write it again :-P |
| 12:26 | <annevk> | smaug: feedback on https://github.com/whatwg/html/issues/1576 would be appreciated |
| 12:28 | <annevk> | zcorpan: Live DOM Viewer uses document.write itself, right? |
| 12:28 | <annevk> | zcorpan: might not be the best way to test the limit |
| 12:28 | <nox> | zcorpan: Nice. :D |
| 12:31 | <nox> | zcorpan: I wonder if it's supposed to be 10 for Edge, |
| 12:31 | <nox> | but they have an off by one error that makes it 11. |
| 12:32 | <zcorpan> | annevk: it does, yeah |
| 12:33 | <zcorpan> | the numbers in my test may be confusing. i think the limit is the given number + 2 |
| 12:57 | <Ms2ger> | zcorpan, have a minute? |
| 12:58 | <zcorpan> | sure |
| 12:58 | <Ms2ger> | zcorpan, do you know off-hand if gecko/chrome are correct in this test case: https://github.com/servo/servo/issues/13944 ? |
| 13:03 | <zcorpan> | Ms2ger: i believe they are correct. offsetHeight should flush layout (that's not actually defined yet though) |
| 13:03 | <zcorpan> | Ms2ger: does Servo behave differently if the 'top' value is different? |
| 13:12 | <Ms2ger> | zcorpan, oh, interesting, explicitly adding display: block helps |
| 13:13 | <zcorpan> | i guess you found what the bug is then :-) |
| 13:13 | <Ms2ger> | I found enough to make it SEP :) |
| 13:14 | <Ms2ger> | position: absolute should force a block-level box, should it not? |
| 13:16 | <zcorpan> | yes |
| 13:17 | <zcorpan> | https://drafts.csswg.org/css-position/#dis-pos-flo |
| 13:18 | <zcorpan> | in this case at least; not for inline-table or "others" (not sure which those are) |
| 13:20 | <nox> | zcorpan: I took the permalink of your test, |
| 13:20 | <nox> | and loaded data:text/html,... instead, |
| 13:20 | <nox> | and indeed, it reported one more call. |
| 13:27 | <annevk> | heycam|away: can you add a link to the editor's draft at the top of https://github.com/heycam/webidl |
| 15:02 | <gsnedders> | do we have any API that parses arbitrary JSON off the network anywhere? |
| 15:09 | <Ms2ger> | fetch().then(|x| x.json())? |
| 15:09 | <Ms2ger> | I don't suppose that's what lambdas look like in JS |
| 15:10 | <caitp> | you can just hear the JS syntax nuts growling as they read this |
| 15:19 | <annevk> | there's also an XMLHttpRequest one |
| 15:19 | <annevk> | you get to chose |
| 15:26 | <gsnedders> | annevk: an XHR one? |
| 15:26 | <gsnedders> | annevk: oh, with .response |
| 15:26 | <annevk> | responseType = "json" |
| 15:27 | <annevk> | used to be pretty cool |
| 15:29 | <gsnedders> | and it only supports utf-8 |
| 15:31 | <annevk> | same for fetch() |
| 15:32 | <gsnedders> | is there any way to detect support of <a ping>? |
| 15:34 | <Ms2ger> | "ping" in document.createElement("a"), theoretically |
| 15:36 | <gsnedders> | what does Fx say with the pref disabled? |
| 17:29 | <Domenic> | Anyone have examples of repeated initialisms in APIs? So far we have XMLHttpRequest (inconsistent), HTMLHRElement (consistent), RTCDTMFSender (consistent). I was hoping for some that are properties not just class names. |
| 17:30 | <wanderview> | JakeA: shouldn't the throw() in the fetch event promise.catch here trigger an interception failure? https://googlechrome.github.io/samples/service-worker/prefetch/service-worker.js |
| 17:49 | <wanderview> | JakeA: I ask, because that service worker in chrome produces a "page not found" error page instead of the "something went wrong" page you normally get for a broken SW interception |
| 17:50 | <wanderview> | JakeA: the top level page for the SW is this: https://googlechrome.github.io/samples/service-worker/prefetch/index.html |
| 18:41 | <Domenic> | annevk: https://github.com/whatwg/url/commit/f20ad7efa8514ca88ceab0c3daad025ab0beb80e doesn't seem to have deployed :-/ |
| 18:52 | <annevk> | Domenic: done manually |
| 18:52 | <annevk> | I really wonder why this fails so often |
| 18:52 | <annevk> | The setup is super basic |
| 19:14 | <Domenic> | TabAtkins: given a lt of "becomes connected" can Bikeshed figure out <a>become connected</a>, or should I add alternate lt=""s? |
| 19:24 | <tobie> | Domenic: the latter, unfortunately. |
| 19:24 | <Domenic> | ah well |
| 19:31 | <Domenic> | botie: ask rniwa if he's writing tests for the [CEReactions] on elements/document.close/etc.? |
| 19:31 | <botie> | will do |
| 19:33 | <tobie> | Does WebIDL specify what happens when you pass to a function an object with properties which aren't members of the dictionary said function expects as argument? |
| 19:36 | <tobie> | so namespace Bar { foo(optional Options options); } dictionary Options { DOMString a = "test"; } |
| 19:36 | <tobie> | and then: Bar.foo({ a: "test", b: 123 }); |
| 19:37 | <tobie> | Does this throw, is b ignored, or is this spec-dependant (and if so, are there best practices)? |
| 19:40 | <Domenic> | b is ignored |
| 19:42 | <Domenic> | tobie: The conversion from ES object to Web IDL dictionary is at https://heycam.github.io/webidl/#es-dictionary; as you can see it loops through the members of the dictionary doing a Get() for each one, but doesn't do any enumeration of the object in general |
| 19:42 | <Domenic> | I guess right now it does a [[Get]]; should be Get() |
| 19:47 | <tobie> | Domenic: oh right, that's specific to the ES conversion. Makes sense. Thanks. |
| 19:48 | <TabAtkins> | Domenic: Yeah, I'm a little afraid of applying the conjugation code to non-final words. Possible, tho. |
| 19:49 | <TabAtkins> | It would mean an exponential (in the number of words) explosion in the number of combinations I try for every single autolink. |
| 19:50 | <Domenic> | I feel like just the first word would be a sweet spot. But I can see the difficulty yeah. |
| 19:55 | <TabAtkins> | I could be convinced by someone gradually accumulating some examples of failed conjugations in an issue. |
| 19:59 | <Domenic> | People seem to just create a bunch of link-defaults |
| 19:59 | <Domenic> | I think this is mostly HTML's fault |
| 19:59 | <Domenic> | Because people are used to creating giant link-default blocks for HTML |
| 20:00 | <Domenic> | so they just reach for link-defaults blocks whenever Bikeshed isn't working perfectly |
| 21:04 | <Domenic> | For an enum default value meaning "auto", do people have opinions/precedents for whether we'd use `null` or `""`? |
| 21:06 | <KiChjang> | missing default value of "auto"? |
| 21:06 | <KiChjang> | or invalid default value of "auto"? |
| 21:14 | <Domenic> | It's an IDL enum, not a element attribute, so both I guess |
| 21:16 | <KiChjang> | it doesn't quite feel right if you use "null" or "" for "auto" |
| 21:17 | <KiChjang> | because the syntax is obviously different, so intuitively to me, they mean different things |
| 21:18 | <KiChjang> | null means more like a value is missing |
| 21:18 | <KiChjang> | "" means this enum variant is the empty string, but the value isn't missing |
| 21:20 | <KiChjang> | what the empty string means will vary in different contexts though, so i'm not feeling too great about including it as an enum variant either |
| 21:38 | <zcorpan> | Domenic: https://fetch.spec.whatwg.org/#enumdef-requesttype |
| 21:38 | <Domenic> | zcorpan: ooh good one, thanks |