| 03:12 | <hgl> | i was testing if changing a range object (e.g., insert a node) returned by selection should change the selection, and found out browsers behave differently on this one. Should the spec codify this? |
| 03:38 | <MikeSmith> | hgl: yeah. But which spec? The Selection API spec? |
| 03:38 | <hgl> | I believe so, already opened an issue: https://github.com/w3c/selection-api/issues/25 |
| 03:39 | <hgl> | hope that's the correct place. |
| 03:42 | <MikeSmith> | hgl: yeah I think that's the best |
| 03:43 | <hgl> | great, thank you. |
| 03:43 | <MikeSmith> | hgl: also, rniwa's usually around here during us/west hours |
| 03:44 | <hgl> | cool, would be great if i get to talk to him. :) |
| 03:45 | <MikeSmith> | it's only 7:45pm in Cupertino so he may still drop in after he stops by his favorite taco place for dinner :) |
| 03:47 | <MikeSmith> | Ben Peters is doing a lot of related spec work on the rich-textediting-related stuff, but I see he's following https://github.com/w3c/selection-api so he'll see your issue |
| 03:49 | <hgl> | Thank you for heads up. You know i actually have another question relate to rich text editing, but not sure if it should be in the spec, can i discuss with you here? |
| 03:49 | <MikeSmith> | hgl: you can always send messages about issues and comments to public-webapps⊙wo too. Main advantage there is just that a few more people with possible insights will see it. |
| 03:50 | <hgl> | i like irc, especially when get quick response. :) but will try out the mail list. thanks |
| 03:50 | <MikeSmith> | you can discuss it with me but I'm not keeping up clostely with the current work, so the odds are low that I'll be much help |
| 03:51 | <MikeSmith> | hgl: yeah I prefer IRC too and this is definitely the best place to hang out and ask things. But prime time for more people being around here is a few hours earlier than now, when Europe is still awake |
| 03:53 | <MikeSmith> | hgl: you can always just ask here -- drop your question in and you may not get a response right away but a lot of people here read scrollback and the logs so you may get an answer later (if you're still here or the next time you /join) |
| 03:53 | <hgl> | ok. i was wondering if there is a way to get the pixel location of a collapsed selection (i.e., caret). the only way i can think of is to insert an empty span element and calc it's boundingrect, since the range is empty. do you think the spec should make it easier? |
| 03:54 | <hgl> | MikeSmith, that's great, i will definitely hang out here more. |
| 03:56 | <hgl> | for a non-empty selection, i can use getBoundingClientRect() on the range object, very convenient, but if the selection is collapsed, i can no longer use this method. |
| 03:57 | <MikeSmith> | the way you describe his how I've seen others doing it. I'm not sure how much easier it could be made but anything's possible :) Anyway, it's worth asking more people about -- here or elsewhere. At the very least you'll eventually get an explanation about why it's not practical to provide some easier means (if that's the case) -- which wouldn't be totally satisfying but at least would clarify things for |
| 03:57 | <MikeSmith> | you some |
| 03:57 | <MikeSmith> | hgl: right |
| 03:57 | <MikeSmith> | I remember a very recent discussion here about something similar |
| 03:57 | <MikeSmith> | maybe roc knows |
| 03:58 | <roc> | Range.getBoundingClientRect should work |
| 03:59 | <roc> | it should return a meaningful rectangle even for a collapsed selection, if the collapsed selection is in a text node. |
| 03:59 | <hgl> | if the range is collapsed, return rect is 0,0,0,0 |
| 03:59 | <roc> | which browser? |
| 03:59 | <hgl> | i have tested it. |
| 03:59 | <hgl> | safari |
| 04:00 | <roc> | got a link to a testcase? |
| 04:00 | <hgl> | let me set up a demo, wait a sec |
| 04:00 | <rniwa> | hgl: it’s defined in http://w3c.github.io/selection-api/#user-interactions |
| 04:00 | <roc> | I'm pretty sure that's a Webkit bug, and it should work in Firefox. |
| 04:00 | <rniwa> | hgl: the spec matches WebKit/Blink for now |
| 04:02 | <roc> | hgl: http://dev.w3.org/csswg/cssom-view/#dom-range-getclientrects |
| 04:03 | <roc> | note that if your collapsed range is not in a text node then you will get nothing. How caret works when you're not in a text node is very underspecified and not very interoperably implemented either... |
| 04:04 | <hgl> | ok, info avalanche, let me process it one by one. |
| 04:05 | <roc> | :-) |
| 04:06 | <hgl> | rniwa, "with suitable start and end", "suitable" here seems a bit vague. why not go with firefox and change the selection accordingly? seems more accurate and easy to define. |
| 04:07 | <rniwa> | hgl: I think there is a compat issue as far as I could recall. |
| 04:07 | <rniwa> | hgl: If I remember correctly, Gecko wanted to change their behavior. |
| 04:08 | <hgl> | ok, fair enough. |
| 04:08 | <rniwa> | also* wanted |
| 04:08 | <hgl> | rniwa, i guess, i could live with that behavior. |
| 04:10 | <hgl> | roc, please give me a minute to set up the demo. |
| 04:11 | <hgl> | rniwa, i should read the spec more carefully, shouldn't have missed that paragraph. :) |
| 04:12 | <rniwa> | hgl: no problem. |
| 04:12 | <rniwa> | hgl: i’ll close your issue then :) |
| 04:12 | <rniwa> | hgl: I file these bugs all the time as well |
| 04:12 | <rniwa> | some specs are hard to read |
| 04:12 | <hgl> | rniwa, please do. thanks. |
| 04:14 | <jarek> | btw, are there an plans to make getBoundingClientRect() return DOMRect instance? |
| 04:16 | <roc> | jarek: you mean in Webkit? |
| 04:16 | <jarek> | nvm, I just checked the spec |
| 04:16 | <jarek> | it is already mentioning DOMRect |
| 04:16 | <jarek> | roc: no, I meant in DOM standard |
| 04:16 | <roc> | yeah, it already does :-) |
| 04:18 | <hgl> | roc, http://jsbin.com/jafebipexa/3/edit |
| 04:18 | <hgl> | roc, click on the hello text, and see the 0 0 0 0 in output div |
| 04:19 | <hgl> | roc firefox doesn't have selectionchange event, so currently don't work on it. |
| 04:20 | <hgl> | roc, both safari and chrome return 0 0 0 0 |
| 04:21 | <hgl> | the problem with inserting a span element to calc the position is that it will trigger a selectionchange event |
| 04:21 | <roc> | if you trigger the code another way (e.g. setInterval) you'll see that getBoundingClientRect() works in FF |
| 04:22 | <roc> | I think the spec agrees |
| 04:23 | <hgl> | roc, yes you are right, so do you think it's a bug in webkit and blink? |
| 04:25 | <roc> | yeah, that is |
| 04:28 | <hgl> | roc, cool. glad to know the spec actually specified it. all i need to do now is to open an issue and figure out a polyfill, thanks. :) |
| 04:30 | <hgl> | MikeSmith, thanks again for the liaison. :) |
| 04:53 | <hgl> | now, flame suite on. if i got a question with a spec, should i go to whatwg or w3c? |
| 04:57 | <roc> | depends on the spec |
| 05:01 | <hgl> | all questions relate to these https://whatwg.org/specs/ should be submitted to whatwg, and all else to w3c? |
| 05:04 | <roc> | yeah I would |
| 05:05 | <hgl> | ok, got it. thanks. |
| 06:33 | <hsivonen> | Hixie, annevk: Could you, please, change the <a href> and <form action> references to html5.validator.nu from http to https at https://validator.whatwg.org/ ? |
| 06:35 | <Hixie> | done for validator.whatwg.org, any others? |
| 06:36 | <hsivonen> | Hixie: thanks. I'm not aware of others. |
| 06:36 | <Hixie> | coolio |
| 06:38 | <hsivonen> | confirmed: no more cross-scheme form submission warning when using https://validator.whatwg.org/ |
| 07:40 | <hsivonen> | annevk: http://webforms2.testsuite.org/ redirects to Test the Web Forward. Are the tests actually somewhere in the Web Platform Test git repo? |
| 07:45 | <annevk> | hsivonen: I'm not sure, I assume someone has created webforms2 tests |
| 08:10 | <hsivonen> | annevk: OK. I take it that you tests have vanished from the Web then |
| 08:10 | <hsivonen> | s/you/your/ |
| 08:11 | <hsivonen> | MikeSmith: https://about.validator.nu/ *finally* gives up-to-date build instructions (by linking to github instead of trying to offer them inline) |
| 08:11 | <hsivonen> | MikeSmith: also, I added clarifications that the terms of service and privacy policy do not apply to the W3C instance |
| 09:25 | <MikeSmith> | hsivonen: very coolー thanks |
| 09:28 | <annevk> | hsivonen: yeah, they didn't seem that useful anymore with all the automated tests |
| 09:28 | <MikeSmith> | hsivonen: btw I'd like to change validator.w3.org/nu/ over to TLS some time soon |
| 09:29 | <annevk> | hsivonen: kudos on getting TLS btw |
| 09:30 | <annevk> | foolip: really, we're going to bring back CDATASection? :-( |
| 09:30 | <Ms2ger> | Boo |
| 09:30 | <annevk> | foolip: it's not entirely clear to me why we'd restore the node and then not have createCDATASection return an instance of it |
| 09:40 | <annevk> | foolip: oh wait, it was one or the other |
| 09:52 | <zcorpan> | anyone know how web components work? https://www.w3.org/Bugs/Public/show_bug.cgi?id=27376 |
| 09:52 | <Ms2ger> | No |
| 09:56 | <zcorpan> | where is serializeAsCDATA now? |
| 10:01 | <Ms2ger> | Probably nowhere |
| 10:33 | <annevk> | zcorpan_: there's no CSS syntax for web components; that idea had a different name and got dropped |
| 10:33 | <annevk> | zcorpan_: decorators |
| 10:34 | <zcorpan_> | annevk: comment in the bug please :-) |
| 10:47 | <hsivonen> | annevk: I'm particularly happy to end the shame of running a site that had login without https (v.nu bugzilla) |
| 10:48 | <annevk> | zcorpan_: done |
| 10:48 | <zcorpan_> | thx |
| 10:52 | smaug____ | learns something new, rel=nofollow |
| 10:52 | <smaug____> | not that I understand the use case |
| 11:01 | <Ms2ger> | smaug____, search engines? |
| 11:09 | <zcorpan_> | Ms2ger: do you remember if serializeAsCDATA was for compat or just someone complained? |
| 11:16 | <mathiasbynens> | does `itemtype` accept any URL as vocabulary? |
| 11:16 | <mathiasbynens> | schema.org now supports TLS, so it would be nice to use https:// instead of http:// in the spec examples if that’s possible https://html.spec.whatwg.org/multipage/semantics.html |
| 11:17 | <Ms2ger> | zcorpan_, it seems plausible that serializing cdata sections as text would break things |
| 11:17 | <annevk> | mathiasbynens: you could file a bug |
| 11:17 | <Ms2ger> | zcorpan_, but nobody tried, so no idea if that's actually the case |
| 11:17 | <annevk> | mathiasbynens: changing the URL changes the meaning as well though |
| 11:17 | <Ms2ger> | zcorpan_, IIRC, glazou cared quite strongly |
| 11:18 | <zcorpan_> | Ms2ger: ok, i'll take that as "someone complained" :-) |
| 11:19 | <mathiasbynens> | annevk: that’s what i was wondering about – does it change the meaning if both URLs provide the same vocabulary/response |
| 11:23 | <mathiasbynens> | filed https://www.w3.org/Bugs/Public/show_bug.cgi?id=27388 |
| 11:23 | <annevk> | mathiasbynens: per spec it's just a string that happens to look like a URL |
| 11:23 | <annevk> | mathiasbynens: there's no fetching protocol involved afaik |
| 11:24 | <annevk> | mathiasbynens: we don't change http://www.w3.org/1999/xhtml either even though it's all sorts of wrong |
| 11:26 | <mathiasbynens> | i (naively) assumed vocabularies were somehow different |
| 11:26 | <mathiasbynens> | is there a canonical list of vocabulary identifiers/URLs? |
| 11:30 | <zcorpan_> | maybe we should change the namespace to https://n.whatwg.org/html |
| 11:48 | <annevk> | mathiasbynens: DNS? |
| 11:49 | <Ms2ger> | zcorpan_, why not about:blank? |
| 11:50 | <zcorpan_> | Ms2ger: it doesn't use TLS |
| 11:50 | <Ms2ger> | abouts:blank |
| 11:50 | <zcorpan_> | now we're talking |
| 11:50 | <mathiasbynens> | i meant vocabularies that are commonly supported by crawlers etc. |
| 12:32 | <zcorpan_> | hsivonen: maybe you can emit a warning in the validator when accessed over http? |
| 12:36 | <hsivonen> | zcorpan_: hmm. I wonder if API users would want to see an extra warning that doesn't pertain to what they are validating... |
| 12:37 | <zcorpan_> | hsivonen: of course they don't. which is why it might be a working incentive for them to switch to https before you force them to switch. they probably won't see your tweet |
| 12:43 | <jgraham> | zcorpan_: UnicodeEncodeError: 'charmap' codec can't encode character u'\xe5' in position 81: character maps to <undefined> |
| 12:44 | <jgraham> | From wpt html/infrastructure/urls/resolving-urls/query-encoding/resources/css-tmpl.py |
| 12:44 | <jgraham> | From /html/infrastructure/urls/resolving-urls/query-encoding/windows-1251.html |
| 12:44 | <zcorpan_> | hmm |
| 12:47 | <zcorpan_> | jgraham: filed https://github.com/w3c/web-platform-tests/issues/1403 i'll look at it next week |
| 13:26 | <buyoancing> | The website https://html.spec.whatwg.org/ has a SSL configuration that allows only RC4 (which is considered weak/broken) and has some other problems, see https://www.ssllabs.com/ssltest/analyze.html?d=html.spec.whatwg.org |
| 13:27 | <buyoancing> | "As of 2013, there is speculation that some state cryptologic agencies may possess the capability to break RC4 even when used in the TLS protocol. Microsoft recommends disabling RC4 where possible." |
| 13:33 | <buyoancing> | Is there someone here who can fix the configuration? |
| 13:34 | <buyoancing> | StartSSL also offers an intermediate certificate with SHA256 btw |
| 13:35 | <annevk> | buyoancing: DreamHost |
| 13:35 | <annevk> | buyoancing: Hixie could fix the intermediate certificate problem |
| 13:36 | <annevk> | (I pointed him to the incorrect intermediate certificate initially, since StartSSL instructions were not super clear :/) |
| 13:37 | <buyoancing> | i can provide some apache SSL directives that will give the site an A+ score |
| 13:38 | <annevk> | buyoancing: I don't think DreamHost allows that level of configuration |
| 13:39 | <annevk> | buyoancing: we need them to upgrade their VPS servers just as they did for their Shared servers |
| 13:39 | <annevk> | buyoancing: they've upgraded all Shared servers over the past month or so, so hopefully VPS will happen soon |
| 13:39 | <buyoancing> | I see |
| 13:39 | <buyoancing> | Anyway, if someone can point them in the right direction, http://pastebin.com/YYitWhgQ has some clues |
| 13:40 | <annevk> | That -SSLv2 is still required is somewhat disturbing |
| 13:41 | <buyoancing> | Is that Apache's fault or mod_ssl or OpenSSL? |
| 13:42 | <annevk> | Apache / mod_ssl I suspect, unless they just pass this straight through to OpenSSL but that doesn't seem like a good idea |
| 13:42 | <buyoancing> | perhaps it's required due to the "ALL" before "-SSLv2" ? |
| 13:44 | <buyoancing> | "all" is the default |
| 13:44 | <buyoancing> | https://httpd.apache.org/docs/current/mod/mod_ssl.html#sslprotocol |
| 13:44 | <annevk> | buyoancing: I have no idea; https://wiki.mozilla.org/Security/Server_Side_TLS#Apache recommends the same |
| 13:44 | <buyoancing> | it says: "all" is a shortcut for ``+SSLv3 +TLSv1'' or - when using OpenSSL 1.0.1 and later - ``+SSLv3 +TLSv1 +TLSv1.1 +TLSv1.2, respectively |
| 13:45 | <buyoancing> | anyway, thanks |
| 13:45 | <buyoancing> | TTYL |
| 13:47 | <annevk> | Oh he left |
| 13:47 | <annevk> | Well I filed https://www.w3.org/Bugs/Public/show_bug.cgi?id=27390 anyway |
| 14:38 | <annevk> | This prototype.includes decision fucks over DOMTokenList and DOMStringList (if we can't remove it) |
| 14:40 | <annevk> | Maybe TC39 needs a "1Platform" slogan |
| 14:49 | <jgraham> | I think the slogan of the web is something like TSBOWTDI;BTWIDFEAAOTSMOTFLRA (There Should Be One Way To Do It; But That Way Is Different For Every API And Often There's More Than One For Legacy Reasons Anyway) |
| 15:37 | <smaug____> | How unstable or stable might http://drafts.csswg.org/css-scoping/ be |
| 15:37 | <smaug____> | TabAtkins might have an opinion |
| 17:16 | <annevk> | Anyone here with input on the security considerations for Encoding? https://www.w3.org/Bugs/Public/show_bug.cgi?id=19961#c3 |
| 17:23 | <hsivonen> | annevk: isn't a VPS all about having your own virtual server? is it really Dreamhost's responsibility to update the TLS config on Hixie's VPS? |
| 17:24 | <annevk> | hsivonen: as far as I can tell DreamHost VPS just means you're guaranteed some set of system resources |
| 17:24 | <hsivonen> | annevk: ok. pretty misleading to call that "VPS" |
| 17:24 | <annevk> | hsivonen: you could copy yourself on the bug I filed, perhaps Hixie knows more |
| 17:25 | <hsivonen> | I guess I can't make Validator.nu emulate Firefox's initial TLS handshake until whatwg.org moves away from its RC4-only state |
| 17:26 | <hsivonen> | not that I'm in a particular hurry to do that |
| 17:26 | <hsivonen> | but I want to at some point |
| 17:26 | <annevk> | WHATWG is holding back the web, you heard it here first |
| 17:27 | <annevk> | It's a shame DreamHost took such a long time to start paying attention to this stuff, but then so did I |
| 17:42 | <jgraham> | From Dreamhost's page it at least looks like you would expect it to be possible to update the TLS config |
| 17:43 | <jgraham> | Since it seems like it's a VM on which you have root |
| 17:57 | <annevk> | Yeah sounds likely, not sure if Hixie is game for playing with that |
| 18:01 | <annevk> | hsivonen: I'm trying to restore webforms2.testsuite.org, but it's taking a bit longer than expected |
| 18:01 | <annevk> | hsivonen: no ETA, something is broken on DreamHost's end :/ |
| 18:56 | <Hixie> | i'm reluctant to do anything to change the system config because the changes tend to get blown away easily |
| 19:09 | <hsivonen> | Hixie: blow away as in Dreamhost overwriting stuff on VPSs? |
| 19:09 | <Hixie> | yeah, when you fiddle with the panel for example |
| 19:09 | <Hixie> | or when they're trying to fix some problem |
| 19:10 | <hsivonen> | oh so a VPS has some management tools connected to it even though you are supposed to have root? |
| 19:10 | <Hixie> | you have root, but yes, the management tools also have root. |
| 19:11 | <Hixie> | (i'd much rather have the management tools than have root) |
| 19:11 | <hsivonen> | I see |
| 19:12 | <hsivonen> | I wonder if Dreamhost has always been RC4-only or if this is fallout from the "mitigate BEAST with RC4" meme |
| 21:11 | <JonathanNeal> | I’m working to standardize tests for polyfills - something that could benefit multiple polyfill projects or services. Would anyone be interested in collaborating with me on this? |
| 21:18 | <JonathanNeal> | Or maybe I should say tests for browser features. |
| 21:20 | <Ms2ger> | Too late, we've already done that |
| 21:20 | <Ms2ger> | https://github.com/w3c/web-platform-tests |
| 21:22 | <JonathanNeal> | oh sweet! |
| 21:24 | <JonathanNeal> | Ms2ger: I’m glad I asked. I am hoping to standardize testing suites between the different polyfill projects. I am having trouble finding tests for <Element>.append or Promise. Do you have those, too? |
| 21:24 | <Ms2ger> | I don't know where Promise tests are hidden |
| 21:25 | <Ms2ger> | Has anyone implemented append()? |
| 21:26 | <Ms2ger> | Don't see any functional tests for it in wpt |
| 21:26 | <JonathanNeal> | “are hidden” - am i supposed to read into that? |
| 21:26 | <jgraham> | Ms2ger: append? |
| 21:26 | <Ms2ger> | jgraham, on nodes |
| 21:29 | <JonathanNeal> | Ms2ger: I don’t think either are there, at least after surfing through commit logs. |
| 21:29 | <Ms2ger> | I think some Promise tests may exist outside wpt |
| 21:29 | <Ms2ger> | In my view, that makes them hidden :) |
| 21:31 | <JonathanNeal> | some, like there are others inside wpt? |
| 21:31 | <jgraham> | Isn't promise in TC39 now? |
| 21:32 | <jgraham> | I don't think they are interested in contributing to wpt |
| 21:32 | <JonathanNeal> | Well, Domenic has tests. I could port them. |
| 21:32 | <JonathanNeal> | If they’re not already perfect ootb. |
| 21:33 | <astearns_> | jgraham: I thought test262 had looked into contributing to wpt |
| 21:34 | <JonathanNeal> | Wow, and with that (TC39, test262) you’ve almost lost me. |
| 21:35 | <jgraham> | astearns_: If so you know more than me |
| 21:35 | <JonathanNeal> | Well, I would like to help, or at least integrate with or leverage those tests. |
| 21:35 | <jgraham> | I thought that tobie enquired about it at some point but nothing happened |
| 21:36 | <JonathanNeal> | How can I help? |
| 21:37 | <jgraham> | JonathanNeal: Help with what? |
| 21:38 | <JonathanNeal> | Making some of those tests un-“hidden”. |
| 21:38 | <jgraham> | JonathanNeal: See if they're at https://github.com/tc39/test262 |
| 21:38 | <jgraham> | Then see if any vendors import those tests, I guess |
| 21:39 | <JonathanNeal> | oh my, that’s a rough testing suite to understand. |
| 21:39 | <jgraham> | Looks like Mozilla has http://dxr.mozilla.org/mozilla-central/source/js/src/tests/test262 but I don't know how up to date it is |
| 21:41 | <astearns_> | hmm - can't find a reference to what I was thinking of |
| 21:41 | <JonathanNeal> | Are those intended to be human readable? |
| 21:42 | <JonathanNeal> | human readable / searchable |
| 21:42 | <astearns_> | bterlson: do you remember whether test262 was looking into contributing to wpt? Failing that, do you know the state of promises tests? |
| 21:53 | bterlson | reads backlog |
| 21:53 | <JonathanNeal> | jgraham: it’s hard to tell since so many of the tests use assertions via Promise |
| 21:54 | <JonathanNeal> | But maybe in es6/ch25/25.4/25.4.4/25.4.4.1 |
| 21:54 | <bterlson> | astearns_: I don't htink there have been any discussions about contributing to wpt, but they are OSS and could be adapted by anyone I believe |
| 21:54 | <bterlson> | yes, those are the promise tests |
| 21:55 | <JonathanNeal> | And they should be hidden from wpt? |
| 21:55 | <bterlson> | note that once https://github.com/tc39/test262/issues/107 is finished (probably this weekend) the promise tests will be under lib/Promise |
| 21:55 | <bterlson> | also you may want to check out https://github.com/bterlson/test262-harness if you don't want to use the included python harness |
| 21:55 | <bterlson> | JonathanNeal: was that question directed at me? |
| 21:56 | <sicking> | annevk: ping |
| 21:56 | <JonathanNeal> | bterlson: anyone, since i don’t understand the organizational structure of these tests. |
| 21:57 | <bterlson> | JonathanNeal: I don't know what "hidden from wpt" means, exactly |
| 21:57 | <bterlson> | JonathanNeal: the current organization is based on a hybrid of ES5 and ES6 spec section numbers |
| 21:57 | <JonathanNeal> | bterlson: Ms2ger wrote “I don't know where Promise tests are hidden” and “I think some Promise tests may exist outside wpt” “In my view, that makes them hidden :)” |
| 21:57 | <bterlson> | but as per 107 we are moving to a more undertandable organization structure |
| 21:58 | <bterlson> | I think I disagree with that view :-P |
| 21:59 | <bterlson> | the entire goal of test262 is to be exactly what you want it sounds like: a test suite that implementers (including, perhaps especially, polyfill authors) can use |
| 21:59 | <Ms2ger> | Would you claim that JS isn't part of the web? |
| 22:00 | <JonathanNeal> | I help contribute to https://github.com/Financial-Times/polyfill-service and I’d like to see some standardization of tests. wpt seems very promising but possibly quite behind. Not sure if you were saying tc39 is becoming more readable or if you were talking about wpt or whatever 107 is. |
| 22:00 | <bterlson> | JS is part of the web yes. But its domain is not exclusive to the web. |
| 22:01 | <bterlson> | JonathanNeal: test262 is standardized tests, and is under active development, and is the official test suite for ECMAScript. Anything polyfilling ECMAScript functionality should be running official ECMA tests. |
| 22:01 | <JonathanNeal> | Ideally, I would rather contribute a test to some independent repository and reference it in whatever service I’m using. I would, however, like the tests to be somewhat readable to developers, and be possible to test across all browsers/versions to generate a compatibility list. |
| 22:02 | <bterlson> | issue 107 that I linked above is just the details about us changing the organization to be more striaght forward (eg. test/library/Promise for promise tests rather than test/suite/es6/ch25/...) |
| 22:02 | <bterlson> | Independent of a standards body you mean? |
| 22:02 | <JonathanNeal> | No, independent of the polyfill implementation. |
| 22:02 | <bterlson> | Oh, right, I agree |
| 22:03 | <bterlson> | sounds like test262 is what you want. There will be some problems to solve TBH, but we share the same goals and so would work to address any problems |
| 22:04 | <bterlson> | once issue 107 is addressed, identifying tests by feature and getting pass rates across browsers (or any other JS hosts incluidng node) will be fairly easy. Needs harness work, though I have already started it. |
| 22:05 | <Hixie> | wtf http://www.w3.org/2011/tracking-protection/drafts/tracking-dnt.html |
| 22:06 | <JonathanNeal> | For instance, there’s https://github.com/polyfills/polyfills and Yahoo is building their own service https://twitter.com/kaelig/status/534335699512999936 and there are numerous other polyfill libraries out there. |
| 22:06 | <bterlson> | and meanwhile, the actual polyfills are in various stages of spec compliance, ranging from pretty good to "what is this doing?" |
| 22:09 | <bterlson> | JonathanNeal: feel free to email me too if you have long form concerns you want to hash out |
| 22:09 | <JonathanNeal> | bterlson: I’ve been a sinner and saint in regards to spec compliance. Sure, let me PM you for deets. |
| 22:10 | <JonathanNeal> | But the bigger subject is that we have this big, massively distributed polyfills, and the tests they use also generate compatibility charts, like the kind you see at kangax or caniuse. |
| 22:10 | <JonathanNeal> | *have these |
| 22:10 | <bterlson> | Yes. I am hoping that test262 can ultimately feed data into Kangax's chart at least. He's open to the idea. |
| 22:10 | <JonathanNeal> | So, it would make sense for us to be sharing at least SOME tests, and to be addressing the varying levels of spec compliance. |
| 22:11 | <bterlson> | I'd argue that there is little reason not to share all tests. Eg. if ECMA spec says test should pass then it should go in test262. Better for everyone that way. Plus implementers run test262 a lot. |
| 22:11 | <JonathanNeal> | For instance, I think it’s reasonable to see a polyfill that matches functionality, but discards aspects of conformance, like the iterability of a method or property, or the argument of some method. |
| 22:12 | <JonathanNeal> | Yea, I’ve worked to make all of our tests CC0, so no signing an agreement just to play. |
| 22:12 | <bterlson> | I don't know how it can match functionality but also take different arguments |
| 22:12 | <bterlson> | thankfully the test262 CLA is super simple |
| 22:12 | <JonathanNeal> | *argument lengths |
| 22:12 | <bterlson> | ahh |
| 22:13 | <JonathanNeal> | function() {} vs function properName() {}, or function () { /* parse arguments */ } function (technicalExpectsAtLeastOneArgument) {} |
| 22:13 | <bterlson> | if s/iterability/enumerability/ then I agree with reservations (it probably depends on if iterating over the keys of the object is a common thing) |
| 22:13 | <bterlson> | but yeah, no point tanking perf just to have proper enumerability in a case no one will run in to |
| 22:14 | <JonathanNeal> | Right, so having a consistent chart, like “A conformance meets these demands (tests), AA conformance meets these demands” even if the browser can’t technically meet AA conformance it’s good to know that a polyfill meets the basic requirements. |
| 22:14 | <JonathanNeal> | We’ve run into that ourselves with iterability tests in older browsers. |
| 22:15 | <JonathanNeal> | We don’t want to say “sorry, this polyfills fails” when it doesn’t, it just doesn’t conform to iterability. |
| 22:15 | <bterlson> | my theory in this area is that developers want to make the call about what "good enough" is for themselves |
| 22:15 | <bterlson> | it's a hard thing to prescribe |
| 22:16 | <JonathanNeal> | Yes, but some accessibility standards have adopted a pretty good approach, like I’ve referenced with A, AA, AA compliance. |
| 22:16 | <JonathanNeal> | *AAA, pardon all the typos. |
| 22:16 | <bterlson> | but it takes a judgement call to decide what AAA means, so it seems possible that depending on the tradeoffs made, an A polyfill might be better than AAA. Maybe better just to give information on what passes and fails. |
| 22:17 | <bterlson> | can you link me to these tests though? Curious |
| 22:17 | <JonathanNeal> | And this seems like a problem for a standards body to address, and I definitely have an immediate need. With Yahoo entering the scene, it seems like we should take it a bit more seriously. |
| 22:17 | <bterlson> | agreed |
| 22:17 | <bterlson> | well, it's why test262 exists in the first place |
| 22:18 | <JonathanNeal> | https://github.com/polyfills/polyfills/blob/master/test/test.js (bundled into one file) https://github.com/Financial-Times/polyfill-service/tree/master/polyfills (packaged with each polyfill) |
| 22:18 | <bterlson> | although until recently we've not had a harness that js devs want to use, or done a good job of making the tests accessable |
| 22:19 | <JonathanNeal> | And don’t forget https://github.com/es-shims/es5-shim/tree/master/tests |
| 22:19 | <JonathanNeal> | Those, I think, are the most widely cited (not including Domenic’s https://github.com/promises-aplus/promises-tests which are exclusively for promises) |
| 22:20 | <bterlson> | yeah I briefly talked to Jordan about test262 couple days back at tc39. I think we can share tests with es5/6/7-shim. |
| 22:20 | <JonathanNeal> | In the process of creating our own tests, we also noticed that sometimes the caniuse or mdn reports were off (possibly even a few older kangax es5 ones). I think this is because those tests are not automatically generated. |
| 22:21 | <bterlson> | the A+ promise tests we decided not to port on the grounds that it also tests promise interaction with non-ECMAScript things like setTimeout |
| 22:22 | <Domenic> | and also all vendors are running them anyway |
| 22:22 | <JonathanNeal> | Well, for those of you working on these official suites, I would love to involve you in *our discussion or get *us talking to you about using your tests. *our/us as in the various polyfill libs. |
| 22:23 | <bterlson> | yes and that |
| 22:23 | <bterlson> | yep, love to be included in that discussion :) |
| 22:23 | <JonathanNeal> | Domenic: Thoughts? Do you think I’m accurating describing a real problem or just off in my own la la land? Please be honest. |
| 22:24 | <Domenic> | JonathanNeal: it would be good if polyfills used and contributed to the standard test repos (viz. web-platform-tests and test262). If you can make it happen, go for it. |
| 22:25 | <bterlson> | (viz. used properly is a beautiful thing to behold) |
| 22:26 | <JonathanNeal> | If the tests can (as I imagine they should) generate caniuse-like support maps for all browsers, then I can definitely help make that happen. |
| 22:28 | <JonathanNeal> | Two things on the testing suite’s end could really help. If the tests could allow lower bands of conformance (re: functional, without conforming to iterability, argument length, or argument name). |
| 22:28 | <JonathanNeal> | *method name |
| 22:29 | <JonathanNeal> | A good case for method name is Promise prototype’s catch. You can’t name a method catch. |
| 22:29 | <JonathanNeal> | Well, you could, but then your JS will surely fail in almost any browser. |
| 22:30 | <bterlson> | JonathanNeal: we'll get there in a few weeks. Step 1 is identify tests by feature name, which is almsot complete. Step 2 is teach the harness how to pull categories from a config file and report pass rates by category. |
| 22:30 | <JonathanNeal> | The second thing that could really help would be organization of the tests. Some kind of common configuration. Right now, things seem pretty scattered. |
| 22:31 | <bterlson> | JonathanNeal: IE8 is the only browser I know of that doesn't like reserved words for property names? |
| 22:31 | <JonathanNeal> | Try function catch() {} in whichever browser you please. |
| 22:31 | <bterlson> | that's not a property name |
| 22:32 | <bterlson> | but say var obj = {catch: 1} is fine |
| 22:32 | <JonathanNeal> | Sorry, I thought I said method name. |
| 22:32 | <JonathanNeal> | Method name, like right now you can test Promise.prototype.catch.name === “catch" |
| 22:32 | <bterlson> | methods are usually attached to objects right? so foo.catch is fine (IE>8) |
| 22:32 | <JonathanNeal> | obj.foo can be function bar() {} |
| 22:33 | <JonathanNeal> | So, if you test for someFunction.name and someFunction.length you could have a functional polyfill that fails. |
| 22:34 | <JonathanNeal> | Or, likewise, a browser that fails the test on those grounds. |
| 22:34 | <Domenic> | You can name a method catch in all browsers. You just can't name a function catch. |
| 22:35 | <Domenic> | Anyway, tests are meant for browsers, not polyfills, so polyfills will always fail some tests |
| 22:35 | <Domenic> | If polyfill authors want to collate a list of "known to be impossible to pass" (or hard to pass) tests, they should do that |
| 22:35 | <Domenic> | e.g. like how V8 maintains a list of "known to be slow to run" test262 tests. |
| 22:35 | <Domenic> | that's not really part of test262/wpt's job |
| 22:35 | <bterlson> | JonathanNeal: Re: your second item, you will find that after this weekend (I hope) the organization will make much more sense. |
| 22:36 | <JonathanNeal> | Which testing suite is this for again, bterlson? Is this the one that asked me to sign something? |
| 22:37 | <JonathanNeal> | That later question is less important. |
| 22:37 | <JonathanNeal> | At least, to me. |
| 22:38 | <bterlson> | JonathanNeal: test262 is what I'm talking about |
| 22:38 | <JonathanNeal> | Domenic: agreed that “known to be X” might need to exist somewhere else, as long as it’s all iteroperable. |
| 22:38 | <JonathanNeal> | If, at the end of the day, I just copied and pasted the tests I liked from wpt or 262, it doesn’t do us much good. |
| 22:39 | <JonathanNeal> | In other words, I want browser feature support charts and polyfill support charts that can get gold, silver, or bronze medals that actually mean something and aren’t awarded by ourselves. |
| 22:40 | <bterlson> | Sounds like we want the same things :) |
| 22:41 | <bterlson> | JonathanNeal: I think the form your "known-tests" would take is a config file for the harness that users can pick up to tell the harness to run only the specific set of tests for whatever "conformance level" you've defined |
| 22:42 | <JonathanNeal> | Yes, except that I would need the conformance level to be beyond my implementation, too. |
| 22:42 | <JonathanNeal> | conformance levels |
| 23:11 | <Hixie> | annevk: it strikes me that realms and global objects can't be the same thing because you can change the global object during runtime of a script (using document.open) but that presumably doesn't change the realm |
| 23:11 | <Hixie> | not sure if it changes the global environment record |
| 23:40 | <Hixie> | we have 31 domains with a cert |
| 23:40 | <Hixie> | each of which takes about 12 seconds to update |
| 23:40 | <Hixie> | i am going insane |
| 23:41 | <Hixie> | and dreamhost's !@#$%^ portal won't let you just open each one in a new tab and update them all, because they store state in !@#$%^ cookies |
| 23:41 | <Hixie> | good work team |