freenode :: #whatwg

13 Feb 2017 seems to be down
00:34xfq"A database query error has occurred. This may indicate a bug in the software."
04:35foolipannevk, jgraham, gsnedders, I tend to use "web-platform-test", although as the first part of a title it looks a bit odd. I did it anyway though.
04:49MikeSmithfoolip: I noticed thats what jgraham normalized it into a recent change the wpt
04:50MikeSmithoh but you do mean web-platform-tests, right?
06:04MikeSmithin WebIDL blocks, do we make any attempt to put the members in any kind of order?
06:05MikeSmithor is it just, whatever
06:05MikeSmithe.g., looking at
06:05MikeSmithis there any discernable order there?
06:06MikeSmiththe members are obviously not alphabetically sorted, but are they sorted in some other logical way?
06:07MikeSmithdo we not care?
06:08MikeSmithin the case of the HTMLLinkElement interface, is there any good reason *not* to just re-sort them alphabetically?
06:09MikeSmithit otherwise seems suboptimal to make N different readers have to hunt through that looking for whatever specific member they might have come looking for
06:15DomenicMikeSmith: the order impacts their web-exposed iteration order
06:15DomenicSo it's largely the historical order they were added in
06:15DomenicWe know of at least one case where web content depends on this, for some canvas stuff
06:16MikeSmiththat seems .. fragile
06:16DomenicBecause of minified code contests that do something like var x = Object.keys(CanvasContext2D.prototype) and then they can use x[5] instead of typing out a "long" method name
06:16Domenicyeah it's not great...
06:16MikeSmithah geez
06:17DomenicMikeSmith: excited to look at your PR in more depth tomorrow, the one thing I'm worried about is whether the reorganization is consistent with other elements. E.g. in general we keep everything about an attribute together instead of separating authoring and implementer stuff. We'll see though.
06:18MikeSmithyeah I think that link section may be a very special case
06:18annevkbz had some ideas on randomizing order of IDL blocks except where it mattered
06:18annevkDoubt that will ever go anywhere
06:19annevkThis is also a problem with partial interfaces and the unclear order in which they are supplied
06:19MikeSmithI personally find the current organization of that link section to be sorta random and very hard to read through. It jumps all over the place. I think thats partly because in adding things over the last year or so, weve kinda just dumped stuff in there wherever
06:20MikeSmithannevk: yeah when I take members from a partial interface in another spec, and I add them to the element interface in the HTML spec, how to do I know where to add them?
06:21MikeSmithjust tack them on the end?
06:21annevkSure, but what about multiple partials
06:21annevkNobody has been looking towards solving this
06:22annevkAlong with element attributes and maybe some operations in JavaScript it's one of the few holes we have with regards to ordering
06:23MikeSmithkinda surprised it hasnt caused more problems
06:24MikeSmithanyway I guess for the IDL for something that has been implemented but is not in the HTML spec yet, we may need to look at where it was added in the IDL code for the engine(s) that implemented it?
06:24MikeSmithI mean especially if theyve shipped it
06:25annevkMikeSmith: depends, but that never hurts
06:26MikeSmithDomenic: annevk as far as in general we keep everything about an attribute together instead of separating authoring and implementer stuff I am starting to wonder if we really want to do it that way because its best or instead just because thats the way its always been
06:27DomenicYeah, maybe, but I'd rather change all that at once than piecemeal, if we were to do that..
06:27MikeSmithI think think especially that some of this <span no-dev> stuff in the middle of paragraphs is kind of a red flag
06:27MikeSmithwell I am happy to start going through and looking at it in other places too
06:28MikeSmithbut anyway I think we do have evidence that it confuses implementors sometimes
06:28MikeSmithIMHO where there are UA requirements in the spec we should be making it as easy as possible for implementors to see them
06:29annevkMikeSmith: changing it into &quot;Writing&quot; and &quot;Processing&quot; sections would work for me
06:29MikeSmithyeah that would be a lot of work if we did it globally
06:30annevkMikeSmith: but we should indeed do it in some principled way with a plan, since it&#39;s a bit too much do it piecemeal
06:30MikeSmithyeah I guess we can start by looking at this link section as a test case
06:30annevkMikeSmith: or we could see about identifying the distinction more clearly somehow
06:31MikeSmithbtw in this case I didnt patch the link section just because I didnt like the current aesthetics but instead because I was looking at how to integrate the rel=serviceworker stuff into it
06:32MikeSmithscope, workertype, usecache
06:33MikeSmithand once I started trying to add that serviceworker stuff, I found myself .. getting a bit lost/distracted by the current structure
06:35Domenicwow they added three attributes??
06:35MikeSmithannevk: I think one relatively easy way for identifying the distinction more clearly is put specific (sub)headings on the UA-requirements stuff
06:35MikeSmithDomenic: yeah, seems so
06:36DomenicI see they went with workertype instead of type, that&#39;s annoying
06:37DomenicWait what
06:37Domenic&quot;If useCache is not a valid boolean value, queue a task to fire an event named error at the link element, and abort these steps.&quot;
06:37MikeSmithannevk: which is what I did in that link patch (short of adding separate &quot;Writing&quot; and &quot;Processing&quot; sections; because the distinction can be made implicitly somewhat just be adding specific subheadings like that)
06:37MikeSmithDomenic: yeah
06:37DomenicSo you have to do usecache=&quot;true&quot; or usecache=&quot;false&quot;???!?
06:37MikeSmithsaw that too
06:37DomenicFiling bug now...
06:37MikeSmithyeah I hope that was not the intent
06:38MikeSmiththanks for saving me the time of filing a bug for that myself :)
06:38MikeSmiththat sentence should just be dropped
06:38MikeSmithand the one above it should be changed
06:39MikeSmithjust to say, &quot;If the usecache attribute is present, set the use cache to true.&quot; Period.
06:39MikeSmithI think
06:39MikeSmithbecause it is false by default iirc
06:40MikeSmithbut regardless the wording should just be about the presence of the attribute
06:40Domenicok, going to bed now...
06:40* MikeSmith looks
06:41MikeSmithbtw, the SW spec also treats &quot;classic&quot; and &quot;module&quot; case-sensitively
06:42annevkcase-sensitive seems fine
06:42DomenicYes, they seem to be following the precedent annevk insists on setting in but nobody implements...
06:42DomenicReally not a fan
06:42annevkIf we make IDL ASCII case-insensitive count me on board
06:42MikeSmithI yeah didnt mean to re-open that can o worms
06:43annevkAt some point you want to explain element attributes through IDL enums too
06:43annevkAnd their value space mismatch is just bad
06:43* Domenic closes the tab and backs away...
07:34foolipMikeSmith: yes, I meant web-platform-tests with the s
09:42annevkHow can you enforce a max-width but allow it to go wider for really wide content?
09:43annevkSomething like max-width:max(60em, min-content)
09:54zcorpan_annevk: would be nice to loop in someone from chromium and edge before merging NonTransitional, no?
09:55annevkzcorpan_: yeah I suppose that would be better
09:55annevkzcorpan_: I don&#39;t really know who to copy from Edge
09:56annevkzcorpan_: I guess for Chrome jungshik + sleevi?
09:58zcorpan_annevk: yeah i guess
10:06annevkzcorpan_: done
10:06annevkzcorpan_: I doubt we&#39;ll get a reply from Edge, but who knows
10:08annevkI&#39;ll file bugs on them though
10:08annevkProbably all of them, since everyone has bugs
11:40zcorpan_hsivonen or MikeSmith do you want to reply to the DTD email on the mailing list?
11:56* Ms2ger would tend to ignore
12:00annevkzcorpan_: we did discuss it at some point, I was wondering whether it was actually an issue
12:00annevkzcorpan_: I don&#39;t recall if we reached a conclusion though
12:01zcorpan_annevk: i think we concluded that it&#39;s an issue in DTD but not in RELAX NG
12:09annevkMs2ger: can you synchronously find out whether a blob URL is revoked?
12:10annevkMs2ger: through XHR maybe?
12:10annevkMs2ger: or also less &quot;deprecated&quot; APIs?
12:10Ms2gerI don&#39;t know off-hand
12:11annevkI kinda want to know, but I&#39;m not sure whether I want to write tests since then I get to own it I guess
12:22MikeSmithzcorpan_: I can reply once Ive actually read it, which I havent yet
13:21annevkjgraham: ping for
13:24jgrahamannevk: Noted. Not sure if it will happen today (patches accepted though)
13:25annevkjgraham: I would write a patch, but there is no contributor documentation making it clear how to approach it
13:26annevkjgraham: I guess I also don&#39;t really know what request_handler actually returns or how to find out, so...
13:43annevkzcorpan_: perhaps you should ping hsivonen explicitly in or did you already use email or some such?
13:44annevkzcorpan_: or does assign to review create a sufficient ping in GitHub already? I don&#39;t actually know
13:45zcorpan_i had pinged him here in irc. but i can fire off an email
13:48* zcorpan_ learns has examples where conversion succeeds but validation fails
13:50annevkzcorpan_: ToUnicode always succeeds basically
13:51zcorpan_meant to link to
13:53zcorpan_but yeah. i found it noteworthy since i didn&#39;t see anyone complain about ToUnicode in issue 118
13:54annevkToUnicode is actually not currently exposed anywhere
13:54annevkOh actually, no, I guess HTML uses it
13:55annevkBut maybe not the parser override, just the conversion algorithm
13:58zcorpan_annevk: does anything set the Unicode flag for the host parser?
13:59annevkzcorpan_: domainToUnicode() used to be a thing, but not anymore
13:59annevkzcorpan_: but also, that could have been defined as host parsing followed by &quot;domain to Unicode&quot;
13:59annevkzcorpan_: so I think I&#39;m going to remove it and encourage that pattern instead
13:59zcorpan_annevk: sounds reasonable
15:27annevkI kinda want to replace e-mail with email
15:31Ms2gerNot in Dutch, I hope :)
15:33annevkMs2ger: my written Dutch is not great and I have no ambitions there I think
16:36MikeSmithI read the message about the DTD on the mailing list but I dont really know what Id say in reply except, If people who are using tools that consume DTDs find it useful to have an HTML DTD, then great
16:36MikeSmiththere are lots of possible formalisms
16:37MikeSmithI think the devs working on the MDN backend have some kind of JSON representation they are using to define HTML element content models and attribute
16:38MikeSmithas far as DTDs I guess psgml still exists for Emacs and uses DTDs so if theres a DTD people can use with that, great
16:39MikeSmithpsgml is a pretty nice way to get element and attribute completion when editing HTML docs in emacs
16:58gsneddersMikeSmith: I had the impression that they claimed to have represented everything in the DTD, which I thought hsivonen had said was impossible?
16:58MikeSmithyeah it is impossible of course
16:59MikeSmithcertainly cant represent data-* attributes
16:59MikeSmithnor custom elements
17:00MikeSmithnor handle ARIA role checking when the role value contains multiple tokens
17:00MikeSmithnor do much of any microsyntax/datatype checking of attribute values at all
17:00MikeSmithetc etc
17:03MikeSmiththat said if the goal is not validation but instead just having a DTD that can be plugged into editing tools like psgml to enable context-sensitive element and attribute completion/editing, then it may enought of that right to be very useful
17:03MikeSmith*may get enough of that right
17:05gsnedders touches on using it to parse
18:31annevkI liked the argument that DTDs were more concise
18:32annevkOnly true of course if you define that language first and need it elsewhere, which is a little less clear
22:10AutomatedTesterjyasskin: hey, thanks for the link to promises guide
22:10jyasskinAutomatedTester: np :)
22:10AutomatedTesterjyasskin: are we supposed to reference that guide for certain things
22:10AutomatedTestere.g. promise-calling
22:12jyasskinAutomatedTester: If you use one of the terms defined there, you should cross-reference to it (using, e.g. [=resolve=] in Bikeshed). I haven&#39;t found [=promise-calling=] to be all that useful, but I use [=a new promise=] all the time. (cc Domenic)
22:13AutomatedTesterjyasskin: oh, annevk told me new promise was a webidl issue
22:14AutomatedTesterfetch and url dont have new promise linked
22:14jyasskinMost of the content in promises-guide should move to WebIDL, but it hasn&#39;t yet.
22:15jyasskinI think fetch and url are formally incorrect by not linking to promises-guide, but nobody&#39;s going to implement the wrong thing as a result, so it doesn&#39;t really matter.
22:16AutomatedTesterthats pretty much what annevk said :D
22:16jyasskinHah :)
22:16jyasskinThe nice thing about using [==] is that when the terms move, you&#39;ll just be able to rebuild your spec, and it&#39;ll automatically update the link targets.
22:17AutomatedTesterwhats it in respec?
22:18* AutomatedTester can&#39;t find a link in so guessing I should add it...
22:18jyasskinRespec didn&#39;t do cross-links when I was using it. It might have been added since then, but I don&#39;t know the syntax if so.
22:18TabAtkinsAutomatedTester: Dont&#39; worry about it, because you shouldn&#39;t be using ReSpec. ^_^
22:18jyasskinCross-links are why I switched to Bikeshed.
22:18TabAtkins(But also, yes, ReSpec does not have cross-spec linking.)
22:18jyasskinNow Tab, you&#39;re biased. ;)
22:18AutomatedTesterTabAtkins: I use what I have...
22:19AutomatedTesternot going to bikeshed...oh
22:21tobieAutomatedTester, jyasskin:
22:22AutomatedTestertobie: I didnt realise that site existed... I always search the json file
22:22AutomatedTesterthanks tobie
22:22tobieThere&#39;s also search from within the respec top right menu.
14 Feb 2017
No messages
Last message: 8 days and 1 hour ago