| 00:01 | <annevk> | GPHemsley: I still don't think you should accommodate EOT |
| 00:02 | <annevk> | SimonSapin: ah, good to know about subscribing, guess we should mention that somewhere |
| 00:04 | <Hixie> | http://software.hixie.ch/utilities/js/canvas/ (and hit download) -- pretty inner shadow |
| 00:05 | <SimonSapin> | annevk: might be a temporary issue, I don’t know |
| 00:05 | <annevk> | GPHemsley: fwiw, I think if you use <li><p> more often you might get better visual results |
| 00:10 | <annevk> | thanks abarth and Hixie |
| 00:10 | <annevk> | I guess I will reply to the public-webapps thing again, but later, prolly a "work day" |
| 01:32 | <GPHemsley> | annevk: What's wrong with the visual results I'm currently getting? |
| 09:03 | <annevk> | MikeSmith: the problem with the URL tests looks like the opposite problem to me |
| 09:03 | <annevk> | MikeSmith: "assert_equals: expected "http://192.168.0.257"; but got "http://192.168.0.257/""; |
| 09:03 | <annevk> | MikeSmith: so they expect without slash, but get one |
| 09:04 | <MikeSmith> | yeah I realize that |
| 09:04 | <MikeSmith> | so I changed them so that the original does have a slash |
| 09:05 | <MikeSmith> | and then the result still has the slash too |
| 09:05 | <annevk> | okay |
| 09:05 | <annevk> | these tests do indeed seem quite WebKit specific |
| 09:05 | <annevk> | "http://go\@ogle.com"; expects the slash to disappear apparently... |
| 09:06 | <annevk> | euh, backslash |
| 09:07 | <MikeSmith> | btw do you have any clue why most of those tests don't get run at all in Opera? |
| 09:11 | <MikeSmith> | the changes I made are at https://github.com/sideshowbarker/url-testing/commit/fde8c0b7690a4f04d93d83c959864487b2ef6ade |
| 09:12 | <annevk> | not sure |
| 09:15 | <annevk> | omg CPU heating up like mad |
| 09:35 | <MikeSmith> | annevk: anyway maybe the spec should actually say that for pathless URLs a slash gets appended |
| 09:35 | <MikeSmith> | since that seems to be what browsers actually do |
| 09:36 | <annevk> | MikeSmith: the spec does |
| 09:36 | <MikeSmith> | oh |
| 09:36 | <MikeSmith> | so the test cases were wrong anyway, then |
| 09:37 | <annevk> | see http://url.spec.whatwg.org#concept-url-serializer |
| 09:37 | MikeSmith | reads |
| 09:37 | <annevk> | 'Append "/" concatenated with the strings in url's path (including empty strings), separated from each other by "/" to output.' |
| 09:38 | <MikeSmith> | ah OK |
| 09:48 | <annevk> | GPHemsley: mostly better margins, especially with that table it looks rather ugly currently |
| 09:50 | <annevk> | I guess eventually progress events should move to Fetch |
| 09:50 | <annevk> | I should start make a set of notes on Fetch |
| 10:19 | <alFReD-NSH> | Hi, I have an idea for protection against XSS in HTML. It would be nice if we could set an attribute on an Element, and that would stop any children from executing any `script` element. Is there any proposal similar to this? |
| 10:28 | <MikeSmith> | alFReD-NSH: you know about Content Security Policy? |
| 10:29 | <MikeSmith> | instead of using markup in the document, you add an HTTP header |
| 10:30 | <MikeSmith> | see http://dvcs.w3.org/hg/content-security-policy/raw-file/tip/csp-1.0-specification.html#sample-policy-definitions |
| 10:31 | <MikeSmith> | simplest case is you just send the "Content-Security-Policy: default-src 'self'" header |
| 11:38 | <MikeSmith> | it turns out that in the outline algorithm the steps that say to create an implied heading are actually pointless |
| 11:39 | <MikeSmith> | because the only part of the algorithm that checks to see if any implied heading has been created is a part that says, "if the heading of the last section of the outline of the current outlinee is an implied heading" |
| 11:40 | <MikeSmith> | and that could just as well be handled by not creating implied headings to begin with |
| 11:40 | <MikeSmith> | in which case a section that has no heading in the source will just have no heading in the outline |
| 11:41 | <MikeSmith> | and so that step could just be changed to read, "if the heading of the last section of the outline of the current outlinee has no heading" |
| 11:42 | <MikeSmith> | I know because that's the way I have it implemented |
| 11:42 | <MikeSmith> | and it works |
| 11:42 | <MikeSmith> | the whole "create an implied heading" adds nothing except additional confusion |
| 11:56 | <annevk> | is "implied heading" not something you want to show in the outline? |
| 12:01 | <MikeSmith> | sure it is |
| 12:02 | <MikeSmith> | but for the step of showing the outline, if a section has no heading, then you just show "implied heading" |
| 12:02 | <MikeSmith> | or whatever text -- e.g., "[untitled article element]" |
| 12:03 | <MikeSmith> | the action of building the outline does not need to do that in order for the action of showing it to be able to display something there |
| 12:04 | <MikeSmith> | and there's never any case where the algorithm doesn't say to add an implied heading for a heading-less section |
| 12:05 | <MikeSmith> | so having no heading and having an implied heading are exactly the same in practice |
| 12:08 | <annevk> | ah k |
| 12:17 | <annevk> | out of curiosity given relative reference "" (the empty string) and base URI "http://x";, how does http://tools.ietf.org/html/rfc3986#section-5.2.2 lead to a target URI "http://x/";? |
| 12:18 | <annevk> | R.path is "" and therefore T.path becomes Base.path (which is empty) and that's that |
| 12:20 | <annevk> | oh well, it's obsolete now |
| 12:41 | <MikeSmith> | annevk: I notice that some of the expected results in Chris's URL tests are different than the expected results in the WebKit URL tests |
| 12:42 | <annevk> | MikeSmith: well the spec does not match WebKit either so I guess that makes sense |
| 12:43 | <MikeSmith> | yeah but I think some of them neither match the spec nor match WebKit |
| 12:43 | <annevk> | oh okay, I guess you could patch those? |
| 12:44 | <annevk> | if you want me to take a look at specific cases you could leave them here, I gotta go for a bit but I can check later |
| 12:44 | <annevk> | I was hoping to get to testing in a few weeks, but maybe I should move that around if there's interest now |
| 12:44 | <annevk> | ttyl |
| 12:44 | <MikeSmith> | hai |
| 12:44 | <MikeSmith> | go/@ogle.com is one case I noticed |
| 15:21 | <Ms2ger> | jgraham, ping |
| 15:26 | <annevk> | MikeSmith: http://go/@ogle.com -> http://go/@ogle.com/ |
| 15:27 | <MikeSmith> | ok |
| 15:27 | <MikeSmith> | annevk: Chris's expected result for that is http://go⊙oc |
| 15:27 | <MikeSmith> | for some reason that I don't understand |
| 15:27 | Ms2ger | assumes "a mistake" |
| 15:28 | <MikeSmith> | prolly so |
| 15:28 | <annevk> | maybe the regular expression in STD 66 gives something else? |
| 15:30 | <annevk> | hmm, looks like just a mistake |
| 15:47 | <MikeSmith> | I'll file a bug for it later |
| 15:47 | <MikeSmith> | right now, shochu time |
| 15:48 | <MikeSmith> | btw, I believe I now have a conforming implementation of the outline algorithm in the server-side code for the validator |
| 15:48 | <MikeSmith> | available for testing at http://qa-dev.w3.org:8888/ |
| 15:49 | <MikeSmith> | I had earlier implemented something that generated it on the client side |
| 15:49 | <MikeSmith> | and the results look the same as what I had there before |
| 15:50 | <MikeSmith> | bot now the client-side code is gone and what you see there is generated on the server |
| 15:50 | <MikeSmith> | next thing I plan to add when I have time is a way for users to filter out certain error messages |
| 15:51 | <MikeSmith> | a la http://validator.keegan.st/ |
| 15:55 | <annevk> | MikeSmith: I just noticed we have a URL component under the WebAppsWG as well |
| 15:56 | <annevk> | MikeSmith: should I move all those bugs and will you then nuke that component or do you have a different preference? |
| 15:56 | <zewt> | firefox's "different by design" slogan is pretty embarrassing, heh |
| 15:59 | <zewt> | really the entire download page is painful |
| 16:01 | <annevk> | Is there a bug on updating HTML to use the URL spec? |
| 16:05 | <[tm]> | annevk: can't recall if there is a bug for that yet |
| 16:06 | <[tm]> | maybe an old one, for the old draft |
| 16:06 | <[tm]> | about the WebApps bugs, yeah lets move them |
| 16:07 | <[tm]> | either you can our I still later when I get back to my work cave |
| 16:09 | <annevk> | [tm]: https://www.w3.org/Bugs/Public/buglist.cgi?product=WebAppsWG&component=URL&resolution=--- is now empty |
| 16:09 | <annevk> | [tm]: https://www.w3.org/Bugs/Public/buglist.cgi?product=WHATWG&component=JavaScript can be removed too, matjas uses GitHub instead |
| 16:12 | <[tm]> | ok |
| 16:23 | <jgraham> | Ms2ger: Pong |
| 16:24 | <jgraham> | Ms2ger: I saw your patch and it's kind of OK, but I don't really understand why it's needed. Seems like timeout() should only do something if explicit_timeout is set, which means that timeout_length should be null. In theory. |
| 16:32 | <Hixie> | anyone up to date on srcset implementations? |
| 16:36 | <annevk> | hmm, no open bugs on HTML's public suffix list comparisons being wrong |
| 16:36 | <annevk> | apparently public suffix list uses Unicode rather than the ASCII representation of domains |
| 16:45 | <annevk> | Hixie: do you want a bug for the URL stuff |
| 16:50 | <GPHemsley> | Apparently Mozilla just got the memo this week that WHATWG standards are on GitHub and the Specs todo wiki page exists |
| 16:51 | <Hixie> | annevk: sure |
| 16:51 | <Hixie> | GPHemsley: oh? |
| 16:51 | <GPHemsley> | Hixie: https://wiki.mozilla.org/Firefox/Planning/2012-11-21#Standards |
| 16:52 | <annevk> | looks like they read my blog then :) |
| 16:52 | <GPHemsley> | ah |
| 17:10 | <annevk> | https://www.w3.org/Bugs/Public/show_bug.cgi?id=20072 |
| 17:36 | <Ms2ger> | jgraham, well, I kinda need timeout() to end up calling my completion callback |
| 17:39 | <annevk> | odinho: btw, re the userinfo thing |
| 17:39 | <annevk> | odinho: we want to preserve : |
| 17:39 | <annevk> | odinho: so http://user:@x/ should serialize as such |
| 17:40 | <annevk> | odinho: so http://:@x/ too, only in http://@x/ gets @ removed because there's no corresponding component |
| 17:42 | <Velmont> | So //:@x/ would indeed trigger the "has userinfo" part? And thus fail in redirect? |
| 17:50 | <annevk> | yes |
| 17:51 | <annevk> | curious edge case: //\t@/ \t is ignored by the parser |
| 17:51 | <annevk> | parser also ignores newlines, but you cannot use those in HTTP :) |
| 17:51 | <annevk> | I'm guessing \t is okay |
| 17:54 | <tantek> | Don't know about \t but @t is okay. |
| 17:55 | <annevk> | heh, I was using \t as a shorthand for 0x09 |
| 17:55 | <annevk> | btw http://wiki.whatwg.org/wiki/URL#Schemes |
| 17:58 | <annevk> | http://lists.w3.org/Archives/Public/public-script-coord/2012OctDec/0173.html |
| 17:58 | <annevk> | "We've interested in any and all feedback. In particular, if there are any Web API object features that can not be expressed in terms of the new MOP." I'm guessing document.all! |
| 17:59 | <Hixie> | or Window, probably |
| 18:00 | <annevk> | I feel this is somewhat backwards though, shouldn't they have gone through some effort to identify areas? |
| 18:03 | <Hixie> | annevk: what makes you say they didn't? |
| 18:04 | <annevk> | I would have expected them to have stated they looked if they had and conclude with "if there's anything we've overlooked, please let us know!" |
| 18:05 | <annevk> | but yeah, maybe they have |
| 18:09 | <tantek> | "Meta-Object Protocol" ?!? oh dear. |
| 18:17 | <Hixie> | annevk: eh, i often don't mention that since it's obvious :-) |
| 18:18 | <tantek> | annevk - re: http://wiki.whatwg.org/wiki/URL#Schemes, since you've explicitly included the "mailto" scheme, are various IM schemes in scope, e.g. "aim", "xmpp", "ymsgr", "msnim", "skype" (examples documented here: http://microformats.org/wiki/hcard-examples#AOL_Instant_Messenger_.28AIM.29 ) ? and telephony, e.g. tel:, fax:, callto: ? and the latest fun one is geo: RFC5870. |
| 18:19 | <annevk> | Hixie: guess I'm still somewhat skeptical about TC39, although it seems it's getting better |
| 18:21 | <annevk> | tantek: so yeah, I guess so |
| 18:23 | <annevk> | tantek: kinda depends on how we end up specifying all this. E.g. if you XHR() a skype URL, results in no resource, and then results in a network error |
| 18:47 | <jgraham> | Ms2ger: But do you set explicit_timeout in your testharnessreport.js? |
| 18:47 | <Ms2ger> | Doesn't look like it, is that new? |
| 18:48 | <annevk> | tantek: updated http://wiki.whatwg.org/wiki/URL#Schemes |
| 18:54 | <jgraham> | Ms2ger: I thought it was added because you asked for it |
| 18:54 | <Ms2ger> | Oh, heh |
| 18:55 | <jgraham> | I thought the idea was that you would disable the timeout in testharness.js and mochikit would call timeout() or something |
| 18:55 | <jgraham> | *mochitest |
| 18:56 | <Ms2ger> | We've got "timeout": 1000000 |
| 18:59 | <jgraham> | Oh |
| 18:59 | <jgraham> | Well yeah, that won't work |
| 19:02 | <jgraham> | Ms2ger: I have no idea how to set r+ on bugzilla, but if I did I would |
| 19:02 | <Ms2ger> | Change the select on https://bugzilla.mozilla.org/attachment.cgi?id=684870&action=edit :) |
| 19:03 | <jgraham> | Maybe I don't have enough permissions? |
| 19:03 | <jgraham> | (I don't see a select) |
| 19:03 | <Ms2ger> | Mm, possibly |
| 19:04 | jgraham | comments |
| 19:47 | <tantek> | annevk, ok, cool. and "geo" ? http://tools.ietf.org/html/rfc5870 |
| 19:48 | <annevk> | tantek: I guess that would end up like mailto |
| 19:48 | <tantek> | sort of. "geo" is descriptive, like a noun. whereas in practice, "mailto:" is more like an an action, like "open a new message to" |
| 19:49 | <tantek> | all the IM/telephony/email URLs are implemented as actions |
| 19:49 | <annevk> | tantek: in the end though, the first list is what matters, if it's not part of the first list, it's going to be handled externally |
| 19:49 | <tantek> | ok cool. didn't know if the distinction mattered for this spec. |
| 19:51 | <annevk> | I think what mostly matters is whether the URL maps to a resource (a bag of bits as Hixie calls it) |
| 19:53 | <tantek> | geo: is like data: in that respect then |
| 19:53 | <tantek> | it maps to a bag of coordinates bits |
| 19:54 | <annevk> | no it isn't |
| 19:54 | <annevk> | geo does not represent something that has a MIME type |
| 19:55 | <tantek> | hmm - though the RFC does liken it to mailto: "Accessing information about a particular location or triggering further services shouldn't be any harder than clicking on a 'mailto:' link and writing an email straight away." |
| 19:59 | <tantek> | a geo URL "identifies a physical location rather than a network resource" |
| 19:59 | <tantek> | from: http://tools.ietf.org/html/rfc5870#section-9 |
| 20:05 | <SimonSapin> | I’m writing an algorithm in English for css3-page that has lots of nested "if"s. How should I mark it up? |
| 20:06 | <SimonSapin> | I’m not satisfied with the presentation of http://dev.w3.org/csswg/css3-page/#margin-dimension and I’m rewriting the content anyway |
| 20:07 | <annevk> | SimonSapin: using <ol> might help a lot |
| 20:09 | <SimonSapin> | annevk: ok for <ol> instead of <ul>. Great for a sequence of steps. Would you use the same for a set of exclusive conditions? Eg. "if foo, <nested block> Otherwise, if bar, <nested block>" |
| 20:10 | <annevk> | depends on if there's more steps I suppose |
| 20:11 | <annevk> | if it's just two steps I might if it helps for clarity, otherwise just two paragraphs |
| 20:11 | <SimonSapin> | two paragraphs inside a <li> ? |
| 20:11 | <Yuhong> | annevk: Killing EOT support from IE from what modes? |
| 20:14 | <annevk> | Yuhong: all |
| 20:14 | <annevk> | SimonSapin: uhuh, <p> is your friend |
| 20:15 | <Yuhong> | annevk: Sometime MS does do this, but they don't take it lightly: |
| 20:15 | <Yuhong> | http://blogs.msdn.com/b/ie/archive/2012/06/04/legacy-dx-filters-removed-from-ie10-release-preview.aspx |
| 20:15 | <annevk> | Yuhong: aligning browsers is a long term goal, I don't expect immediate compliance |
| 20:16 | <Yuhong> | annevk: But why features in one browsers that don't exist at all in other browsers matter? |
| 20:17 | <annevk> | Yuhong: if that browser gets a dominant market position the other browsers will be forced to reverse engineer the proprietary features as they will get more wide deployment (see NN4, IE6, "mobile" WebKit) |
| 20:19 | <Yuhong> | annevk: Anyway, if you get appointed to the TAG, I suggest working on this: http://www.w3.org/wiki/Evolution |
| 20:20 | <Yuhong> | annevk: As I said before, it will help to prove that the W3C process and spec versioning is flawed. |
| 20:22 | <Ms2ger> | SimonSapin, depending on what kind of if, maybe a dl? |
| 20:23 | <SimonSapin> | Ms2ger: it’s the big nested list in http://dev.w3.org/csswg/css3-page/#margin-dimension |
| 20:23 | <SimonSapin> | <ol> is interesting, to be able to talk about "case 3.1.2 of the algorithm" |
| 20:24 | <Yuhong> | annevk: And on the Encoding Standard and IE, I think IE is pretty close already. |
| 20:25 | <Yuhong> | annevk: Do anyone really care that IE don't support JIS X 212 at this point? |
| 20:26 | <Yuhong> | annevk: In fact many of the encodings itself comes from IE. |
| 20:28 | <Yuhong> | "(see NN4, IE6, "mobile" WebKit)" Actually, it is NN1-3, IE 4-5, and yea mobile WebKit. |
| 20:29 | <Ms2ger> | SimonSapin, IMO, the 40-something em max-width is too small if you actually write things out |
| 20:29 | <annevk> | SimonSapin: any kind of list of steps really has to be <ol> imo |
| 20:30 | <Yuhong> | NN4 and IE6 was when the browsers began to stagnate. |
| 20:30 | <Yuhong> | And became a boat anchor. |
| 20:31 | <SimonSapin> | Ms2ger: Do you mean the max-width on body? That’s in all CSS EDs… |
| 20:31 | <Ms2ger> | SimonSapin, yeah, but most CSS EDs don't specify much |
| 20:31 | <Yuhong> | And even now, IE9 is not available for XP, causing IE8 to be another boat anchor. |
| 20:31 | <Yuhong> | Causing problems for things like XHTML (remember polyglot?). |
| 20:31 | <Ms2ger> | It works fine with short paragraphs that just hand-wave a bit |
| 20:33 | <Yuhong> | And DOM level 2, which is probably why jQuery 2.0 is not supporting IE8. |
| 20:33 | <SimonSapin> | Ms2ger: It’s currently at 50em and I think that more would be harder to read, but that’s not what I’m interested in right now |
| 20:34 | <Yuhong> | To be honest, XP was in extended support by the time IE9 was released. |
| 20:34 | <Ms2ger> | SimonSapin, and 70px+2em padding ;) |
| 20:34 | <Yuhong> | And IE9 required a platform update to work on Vista and IE10 requires a platform update to work on Win7. |
| 20:35 | <SimonSapin> | annevk: I see your point about <ol> and sequential steps, but I already use it for "branching" |
| 20:35 | <SimonSapin> | Ms2ger: Ok. But still not interested :p |
| 20:36 | <Yuhong> | And this reminds me that people often use the term "standard compliance" when "boat anchor" is meant. |
| 20:36 | <Ms2ger> | SimonSapin, sure, but that was the first thing that got my way when I looked at it ;) |
| 20:37 | <SimonSapin> | Ms2ger: you could ping fantasai about restyling specs |
| 20:38 | GPHemsley | is not a fan of numbered "otherwise" steps. |
| 20:38 | <Ms2ger> | Maybe I should use the geocities one instead :) |
| 20:38 | <GPHemsley> | Numbered steps should be self-contained, IMO. |
| 20:39 | <SimonSapin> | geocities.css has max-width on body too |
| 20:39 | <Ms2ger> | Foiled again |
| 20:42 | <SimonSapin> | GPHemsley: My algo has a "tree" of conditions, and only leaves have steps |
| 20:43 | <SimonSapin> | Numbering the steps is not that interesting. There are not many in one given leaf and they are close together |
| 20:56 | <GPHemsley> | SimonSapin: By "tree", you mean if/otherwise? |
| 20:56 | <GPHemsley> | if/elseif/otherwise |
| 20:56 | <GPHemsley> | or whatever |
| 21:08 | <SimonSapin> | GPHemsley: yes |
| 21:09 | <GPHemsley> | SimonSapin: And by "leaves", you mean...? |
| 21:10 | <SimonSapin> | each elsif contains either a nested series of if/elseif/else (a subtree) or a sequence of steps (a "leaf") |
| 21:15 | <GPHemsley> | SimonSapin: So... <li><p>if x, do: <ol></ol></p><p>elseif y, do: <ol></ol></p><p>otherwise, do: <ol></ol></p></li> ? |
| 21:15 | <GPHemsley> | (or thereabouts) |
| 21:20 | <SimonSapin> | GPHemsley: so your advice is to use <p> for if/else and <ol> for nesting? Even if one <ol> only has one <li> made of if/else? |
| 21:21 | <GPHemsley> | SimonSapin: I'm actually just trying to clarify the advice that others have given plus whatever you've done, as I've been trying to avoid "otherwise" clauses for want of how to mark them up. :) |
| 21:22 | <SimonSapin> | I can make conditions that happen to be exclusive and not write "otherwise" explicitly |
| 21:22 | <SimonSapin> | but that doesn’t solve how to nest them |
| 21:23 | <GPHemsley> | SimonSapin: Oh, yeah, that's what I've been doing: Making exclusive ifs |
| 21:24 | <GPHemsley> | <li>if x and y, do: <ol></ol></li><li>if x and z, do: <ol></ol></li> |
| 21:25 | <GPHemsley> | But perhaps <ul> does work here |
| 21:25 | <SimonSapin> | so you do not nest and repeat x as needed? |
| 21:25 | <SimonSapin> | What if x is lengthy to describe? |
| 21:26 | <GPHemsley> | hmm |
| 21:26 | <GPHemsley> | SimonSapin: I've been able to avoid such a situation thus far :) |
| 21:26 | <GPHemsley> | hmm |
| 21:27 | <SimonSapin> | well, I did not pick the easiest first edit :] |
| 21:27 | <GPHemsley> | <li>If x, do: <ol><li>If y, do: <ol></ol></li><li>If z, do: <ol></ol></li></ol></li> |
| 21:29 | <GPHemsley> | I have managed to have only a single instance of the word "otherwise" in mimesniff |
| 21:29 | <GPHemsley> | and 0 instances of "else" |
| 21:29 | <GPHemsley> | " A no-sniff flag, which defaults to set if the user agent does not wish to perform sniffing on the resource and unset otherwise. " |
| 21:29 | <SimonSapin> | Looks good. |
| 21:29 | <SimonSapin> | And for sequential steps at the inner most level, switch to <p> or keep <ol>? |
| 21:30 | <GPHemsley> | But I do recall having gone through extra effort to ensure that the default step come after a bunch of steps that end in "abort these steps" or "return x" |
| 21:30 | <GPHemsley> | SimonSapin: Keep <ol> for sequential steps |
| 21:31 | <GPHemsley> | sequential steps should always be <ol> |
| 21:31 | <GPHemsley> | it's the if/elseif/else markup that's murky |
| 21:31 | <GPHemsley> | but if it's just "do x. do y. do z.", that part's clear |
| 21:32 | <GPHemsley> | Avoid having more than one instruction per <li> |
| 21:32 | <GPHemsley> | (unless it's an abort or similar) |
| 21:32 | <GPHemsley> | and FTR, this is all IMO |
| 21:33 | <SimonSapin> | your opinion is better than me trying stuff at random :) |
| 21:33 | <GPHemsley> | :) |
| 21:34 | <SimonSapin> | Markup is settled, back to figuring out the actual algorithm. (Layout is hard.) |
| 21:34 | <GPHemsley> | :) |
| 21:35 | <GPHemsley> | I have a lot of steps that go "1. x = y; 2. if x is not undefined, abort. 3. x = z; if x is not undefined, abort. ..." |
| 21:36 | <GPHemsley> | but I definitely use abort/return in place of if/elseif/else |
| 21:36 | <GPHemsley> | don't know if that's good, but it avoids the markup issue |
| 21:49 | <annevk> | SimonSapin: not sure what branching is, but any spec algorithm I know uses <ol> |
| 21:49 | <SimonSapin> | branching is if/elseif/else |
| 21:52 | <annevk> | so I guess what we want is that about/blob/data URLs are treated on equal footing; e.g. if your API allows one of those it should allow the others too |
| 21:52 | <annevk> | javascript is a special case, only allowed when okay |
| 21:52 | <annevk> | then ws/wss are only allowed in WebSocket |
| 21:52 | <GPHemsley> | annevk: Are you splitting URLs up into types? hierarchical, actions, scripting? |
| 21:52 | <annevk> | http/https work almost anywhere |
| 21:53 | <annevk> | GPHemsley: more like "can be fetched" / "cannot be fetched" |
| 21:54 | <GPHemsley> | and you're leaving room for new schemes, right? |
| 21:54 | <annevk> | not sure what that means |
| 22:00 | <annevk> | What I'm trying to figure out what the best model is for XHR / <img> / <script> / <object> / <iframe> / <svg:image> and anything else that takes a URL, going forward |
| 22:01 | <SimonSapin> | annevk: for these anything that can not fetch a resource is an error, isn’t it? |
| 22:02 | <SimonSapin> | where "can fetch a resource" means return a sequence of bytes with a content-type |
| 22:02 | <annevk> | SimonSapin: yeah, although <iframe> and <object> might be special come to think of it |
| 22:02 | <SimonSapin> | how so? |
| 22:03 | <annevk> | some stuff goes directly to "fetch", some stuff does "navigate" first |
| 22:03 | <annevk> | <iframe> does "navigate" |
| 22:03 | <annevk> | <object> might too, not entirely sure |
| 22:03 | <SimonSapin> | does "navigate to mailto:" make sense? |
| 22:03 | <annevk> | yes, that is what <a> does |
| 22:03 | <SimonSapin> | in <iframe> I mean |
| 22:05 | <annevk> | seems that depends on the UA |
| 22:07 | <SimonSapin> | oh, I remember a story about Android loading tel: URLs in <iframe> which was enough to trigger IMEI magic numbers |
| 22:08 | <annevk> | this area has been filled with exploits, which is in part why we should define it more carefully I think |
| 22:09 | <SimonSapin> | I don’t see a use case for tel: or mailto: on anything but <a> |
| 22:09 | <annevk> | I guess I have to talk with Hixie at some point how to best organise it, for now just defining the processing steps for various schemes is prolly enough work |
| 22:09 | <annevk> | SimonSapin: sure, but what if the <a> has a target that points to the <iframe>? |
| 22:09 | <annevk> | SimonSapin: in the end all <a> does is manipulate the top-level <iframe> that is not exposed |
| 22:10 | <annevk> | SimonSapin: there's something to be said for magic top-level <iframe> and normal <iframe>s to have a different model of course |
| 22:26 | <GPHemsley> | annevk: What are the problems with EOT? |
| 22:27 | <annevk> | it's a proprietary format |
| 22:28 | <GPHemsley> | even given http://www.w3.org/Submission/EOT/ ? |
| 22:28 | <annevk> | yeah, because nobody plans to implement it |
| 22:28 | <annevk> | you can have all the documentation you want, but if there's just one implementation it's proprietary |
| 22:28 | <GPHemsley> | o_0 |
| 22:29 | <GPHemsley> | Yet another use of "proprietary" that does not coincide with how I understand the word |
| 22:32 | <zewt> | "proprietary" just means "owned", it has nothing to do with how many implementations there are |
| 22:37 | <annevk> | you can call it something else if you like, substance is the same |
| 22:37 | <GPHemsley> | well, not really... because I was looking for reasons as to why nobody plans to implement it |
| 22:38 | <annevk> | because we have too much font formats already |
| 22:38 | <annevk> | well, woff is getting a new version of sorts I believe, but nobody has plans for EOT |
| 22:39 | <GPHemsley> | having "too many already" is far from a substantial reason |
| 22:40 | <annevk> | o_O |
| 22:40 | <annevk> | nn |
| 22:40 | <GPHemsley> | to also have too many image formats, video formats, encodings, ... |
| 22:40 | <GPHemsley> | s/to /we / |