| 00:34 | <annevk3> | "HTML5 Web Workers" You'd think that the Google guys of all people know that Web Workers was never in HTML5 to begin with :) |
| 00:34 | <Hixie> | it's in html5 now! |
| 00:35 | <Hixie> | well it's in webapps/current-work/source |
| 00:35 | <Hixie> | but yeah |
| 00:35 | <Hixie> | we need a new name to refer to all these technologies |
| 00:40 | <Hixie> | annevk3, i might need you to do something for sync xhr (namely, release the storage mutex) |
| 00:40 | <Hixie> | i'll let you know later. |
| 00:46 | <dave_levin> | annevk3: The html5 prefix was an oversight. I'll see if I can get it fixed. |
| 02:45 | <MikeSmith> | Hixie: i'm wondering what an "earthquake prevention kit" is |
| 03:24 | <jcranmer> | (long story) |
| 06:47 | gsnedders | needs to wake up |
| 06:51 | zcorpan | enumerates window in ie8 |
| 06:51 | <zcorpan> | window.maxConnectionsPerServer=6 |
| 06:51 | <zcorpan> | onhelp=null |
| 06:52 | <zcorpan> | offscreenBuffering=auto |
| 06:53 | <takkaria> | gsnedders: boo |
| 06:54 | <zcorpan> | they have a clientInformation attribute that seems identical to navigator |
| 06:54 | gsnedders | jumps in fright |
| 06:54 | <gsnedders> | All these southerners! |
| 07:26 | <gsnedders> | I need to learn to spell. |
| 07:26 | <hsivonen> | interesting comments at http://almaer.com/blog/canvas-3d-standards-and-where |
| 07:28 | <hsivonen> | annevk3: Re: validator.nu messages: perhaps I should just bite the bullet and merge George Bina's improvements directly on the v.nu branch of Jing... |
| 07:31 | <hsivonen> | from the text editor authoring perspective, I'd be unhappy with required quotess on attributes |
| 07:51 | <Hixie> | annevk3: yt? |
| 07:58 | <zcorpan> | hsivonen: checking just quotes would be pointless. it's better to check full polyglotness |
| 08:00 | <hsivonen> | zcorpan: who'd be my users for a full polyglotness check? I can think of two potential users right now. |
| 08:01 | <zcorpan> | hsivonen: who other than sam? |
| 08:04 | <hsivonen> | zcorpan: the other polyglot publisher I had in mind was Jacques Distler |
| 08:04 | <zcorpan> | ah |
| 08:05 | <zcorpan> | yep and it would be nice for people who come from xhtml 1.0 validation and don't want to regress their "strictness" checking |
| 08:06 | <zcorpan> | although i would only include the checkbox in the advanced interface :) |
| 08:06 | <hsivonen> | zcorpan: oh yeah, I have an open feature request on XHTML strictness stuff |
| 08:07 | <hsivonen> | zcorpan: I have feedback (in addition to feedback from you) that it would be a killer XHTML 1.0 to HTML5 evangelism feature |
| 08:07 | <zcorpan> | hsivonen: but did you plan to implement that as full polyglotness checking or just warning on implied tags? |
| 08:08 | <gsnedders> | jgraham / annevk3: Can one of you add seanecoates to html5lib? |
| 08:09 | <hsivonen> | zcorpan: I was thinking of implementing warning on implied tags. |
| 08:13 | <zcorpan> | hsivonen: partial polyglotness checking is already possible but it's hard to discover and requires several changes from the default |
| 08:14 | <zcorpan> | hsivonen: i think it would be nice with a checkbox that validated as both html and xhtml and maybe check stuff like <pre>\n |
| 08:15 | <jgraham> | gsnedders: who is he? |
| 08:16 | <MikeSmith> | zcorpan: about http://simon.html5.org/test/validator/content-model/label/input-in-label-with-for.html |
| 08:16 | <jgraham> | And is that a gmail address? |
| 08:17 | <MikeSmith> | zcorpan: the change I made to assertions.sch will catch that as an error |
| 08:18 | <MikeSmith> | but need to change it in the java assertions code that v.nu actually uses, in order for v.nu to be able to report it as an error |
| 08:18 | <hsivonen> | MikeSmith: I deployed your patches from yesterday. Thanks! |
| 08:18 | <MikeSmith> | hsivonen: cool |
| 08:18 | <hsivonen> | zcorpan: I think people who'd appreciate <pre>\n correctness as opposed to being confused by it would be very few :-/ |
| 08:20 | <MikeSmith> | hsivonen: the case where the id() checking assertions.sch results in suboptimal reporting is <label for=foo>bar <input id=foo><input id=foo></label> |
| 08:20 | <MikeSmith> | the duplicate IDs will get reported by the validator (if it's doing ID checking) |
| 08:21 | <hsivonen> | MikeSmith: ok (yes, the filter does duplicate checking) |
| 08:21 | <MikeSmith> | so assertions.sch, checking by itself, will only report the second <input id=foo> instance as an error |
| 08:22 | <olliej> | annevk3: yt? |
| 08:22 | <zcorpan> | hsivonen: yes probably |
| 08:23 | <MikeSmith> | hsivonen: understood. I meant in the general case -- where somebody might be using assertions.sch outside of v.nu |
| 08:23 | <hsivonen> | MikeSmith: I think it's ok to assume a separate duplicate checking layer |
| 08:23 | <MikeSmith> | OK |
| 08:24 | <MikeSmith> | so actually I mean for the case <label for=foo>bar <input id=goo><input id=goo></label> |
| 08:24 | <MikeSmith> | where the ID and @for don't match |
| 08:25 | <MikeSmith> | in that case, it will report "The “for” attribute of the “label” element must refer to a form control." |
| 08:25 | <MikeSmith> | which it was already doing, and is expected |
| 08:25 | <MikeSmith> | but also will report |
| 08:29 | <MikeSmith> | will report 'An "input" descendant of a "label" element with a "for" attribute must have a unique ID value that matches that "for" attribute.' only for the second <input> instance |
| 08:30 | <MikeSmith> | I think it would be better if it reported it for the first as well, but I don't know how to make that happen just with XPath and id() |
| 08:30 | <MikeSmith> | anyway, it's not a big deal |
| 08:33 | <MikeSmith> | hsivonen: fwiw, what it does exactly is, when it hits the test "not(count(.|id(ancestor::h:label/@for)) = count(id(ancestor::h:label/@for)))", it passes for the first <input> instance with 1=1 |
| 08:34 | <MikeSmith> | but when it hits the test for it for the second <input> instance, it fails as expected with 2 !=1 |
| 08:34 | <MikeSmith> | (2 nodes that have an ID that's the same as the value of @for) |
| 08:35 | <hsivonen> | hmm. what does .|if(...) mean? |
| 08:35 | <MikeSmith> | it seems like that test should fail for <input> instance as well, but I don't understand why it doesn't |
| 08:35 | <hsivonen> | um .|id(...) |
| 08:35 | <gsnedders> | jgraham: Rather well known PHP developer. That is his Google username. |
| 08:36 | <MikeSmith> | hsivonen: it means, take the union of the context node and any label nodes with @for that ancestors of the context node |
| 08:37 | <MikeSmith> | and make it a nodeset |
| 08:37 | <MikeSmith> | and then the count() just counts the number of nodes in that nodeset |
| 08:37 | <hsivonen> | ah so that the context node disappears in the count if it's already in the set returned by id()? |
| 08:37 | <MikeSmith> | right, exactly |
| 08:37 | <hsivonen> | ok makes sense. |
| 08:38 | <MikeSmith> | that's the main way to check for node equivalence in pure XPath |
| 08:38 | <MikeSmith> | afaik |
| 08:38 | <annevk3> | olliej, Hixie, pong (for a few minutes) |
| 08:38 | <hsivonen> | ok. I don't find XPath very intuitive, but then I have never learned it properly. |
| 08:38 | <MikeSmith> | well, I've never learned it properly either |
| 08:39 | <hsivonen> | MikeSmith: you have better solutions that I had. The schematron checks around ARIA IDREFS are still subtly wrong |
| 08:40 | <hsivonen> | s/that/than/ |
| 08:40 | <Hixie> | annevk3: xhr |
| 08:40 | <hsivonen> | (the Java checks around ARIA IDREFS should be correct, though) |
| 08:40 | <olliej> | annevk3: apparently getComputedStyle does weird things wrt text-shadow in o9.6 and 10a |
| 08:40 | <Hixie> | annevk3: can we add "Release the storage mutex." as a requirement somewhere near the start of the sync XHR process |
| 08:41 | <Hixie> | annevk3: and can we add "obtain the storage mutex" before you apply cookies and "release the storage mutex" afterwards? |
| 08:41 | <MikeSmith> | well, maybe I can take a look at those in assertions.ssh too at some point. I'm not sure how valuable it will be in the long run to have assertions.sch in sync, but it seems worth trying to for now at least. I can imagine that some might find it useful with other toolchains |
| 08:41 | <Hixie> | annevk3: (for both sync and async) |
| 08:41 | <annevk3> | Hixie, likely, but please email |
| 08:41 | <Hixie> | annevk3: (er, the releases are only for sync) |
| 08:41 | <Hixie> | annevk3: ok |
| 08:42 | <MikeSmith> | hsivonen: and having it might save them some time from trying to go through and re-invent it themselvs |
| 08:42 | <annevk3> | olliej, file a bug? :) |
| 08:42 | <olliej> | annevk3: he has |
| 08:43 | <annevk3> | olliej, https://bugs.opera.com/wizard/ |
| 08:43 | <annevk3> | olliej, ok, great |
| 08:43 | <MikeSmith> | hsivonen: If I could get Norm Walsh to look at it, he could write it in his sleep. me with XPath, it's like I'm rubbing sticks together to make fire |
| 08:44 | <hsivonen> | MikeSmith: I asked on the schematron list, but the suggestions I got were really complex |
| 08:45 | <MikeSmith> | I see |
| 08:45 | <hsivonen> | MikeSmith: in fact, I should take another look and report back to the list. The situation seemed so hard that I've postponed taking a good look for way too long. |
| 08:46 | <Hixie> | annevk3: sent, thanks |
| 08:47 | <Hixie> | well i should sleep |
| 08:47 | <Hixie> | nn |
| 08:47 | <MikeSmith> | hsivonen: well, along with the table-integrity case, maybe it's a case where, while it may be possible in schematron, the complexity and inelegance of doing it in schematron make it a suboptimal means for checking of that particular constraint |
| 08:48 | <hsivonen> | MikeSmith: the ARIA IDREFS stuff is very near the simple end of spectrum of constraints, so I think it's a pretty big failure for XPath if it really is hard to express |
| 08:50 | <hsivonen> | MikeSmith: IIRC, the problem is this: |
| 08:50 | <hsivonen> | with IDREFS, I want to check that 1) all the IDs match and 2) that a condition hold true for all of them |
| 08:50 | <hsivonen> | however, id() tells me if there were zero matches or > 0 matches |
| 08:51 | <hsivonen> | when I want to know that there were as many matches as there were tokens |
| 08:51 | <hsivonen> | except when the tokens are duplicates of each other |
| 08:51 | <MikeSmith> | OK, I see |
| 08:52 | olliej | admires modifications to the semantics of standard web content to suit a single UA |
| 08:52 | <hsivonen> | MikeSmith: IIRC, the problem is that pure XPath 1.0 doesn't allow me to count the # of space-separated tokens in an attribute value |
| 08:52 | <hsivonen> | MikeSmith: even if I had a RELAX NG -level check for duplicate tokens |
| 08:53 | <MikeSmith> | hsivonen: yeah, that limitation in XPath 1.0 sounds right to me. maybe it's possible in 2.0 but that doesn't help us here |
| 08:53 | <hsivonen> | maybe it does if we are OK with using a schematron flavor that is tied to Saxon 9 |
| 08:57 | <MikeSmith> | hsivonen: maybe that'd be a better solution that to struggle with XPath 1.0 limitations |
| 08:58 | <MikeSmith> | but there are still a lot of people using XPath 1.0 tools like libxml2 |
| 08:58 | <MikeSmith> | libxslt2 |
| 08:58 | <hsivonen> | it might also be worthwhile to see what interoperable extensions EXSLT offers around this area |
| 08:58 | <MikeSmith> | ah yeah |
| 08:59 | <MikeSmith> | but that wouldn't do any good for a non-XSLT schematron implementation like the one in jing |
| 09:01 | <hsivonen> | MikeSmith: the implementatino in Jing is an XSLT implementatino |
| 09:01 | <hsivonen> | *tion |
| 09:02 | <MikeSmith> | hsivonen: oh, OK, didn't know |
| 09:04 | <hsivonen> | I wonder if I should try to address some of the SVG-in-text/html concerns by provide HTML2XML as a Web-based app |
| 09:13 | <Philip`> | hsivonen: Like http://services.philip.html5.org/html-to-xhtml/ ? |
| 09:14 | <hsivonen> | Philip`: oh cool. Yeah, that what I had in mind except I considered allowing the user to paste the URL of a remove HTML document |
| 09:14 | <hsivonen> | and have the result download with Content-Disposition: attachment |
| 09:14 | Philip` | didn't do that since he didn't want to have to worry about the security implications |
| 09:18 | <jgraham> | gsnedders: Done |
| 09:20 | jgraham | wishes for some system that didn't involve making uninformed decisions about who to add as contributers whilst still keeping the barrier to entry low |
| 09:22 | <jgraham> | (a DVCS seems like it would help here) |
| 09:23 | <Philip`> | jgraham: Like a DVCS, so everyone can be their own committer, and you judge their work to decide whether to merge it? |
| 09:23 | <Philip`> | Oh, good idea |
| 09:24 | <hsivonen> | is there a hosting service for hg that'd host some kind of html5lib-central? |
| 09:25 | <Philip`> | GitHub? (Not technically Hg but that shouldn't matter much) |
| 09:25 | <jgraham> | BitBucket |
| 09:25 | <hsivonen> | Philip`: I'm aware of GitHub, I'm interested in a service for hg |
| 09:26 | <hsivonen> | jgraham: thanks |
| 09:27 | jgraham | wonders if he could host html5lib on bitbucket and set up a post-commit hook to automatically push changes to google code |
| 09:27 | hsivonen | wonders if it were feasible to put V.nu on bitbucket |
| 09:27 | <hsivonen> | Is there a proper Eclipse plug-in for hg? |
| 09:28 | <hsivonen> | where the built-in CVS thingy and Subclipse count as being proper enough |
| 09:29 | <jgraham> | http://www.vectrace.com/mercurialeclipse/ ? |
| 09:30 | <hsivonen> | jgraham: have you tried it? |
| 09:32 | <jgraham> | hsivonen: No. I have nevber even used Eclipse :) |
| 09:32 | <jgraham> | *never |
| 09:34 | <hsivonen> | ok |
| 09:46 | <hsivonen> | why do widgets use xmldsig instead of signing the zip content bytes? |
| 09:46 | <hsivonen> | roughly like jars are signed |
| 09:50 | <hsivonen> | XML signatures are fun, because they depend on people being able to write bug-free XML serializers |
| 09:51 | <hsivonen> | Philip`: you could have a field day finding bugs that upon fixing theoretically break existing signatures |
| 09:54 | <Philip`> | http://www.smspower.org/ |
| 09:54 | <Philip`> | http://www.smspower.org/scalefix.js - using <canvas> to scale images |
| 09:55 | <Philip`> | (and, incidentally, using getImageData in a way that will break if 1 coordinate space unit != 1 canvas ImageData pixel) |
| 10:06 | <roc> | heh |
| 10:06 | <roc> | I tried to persuade Ollie |
| 10:07 | <olliej> | roc: safari has always supported canvas where 1css px != 1 canvas px |
| 10:07 | <olliej> | rothat was in the original definition |
| 10:07 | <roc> | ok |
| 10:07 | <roc> | it's still going to break |
| 10:08 | <Philip`> | hsivonen: How would that be possible? Presumably people only sign sane documents (not ones intentionally trying to trigger bugs), and it doesn't seem that hard to serialise Canonical XML in sane cases, so there'd never be any need to change it for obscure bugs |
| 10:08 | <roc> | (getImageData, the way it is defined) |
| 10:08 | <hsivonen> | Philip`: note "theoretically" |
| 10:08 | <hsivonen> | Philip`: IIRC, Apache has had a bug in their signing serializer |
| 10:09 | <Philip`> | hsivonen: (I assume by "break" you mean "cause not to work", rather than "allow the creation of an arbitrary document with the same hash" or something) |
| 10:09 | <olliej> | roc: this is going to become another example of a "feature" that drastically harms the future-proofness of a basic tech |
| 10:09 | <hsivonen> | Philip`: so *theoretically* it's possible that someone has signed a document that triggered the bug |
| 10:09 | <olliej> | roeg. moz exposiing and encouraging 1px = 1canvas px basically forces low res canvas for ever |
| 10:09 | <olliej> | roc |
| 10:09 | <hsivonen> | Philip`: after all, I know real people have hit the bug |
| 10:09 | <roc> | we're not encouraging that |
| 10:09 | <roc> | that's just the way it is for Mozilla users and almost all Safari users |
| 10:10 | <roc> | so that's all they test |
| 10:10 | <hsivonen> | Philip`: by break I mean that validating a signature says the signature is bogus |
| 10:10 | <roc> | so their sites will break when used by the other users |
| 10:10 | <Philip`> | olliej: It seems there should still be other solutions, e.g. canvas could be high-res but getImageData returns a low-res 1:1 version of the pixels insteaad (and loses data) |
| 10:10 | <olliej> | yup |
| 10:10 | <roc> | Philip`: That's what I was pushing |
| 10:10 | <hsivonen> | Philip`: and then some system refuses to talk with some other software |
| 10:10 | <olliej> | and so i will eventually have to code a routine that subsamples canvas to avoid breaking compatibility with firefox |
| 10:10 | <olliej> | yay |
| 10:11 | <roc> | no |
| 10:11 | <Philip`> | It's not about compatibility with Firefox, it's about compatibility with people's perceptions of <canvas> regardless of what browser they tested in |
| 10:11 | <roc> | you have to avoid breaking compatibility with all the Safari users who have normal dpi screens |
| 10:12 | <roc> | all the Web developers who tested their site on Safari with a normal screen and assumed that would work for everyone else |
| 10:12 | <olliej> | Philip`: the alternative is that we throw out the idea that canvas actually looks good on high dpi displays |
| 10:12 | <hsivonen> | for the same reason, I don't believe browser-based CSS renderers could ever change the px to pt ratio |
| 10:12 | <olliej> | roc: um, i have yet to see a getimagedemo written by someone other than me that looked at safari first |
| 10:12 | <roc> | hsivonen: we do |
| 10:12 | <olliej> | hsivonen: we do |
| 10:12 | <roc> | on trunk |
| 10:12 | <hsivonen> | whoa! |
| 10:12 | <roc> | on Firefox 3 actually |
| 10:12 | <hsivonen> | does it break sites? |
| 10:12 | <olliej> | hsivonen: have done since S2 |
| 10:12 | <roc> | no |
| 10:13 | <hsivonen> | I'm surprised |
| 10:13 | <roc> | that works because you don't need an unusual screen to see differences |
| 10:13 | <olliej> | hsivonen: everything gets scaled appropriately |
| 10:13 | <roc> | olliej: do you really not understand what I'm saying? |
| 10:13 | <hsivonen> | but who has Safari or Firefox configured with an unusual px to pt ratio? I have no idea how to configure that. |
| 10:14 | <olliej> | hsivonen: because realistically points, inches, etc are in terms of css px |
| 10:14 | <olliej> | hsivonen: absolute units are notional |
| 10:14 | <olliej> | hsivonen: the ratio between css units and device pixels can changes arbitrarily |
| 10:14 | <olliej> | and our canvas impl reflects this |
| 10:14 | <hsivonen> | olliej: device pixels, sure. I meant the ratio of CSS px and CSS pt |
| 10:15 | <olliej> | hsivonen: oh hell no we don't do that, that would i'm fairly sure be distastrous --don't know about moz though |
| 10:15 | <olliej> | roc: are you saying moz will vary css px : css pt ratio? |
| 10:16 | <roc> | yeah |
| 10:16 | <roc> | since FF3 |
| 10:17 | <roc> | we detect your screen DPI |
| 10:17 | <roc> | and set pt appropriately |
| 10:17 | <roc> | we had a few complaints IIRC |
| 10:17 | <hsivonen> | roc: how does the detection work? |
| 10:17 | <roc> | but very little really |
| 10:17 | <olliej> | we explicitly do not |
| 10:17 | <roc> | hsivonen: platform APIs |
| 10:17 | <roc> | you can also override dpi explicitly |
| 10:17 | <roc> | open about:config and modify layout.css.dpi to experiment |
| 10:18 | <olliej> | roc: we just change the entire context so that we may have multiple device pixels per css px |
| 10:18 | Philip` | wonders how the platform knows what size monitor he's plugged in |
| 10:18 | <hsivonen> | roc: do many people on desktop have those APIs returning non-traditional values? |
| 10:18 | <roc> | I think those APIs are pretty accurate on Mac and Windows. X is a mess as usual |
| 10:18 | <roc> | so you get what your screen is |
| 10:19 | <roc> | and screens do vary quite a bit these days |
| 10:19 | <hsivonen> | roc: I thought the Windows API dependend on a user setting that has had two values since Windows 3: 96 and 120 |
| 10:20 | <hsivonen> | Do DVI and VGA have a mechanism for reporting the actual dpi of the device? I had always though something like this could only work with Apple displays and Apple OS |
| 10:20 | <hsivonen> | anyway, it's news to me that those APIs return something useful on Mac and Windows |
| 10:20 | <roc> | modern Windows can do that |
| 10:21 | <Philip`> | Vista lets you pick anything from 96dpi to 480dpi |
| 10:21 | <Philip`> | (via the Control Panel) |
| 10:22 | <hsivonen> | Philip`: but does it autodetect the default? |
| 10:22 | <roc> | hsivonen: we use GetDeviceCaps(dc, LOGPIXELSY) ... I can't guarantee that works correctly on Windows, but it's supposed to |
| 10:22 | <Philip`> | hsivonen: I think it always defaults to 96dpi |
| 10:22 | <hsivonen> | Philip`: that was what I was after |
| 10:23 | <Philip`> | which it labels the "Default scale" (vs 120dpi being "Larger scale", and the others being behind the "Custom DPI" button) |
| 10:23 | <roc> | I know we're getting some sort of right DPI on X because people have complained about it |
| 10:23 | <hsivonen> | roc: my guess is s/right/different/ :) |
| 10:23 | <roc> | looks like on Mac we're not doing the right thing |
| 10:24 | <Philip`> | People like to think of their computer displays as being pixels, not as being a pixel-based approximation to an inch-based vector image, so DPI changes are weird and confusing :-) |
| 10:25 | <roc> | even if we're only getting 96 or 120 most of the time, it's OK, it still breaks people making assumptions about pt :-) |
| 10:25 | <zcorpan> | Hixie: you should have pushed harder for aria-role back in 2007 when it was first suggested |
| 10:26 | <roc> | olliej: anyway, getImageData as specified is stuffed unless you can arrange for most Web developers to see a ratio of CSS pixels to dev pixels that's not 1 |
| 10:27 | <olliej> | roc: yeah, currently it's not sufficient :-/ |
| 10:27 | <Philip`> | Good luck on educating web developers :-p |
| 10:27 | <olliej> | roc: i think i commented on this in whatg ml at some point |
| 10:27 | <roc> | yeah, we had a long discussion about it |
| 10:28 | <roc> | you won, and got this getImageData API |
| 10:28 | <roc> | "won" |
| 10:28 | <olliej> | hehe |
| 10:29 | <roc> | IMHO you should give up, let getImageData be lossy, and in a few years when Web devs have got super-high-dpi screens, create getImageDataEx |
| 10:29 | olliej | cries |
| 10:29 | <olliej> | robut yes |
| 10:29 | <olliej> | i suspect you are right |
| 10:30 | <Philip`> | The API should have been designed to pick a random value between 1 and 1.5, then multiply all the sizes by that value before extracting the pixels and returning the ImageData |
| 10:31 | <Philip`> | That way people couldn't rely on it having a 1:1 mapping to canvas coordinate space units, even if they only test in a single browser |
| 10:31 | <olliej> | Philip`: hehe |
| 10:57 | <Dashiva> | Philip`: So people can run the function N times and extract the original value instead? :) |
| 11:43 | <hsivonen> | w00t! finally passing http://hixie.ch/tests/adhoc/dom/level0/write/003.html |
| 11:44 | <hsivonen> | starting an HTML load and ending it turned out to be much hairier than I imagined |
| 11:50 | <Philip`> | When people argue that writing an HTML parser is much harder than writing an XHTML parser, it sounds like they're actually right (though because of document.write rather than any syntax issues) |
| 11:58 | <wilhelm_> | Hixie: Yes, but I haven't had time to wrap it up. Please ask the same question again on Sunday. (c: |
| 11:59 | <Philip`> | data:text/xml,<html xmlns='http://www.w3.org/1999/xhtml'><body><div><script>var d = document.getElementsByTagName('body')[0]; d.parentNode.removeChild(d)</script></div>test</body></html> |
| 11:59 | <Philip`> | Looking at that in Dragonfly in Opera, it looks like the "test" text is a sibling of the root html element |
| 11:59 | <Philip`> | which surely isn't good |
| 11:59 | <Philip`> | since it kind of defeats the point of having a root element |
| 12:00 | <Philip`> | (Opera 9.64 in particular) |
| 12:01 | <virtuelv> | Philip`: known issue, IIRC |
| 12:03 | <Philip`> | (If I wrap the script in two divs then the text vanishes entirely - I presume the </div>s are causing the open element stack to be popped even if the div elements have already been removed from it via scripts) |
| 12:03 | <Philip`> | I guess this is where an XML parser spec would be nice :-) |
| 12:09 | <hsivonen> | Philip`: actually, in this case, the problem wasn't document.write, but the the interaction of global object association with the document, the order of calls into the parser, managing the readiness of the document and scripts as a group and deferring the load event |
| 12:09 | <hsivonen> | Philip`: all issues that apply to application/xhtml+xml |
| 12:10 | <hsivonen> | Philip`: and part of it was Gecko-specific quirkiness: calls to the parser before the global scope is ready |
| 12:10 | <hsivonen> | Philip`: and onload deferring being tied to the global scope in a non-obvious way |
| 12:22 | <rubys> | hsivonen: just curious: how quickly could you develop and deploy polyglotness testing? Even if it was significantly incomplete, I think it would be valuable for educational reasons. |
| 12:24 | <hsivonen> | rubys: I don't know. Forming a good estimate would involve doing the work of thinking about what the issues are, which is the difficult part. :-) |
| 12:25 | <rubys> | http://wiki.whatwg.org/wiki/HTML_vs._XHTML |
| 12:25 | <rubys> | Again, I do think even if it was significantly incomplete, it would be very valuable. Everywhere I turn I run into people who think they are producing XHTML. |
| 12:26 | <rubys> | It often is very hard to convince them otherwise. |
| 12:26 | <hsivonen> | rubys: are you looking for stuff like <pre>\n and <input type="hidden" value='\n'>? |
| 12:27 | <hsivonen> | rubys: or just attribute quotes and implied tags? |
| 12:27 | <rubys> | I'd like to see flagged. Is it valid in XHTML1? Yes. But does it enable you to read the page with a random XML parser? Generally, no. |
| 12:27 | <hsivonen> | ouch |
| 12:27 | <rubys> | ouch as in that's hard? |
| 12:28 | <hsivonen> | ouch as in you seem to be looking for tokenizer-level things instead of tree builder level |
| 12:28 | <rubys> | "In XHTML: <![CDATA[...]]> is a CDATA section. In HTML, it's a bogus comment. " |
| 12:28 | <Lachy> | that wiki page seems a little incomplete now. It doesn't mention the <pre>\n or <input value="\n"> issues, and needs to be updated with SVG and MathML stuff |
| 12:28 | <hsivonen> | detecting which entity is being expanded is doable, but that code is in the 'here be dragons' territory |
| 12:29 | <hsivonen> | hmm. perhaps I already have a check for apos anyway |
| 12:29 | <hsivonen> | so not too hard, just an area that I don't like poking :-) |
| 12:30 | <rubys> | It would also help resolve the svg in html issue. :-) |
| 12:34 | <rubys> | Lachy: my guess is that the wiki page will always be incomplete. |
| 12:35 | <Lachy> | sure, but I should find some time to keep it as complete as possible |
| 12:36 | <zcorpan> | hsivonen: why not validate twice to do polyglotness checking? |
| 12:36 | <hsivonen> | rubys: for a practical resolution of the issue by writing software, I think it would be more useful to make the converter service that allows remote URLs |
| 12:37 | <hsivonen> | zcorpan: that would be way too simple :-) |
| 12:37 | <rubys> | html5lib comes close to being a converter service |
| 12:37 | <zcorpan> | hsivonen: why reject the simple solution? :) |
| 12:37 | <hsivonen> | zcorpan: I guess the real issue is conneg serving different bytes if the validator dereferences the URI twice |
| 12:37 | <Lachy> | zcorpan, because validation alone doesn't catch all polyglot incompatibilities |
| 12:37 | <rubys> | it could easily be put online. Perhaps even with appengine |
| 12:38 | <hsivonen> | rubys: Philip` already put a converter online |
| 12:38 | <hsivonen> | rubys: it just lacks one feature |
| 12:38 | <hsivonen> | rubys: but to be fair to Philip` I, too, shyed away from that feature earlier for security and legal worries |
| 12:38 | <zcorpan> | hsivonen: couldn't you use the same stream of bytes twice? |
| 12:38 | <Lachy> | zcorpan, e.g., it woudln't catch that <pre>\n issue, since its valid with or without the new line, but the incompatibility is caused by the different processing requirements |
| 12:39 | <hsivonen> | zcorpan: not without some infrastructure work |
| 12:39 | <rubys> | what's that feature? |
| 12:39 | <hsivonen> | rubys: being able to enter an arbitrary URL that dereferences into text/html and the service proxies it to the client as application/xhtml+xml |
| 12:40 | <hsivonen> | rubys: but I think the real risk seems negligible |
| 12:40 | <rubys> | I would add one more worry: server load |
| 12:41 | <hsivonen> | rubys: that shouldn't be worse than the current parse tree service unless someone in start using the service as an open proxy to work around censorship or user tracking |
| 12:41 | <rubys> | Many requested that the feed validator produce a feed of results. I don't believe that I could survive the load of having gazillions of people polling their pages once an hour. |
| 12:42 | <hsivonen> | s/ in // |
| 12:42 | <hsivonen> | rubys: have you suggested that those who ask it use inursite? |
| 12:43 | <anne-m> | use the same stream of bytes and compare the output trees? quite a bit of overhead though |
| 12:43 | <rubys> | didn't know about inursite? |
| 12:44 | anne-m | wonders why typing an opening parenthesis and ampersand don't work in mibbit |
| 12:44 | rubys | likes anne-m's thought |
| 12:44 | <zcorpan> | anne-m: no, assume the trees are the same but have checks in the html parser for the differences |
| 12:45 | <hsivonen> | http://inursite.com/ |
| 12:45 | <rubys> | oh, I found it. I also found the page where Ryan says he's moving on |
| 12:45 | <hsivonen> | oh |
| 12:45 | <anne-m> | zcorpan, that seems more complex |
| 12:45 | hsivonen | hasn't seen a page where Ryan says he is moving on |
| 12:45 | <zcorpan> | anne-m: but less overhead |
| 12:45 | <anne-m> | olliej, just enlarge the <canvas> :) |
| 12:46 | <rubys> | http://theryanking.com/entries/2008/12/01/moving-on-from-inursite/ |
| 12:46 | anne-m | thought we were looking for something simple to toy around with |
| 12:47 | <hsivonen> | comparing the trees makes sense |
| 12:47 | <hsivonen> | with SAX Tree, you'd even get error locations in source |
| 12:47 | <hsivonen> | it would totally break all the careful streamability of V.nu |
| 12:48 | <rubys> | Start two threads, read alternatlely from both, abort on first error? |
| 12:48 | <hsivonen> | a more 'right' way would be moving from the java.io model to a browser-like IO model |
| 12:49 | <hsivonen> | but I don't have a Java XML parser that supported that IO model |
| 12:50 | <rubys> | The way I would approach it is to smart small, perhaps only checking for missing quotes since that seems to be an emotional issue for reasons that escape me, and slowly add a new feature every week or three. Over time, the service would be very helpful for advocacy. |
| 12:52 | <zcorpan> | hsivonen: hmm i wonder why i don't get a fatal error for <html a>x<html b> in html-to-xhtml |
| 12:52 | <hsivonen> | rubys: Ryan's blog post adds more evidence to the pile of evidence of no one having succeeded in making a business of monetizing an online validator by making the users pay |
| 12:52 | <rubys> | I would be willing to help, either in the definition, the coding, or the testing. |
| 12:53 | <hsivonen> | zcorpan: it is using the tree mode |
| 12:54 | <zcorpan> | hsivonen: i get a fatal error for <html>x<html b> |
| 12:54 | <hsivonen> | rubys: I think it would make sense to try out the tree comparison idea first as a separate offline tools that uses the rewindable input stream form nu.validator.htmlparser.rewindable |
| 12:55 | <hsivonen> | rubys: having both parsers build a SAX Tree tree with locator data and comparing the trees |
| 12:55 | <hsivonen> | zcorpan: seems like you've found a bug |
| 12:56 | <rubys> | My biggest concern about tree comparison is error recovery. |
| 12:56 | <hsivonen> | rubys: you could only report the first difference |
| 12:57 | <zcorpan> | hsivonen: want me to file a bug? |
| 12:57 | <hsivonen> | zcorpan: that would help, yes |
| 12:57 | <rubys> | I think I suggested that. Useful for true polygots (a very small set of people), and wouldn't have the quite the same advocacy/educational value. |
| 12:59 | <hsivonen> | aargh. Gecko takes the frame name from a different attribute in XHTML and HTML |
| 13:02 | <hsivonen> | rubys: I guess I disagree about polyglotness as something that people should be advocated/educated to do |
| 13:03 | <rubys> | it's actually the other way around... educate them on how hard it really is |
| 13:03 | <hsivonen> | oh |
| 13:03 | <rubys> | let them come to the realization that it isn't worth it. |
| 13:03 | <rubys> | What I do is hard. You don't see me recommending it to others. |
| 13:04 | <rubys> | My experience differs from this statement: http://wiki.whatwg.org/wiki/FAQ#Will_.28X.29HTML_5_finally_put_an_end_to_the_XHTML_as_text.2Fhtml_debate.3F |
| 13:05 | <rubys> | My hope is that an online tool would help point out that what people believe is XHTML really doesn't address their percieved (and largely non-existent) use case. |
| 13:06 | <Philip`> | rubys: Some evidence suggests education won't work |
| 13:06 | <Philip`> | rubys: http://hixie.ch/advocacy/xhtml uses "<!--//--><![CDATA[//><!--" as an example of how "ridiculously complicated" it is to write inline scripts in polyglot documents |
| 13:06 | <hsivonen> | rubys: there's a risk that if I add a near-impossible validation target, some people will start advocating it as the most righteous target since it is the harders |
| 13:06 | <hsivonen> | t |
| 13:06 | <hsivonen> | hardest |
| 13:06 | <Philip`> | rubys: but I see about 0.5% of sites uses that ridiculous string |
| 13:06 | <Philip`> | *using |
| 13:07 | <rubys> | My "HTML Evolution" is a polyglot document that even uses < is inline CSS without that ridiculous string |
| 13:07 | <rubys> | s/is/in/ |
| 13:07 | <zcorpan> | rubys: you mean > ? |
| 13:07 | <Philip`> | (which is far more than the 0.01% or whatever tiny number of pages it is that actually use application/xhtml+xml from the same sample) |
| 13:07 | <rubys> | yup. s/</>/ |
| 13:08 | <hsivonen> | can someone point me to a definition of window.frames.authorspecifiedframename ? |
| 13:09 | <zcorpan> | hsivonen: isn't that [[Get]] on window? |
| 13:09 | <Philip`> | rubys: The point is that attempting to educate people about how hard it is to do a certain thing correctly will just result in them doing it the hard way that you showed in your examples, and they'll miss the point that they shouldn't be needing to do it at all |
| 13:10 | <hsivonen> | zcorpan: I can't find [[Get]] in the spec. what am I missing? |
| 13:10 | <zcorpan> | http://www.whatwg.org/specs/web-apps/current-work/#dom-window-nameditem |
| 13:10 | <rubys> | looking at the amount of broken HTML pages, it seems to me that a lot more people are lazy than virtuous. |
| 13:11 | <rubys> | Thought process seems to be: "oh, it's hard?" & "it doesn't really matter" => "I don't really need to worry about it" |
| 13:11 | <jgraham> | I think rubys may have more of a point for something complex like a whole document rather than something "simple" like a silly string |
| 13:11 | <jgraham> | Note that few people aim for the strict doctypes |
| 13:11 | <Philip`> | I can imagine that people think "I want to use XHTML because I heard it's good", then they read "if you really want to use XHTML then you'd have to do this ridiculously complicated thing", so they think "okay then, I'll copy-and-paste that ridiculously complicated thing and it'll be fine" |
| 13:12 | <zcorpan> | hsivonen: i think that's an area where gecko does different things in quirks mode |
| 13:12 | <hsivonen> | zcorpan: thanks. It seems that it will be non-trivial to make Gecko comply |
| 13:12 | <hsivonen> | zcorpan: it sure does things differently for XHTML and HTML |
| 13:12 | <Philip`> | (instead of thinking "maybe I shouldn't use XHTML") |
| 13:12 | <hsivonen> | I don't like these gratuitous XHTML vs. HTML differences in the APIs |
| 13:13 | <hsivonen> | rubys: or "oh, it's hard?" => "must be important" => "tell others to do it" |
| 13:13 | <rubys> | "maybe I shouldn't use XHTML"? The education I want to do is to help people who think they are using XHTML that they really aren't. |
| 13:14 | <hsivonen> | rubys: then "J. Random Guru says it's important" => "let's do it" |
| 13:14 | <rubys> | s/J./Z./ :-) |
| 13:14 | <jgraham> | It is generally hard to educate people that fashionable things are bad for them |
| 13:14 | hsivonen | is actively trying to make it easier to write JavaScript that work for both HTML and XHTML |
| 13:15 | <hsivonen> | rubys: :-) |
| 13:16 | <Philip`> | jgraham: Fortunately HTML5 seems to be becoming fashionable :-) |
| 13:18 | <jgraham> | Philip`: Is XHTML becoming unfashionable? |
| 13:18 | <jgraham> | Then we don't have to try and advocate people to not use it? |
| 13:19 | <rubys> | faux-XHTML is still very fashionable |
| 13:19 | <jgraham> | s/XHTML/"XHTML"/ |
| 13:19 | <jgraham> | s/?// |
| 13:19 | rubys | notes that s/?// will delete the first character in the line |
| 13:20 | <zcorpan> | Hixie: "window[name] Returns the indicated child browsing context." - doesn't match the impl requirements |
| 13:20 | rubys | retracts that... it actually is a syntax error |
| 13:20 | Philip` | notes that s/?// will actually give a Quantifier follows nothing in regex; marked by <-- HERE in m/? <-- HERE / |
| 13:21 | <Philip`> | But in IRC we use magical regexps, not Perl-compatible regexps, and so it's perfectly valid to write s//missing word/ etc |
| 13:22 | <zcorpan> | s/what i wrote/what i meant/ |
| 13:22 | <rubys> | s/.*?//g |
| 13:23 | <hsivonen> | Hixie: In Gecko, having a name attribute on a frame hides the id attribute value for the purpose of frame naming for access on the window object |
| 13:23 | <jgraham> | Magic regexps are good. They are much better than real regexps |
| 13:25 | <jgraham> | Real regexps tend to end up looking like http://code.google.com/p/html5lib/source/browse/trunk/python/src/html5lib/sanitizer.py#195 |
| 13:25 | <jgraham> | Which is just bad |
| 13:26 | <rubys> | you do realize that you are talking to the author of that line aren't you? :-) |
| 13:26 | <hsivonen> | speaking of regexps, I've considered generating an ahead of compile time automaton for recornizing quirky public ids |
| 13:27 | <Philip`> | rubys: Did you write the next line too? :-) |
| 13:27 | <hsivonen> | *recognizing |
| 13:27 | <Philip`> | rubys: Particularly the one that caused the DOS vulnerability in http://code.google.com/p/html5lib/issues/detail?id=83 ? ;-) |
| 13:28 | <jgraham> | rubys: Yes :p |
| 13:29 | <rubys> | Philip`: yes |
| 13:30 | <jgraham> | rubys: In truth I would like to avoid using regexps there at all. But I don't know how to do that without writing a CSS parser |
| 13:30 | <jgraham> | Because I am not clever enough to understand complex regexps and therefore I don't like relying on them for security |
| 13:34 | <hsivonen> | Lachy: the wiki page does say: White space characters in attribute values are normalized to spaces in XHTML. |
| 13:34 | <anne-m> | rubys, Hixie's long compat line for <style> is if you want compat with pre-<style> UAs as well |
| 13:35 | <anne-m> | rubys, if you just care about HTML4/XHTML1 the story is simpler as the article indicates |
| 13:35 | <Philip`> | Everyone wants compat with Netscape 2 |
| 13:37 | <zcorpan> | people just copy the sample markup without understanding the prose |
| 13:39 | Philip` | thinks that's why Lachy's guide shouldn't mix HTML and XHTML examples, because people will just copy the markup without understanding the difference, and it will result in more inconsistency and the world will become a worse place |
| 13:41 | jgraham | is hard to persuade that bad HTML is a significant world problem |
| 13:42 | <Philip`> | I didn't mean to say it was a significant problem, just that it was a non-zero problem :-) |
| 13:43 | <jgraham> | Well some peole seem to behave as if it were a real problem |
| 13:43 | <hsivonen> | rubys: the reason why it's safe to teach people to escape > as > is to avoid the ]]> special case in XML http://hsivonen.iki.fi/test/moz/square-square-gt.xml |
| 13:43 | <jgraham> | s/real/significant/ |
| 13:43 | <hsivonen> | ]]> is something that would be annoying to add tokenizer complexity for |
| 13:43 | <Philip`> | If ten million HTML authors each waste an hour fighting with bad markup, that's only the equivalent of sixteen human lives, which isn't enough to really worry about |
| 13:45 | hsivonen | wonders what Real Problem the ]]> rule solves |
| 13:45 | <jgraham> | Philip`: People worrying about their own bad markup is a different problem though |
| 13:45 | <Philip`> | hsivonen: I thought it was for compatibility for existing SGML tools, maybe? |
| 13:46 | <zcorpan> | hsivonen: i think ]]> in xml is only an error because it was an error in sgml |
| 13:46 | <zcorpan> | hsivonen: why it was an error in sgml i don't know |
| 13:46 | <hsivonen> | yay for SGML |
| 13:46 | <zcorpan> | hsivonen: it's allowed in attribute values :) |
| 13:47 | <hsivonen> | great |
| 13:47 | <zcorpan> | and in pi data, iirc |
| 13:49 | <hsivonen> | how does Safari deal with pre-HTML5 XPath expressions from JS matched against an HTML DOM? |
| 13:49 | <Philip`> | The rule against < in attribute values seems weird - http://www.xml.com/axml/notes/NoLTinAtt.html says "The rule in XML is simple: when you're reading text, and you hit a <, then that's a markup delimiter. Not just sometimes, always." which seems clearly false because of <!--<--> and <![CDATA[<]]> and <?x >?> |
| 13:49 | <Lachy> | Philip`, I strongly believe that HTML and XHTML need to be tought together so that readers can learn the differences |
| 13:50 | hsivonen | just saw "XPath expressions targeted at pre-HTML5 browsers need to use the XHTML namespace for XHTML and null for HTML. (HTML5 browsers would use the XHTML namespace even in HTML.)" on the wiki |
| 13:50 | <Philip`> | (Uh, I mean <?x <?>) |
| 13:51 | <zcorpan> | Philip`: yeah, it's a weird rule and it wasn't inherited from sgml :( |
| 13:51 | <rubys> | hsivonen: my experience is that > appears in inline CSS way more often than the sequence ]]> appears |
| 13:52 | <hsivonen> | rubys: I just read the HTML vs. XHTML wiki page. I think implementing proper polyglot checking just to make a point to the SVG WG is not a nice prospect |
| 13:52 | <hsivonen> | rubys: however, it might be worthwhile to make a non-validating parsers-only command line tool for making the point |
| 13:53 | <rubys> | what would the command line tool do? |
| 13:53 | <hsivonen> | rubys: it would parse as XML and as HTML into two SAX Tree trees, compare them and report the line and col of the first difference |
| 13:54 | <rubys> | I don't think that would make any point. |
| 13:54 | <rubys> | or rather, I fail to see what point that would make |
| 13:54 | <hsivonen> | rubys: a person playing with it could see when the trees differ |
| 13:55 | <rubys> | what person would find that useful? Perhaps me and Jacques. |
| 13:57 | zcorpan | would probably play with it just to find bugs |
| 13:57 | <Philip`> | hsivonen: Rather than comparing SAX Tree trees, wouldn't it be easier to just serialise both as XML and use 'diff'? |
| 13:58 | <Philip`> | (using a deterministic serialiser for both) |
| 13:58 | <hsivonen> | Philip`: you wouldn't get the right line and col |
| 13:58 | <hsivonen> | Philip`: and you'd have a dependency on a tool that Windows users are unlikely to have |
| 13:58 | <Philip`> | hsivonen: But you'd see where in the markup the error occurred |
| 13:58 | <Philip`> | which is more useful than line/column numbers |
| 13:58 | <Philip`> | particularly since Windows users use Notepad which doesn't display line/column numbers :-) |
| 13:59 | <hsivonen> | Philip`: well, that's something you can already do with HTML2XML and XML2XML and diff |
| 14:03 | <Philip`> | hsivonen: Seems like that might already address many of the use cases, then |
| 14:04 | <hsivonen> | yeah. my shell script skills are weak. can that be done with pipes without temporary files? |
| 14:05 | <rubys> | It could easily be done in a language like python or ruby |
| 14:05 | hsivonen | wonders if document.evaluate is used a lot on text/html in the wild |
| 14:05 | <Philip`> | I don't know how to do it in sh without temporary files |
| 14:05 | <rubys> | why limit it to sh? |
| 14:06 | <Philip`> | rubys: Because that would take the fewest lines of code, if it was possible |
| 14:07 | <rubys> | http://docs.python.org/library/popen2.html |
| 14:07 | <Philip`> | Even if you're using a real scripting language, I don't think you can use the diff command with two input streams and no temporary files |
| 14:07 | <rubys> | Philip`: see the link I just pasted |
| 14:07 | <rubys> | short answer: yes you can |
| 14:07 | jgraham | notes that subprocess should be used rather than popen2 |
| 14:08 | <Philip`> | rubys: That provides access to stdin, stdout and stderr; how do you use them to pass two separate inputs to diff? |
| 14:09 | <jgraham> | Philip`: Ho do you have two seperate inputs? |
| 14:09 | <Philip`> | jgraham: The inputs are the first file and the second file |
| 14:10 | <Philip`> | If you don't want to pass physical files to diff, the only alternative is '-', which is stdin, and there's only one stdin, and 'diff - -' isn't going to be very helpful |
| 14:10 | <jgraham> | Right, I don't really understand conceptually how you han strean two different inputs |
| 14:11 | <Philip`> | Conceptually it's the same as having two output streams |
| 14:11 | <Philip`> | (one called stdout and one called stdin) |
| 14:11 | <Philip`> | except that Unix tools normally only support one standard input stream |
| 14:11 | <Philip`> | Oops |
| 14:11 | <rubys> | wierd: diff -q <(sort file1.txt | uniq) <(sort file2.txt | uniq) |
| 14:11 | <Philip`> | s/stdin/stderr/ |
| 14:11 | <jgraham> | Philip`: Let me be more specific. I don't understand how, in Unix, you can have >1 stdin |
| 14:15 | <hsivonen> | :-( I broke XPath |
| 14:16 | <hsivonen> | that one will be more annoying to fix than CSS |
| 14:16 | <hsivonen> | since at least CSS already treated HTML nodes as being in the XHTML namespace |
| 14:16 | <hsivonen> | does anyone happen to know how WebKit fakes its XPath text/html compat? |
| 14:18 | <hsivonen> | does anyone happen to know how WebKit fakes its XPath text/html compat? |
| 14:18 | <Philip`> | rubys: Hmm, diff <(cat foo) <(cat bar) is kind of cheating because it's really just using bash syntax to hide the creation of temporary files (well, named pipes) :-p |
| 14:18 | <rubys> | http://www.informit.com/articles/article.aspx?p=99706&seqNum=15 |
| 14:19 | <rubys> | looks to me like they are cloned file handles, not temporary files |
| 14:19 | <Philip`> | Oops, yes |
| 14:19 | <Philip`> | "named pipes (FIFOs) or the /dev/fd method of naming open files" |
| 14:20 | <Philip`> | Anyway... The sensible solution is to use temporary files :-) |
| 14:22 | <Philip`> | rubys: Is that informit page about something interesting once it's finished taking forever to load? :-) |
| 14:22 | <zcorpan> | Hixie: the Scope section calls out <blink> as something that's out of scope, yet the rendering section defines it |
| 14:22 | <rubys> | loads quickly for me |
| 14:22 | <rubys> | "Newer systems provide a directory named /dev/fd whose entries are files named 0,1, 2, and so on. Opening the file /dev/fd/n is equivalent to duplicating descriptor n (assuming that descriptor n is open)." |
| 14:22 | <Philip`> | rubys: Oh, works for me now |
| 14:23 | <rubys> | [the article is dated 2003) |
| 14:24 | Philip` | was unaware of such things |
| 14:24 | <rubys> | neither was I |
| 14:36 | <Philip`> | Whoops, I missed the significance of the <(...) thing because I was trying to work out how to pipeline stuff into it |
| 14:37 | <Philip`> | but actually you can just do diff <(html2xml foo) <(xml2xml foo) |
| 14:37 | <Philip`> | which is neat |
| 14:40 | jgraham | just realised that <(...) was different to < (...) which should be obvious given it is bash and is crazy to everyone else |
| 14:46 | Philip` | discovers cat <<<"hello world" |
| 14:47 | rubys | can't wait until Philip` discovers "echo" |
| 14:49 | <rubys> | http://intertwingly.net/blog/2007/11/28/Bash-Here |
| 14:52 | <takkaria> | you'd have thought all the people using unix for years would know all the cool features by now |
| 14:53 | <Dashiva> | Maybe if there weren't five million different tools with different features to learn, none of them properly documented :) |
| 14:53 | <takkaria> | yeah |
| 14:53 | <Philip`> | "man bash" looks like proper documentation to me :-) |
| 14:53 | <Dashiva> | Looks can be deceiving |
| 14:54 | <Philip`> | I've never bothered learning anything advanced in bash because it's easier to just switch to Perl |
| 14:54 | <Philip`> | (or at least it's easier for me) |
| 14:55 | <Philip`> | It took me years to remember how to do for loops over files in bash |
| 14:55 | <rubys> | yet another language where you continue to learn new cool things that have been around for years |
| 14:55 | <Dashiva> | Should just make a shell that uses a decent language ;) |
| 14:55 | <jgraham> | It's like "if all you hace is a hammer everything looks like a nail". Except s/hammer/hammer and sixteen types of screwdriver and a whole socket set/ |
| 14:55 | <Philip`> | I think I've mostly run out of cool new Perl features to discover by accident :-( |
| 14:56 | <Dashiva> | Don't worry, you still have misfeatures and bugs left to go |
| 14:56 | jgraham | still forgets how to do for loops over files |
| 14:56 | <rubys> | for i in *; do echo $i; done |
| 14:56 | <Philip`> | Dashiva: I've never found many of them either :-) |
| 14:56 | <jgraham> | something like for fn in `ls` do; something done |
| 14:56 | <jgraham> | right? |
| 14:56 | <Philip`> | perl -le'print for <*>' is easier for me to remember |
| 14:57 | <Philip`> | jgraham: Your semicolons are in the wrong places, I think |
| 14:57 | <jgraham> | Yeah, that's the bit I get wrong :) |
| 14:57 | <jgraham> | If there is some logic to it I don't understand |
| 14:57 | <Philip`> | Logic? |
| 14:57 | <Philip`> | That's a foreign concept |
| 14:58 | <rubys> | hint: is 'do' a file name or a reserved word? |
| 14:58 | <jgraham> | The bash maintainers would fit right in here |
| 14:58 | <Philip`> | Is ';' a file name or a reserved word? |
| 14:58 | <Dashiva> | Isn't ; just a newline equivalent? |
| 14:59 | <Dashiva> | Not that bash newline handling is any less insane |
| 14:59 | <Philip`> | I haven't a clue |
| 15:00 | <jgraham> | rubys: I need a bigger hint |
| 15:01 | <Philip`> | jgraham: You can write "for i in 1 2 3; do echo $i; done" - it just takes a space-separated list of tokens |
| 15:01 | <Philip`> | and if you use wildcards then they just expand into the list of tokens |
| 15:02 | <jgraham> | Ah, I knew that. So the point is that the first semicolon delimits the set of tokens |
| 15:02 | <jgraham> | and the second one... |
| 15:03 | <jgraham> | I guess is needed because "done" colud be something else? |
| 15:03 | <jgraham> | Or maybe not |
| 15:03 | <Philip`> | (...and one of the tokens could be the filename 'do', and confusion would ensue) |
| 15:04 | <Philip`> | The 'done' could be an argument to the previous command |
| 15:04 | <Philip`> | for i in do; do echo $i done; done |
| 15:05 | <jgraham> | It may be showing that I have only ever written one serious shell script (more than a few lines) |
| 15:07 | <rubys> | between the do and done you can have as many statements as you want |
| 15:07 | <rubys> | each statement ends with a semicolon |
| 15:08 | <rubys> | for i in * |
| 15:08 | <rubys> | do |
| 15:08 | <rubys> | echo $i |
| 15:08 | <rubys> | done |
| 15:08 | <Philip`> | A semicolon or a newline? |
| 15:09 | <rubys> | yes |
| 15:57 | <Philip`> | http://html5.validator.nu/?doc=http://services.philip.html5.org/html-to-xhtml/ |
| 15:57 | <Philip`> | Hmm, why isn't <textarea oninput> allowed? |
| 16:00 | jgraham | wonders wwhy he has such a low success rate in posting to es-discuss |
| 16:00 | <hsivonen> | Philip`: spec bug or validator bug? |
| 16:04 | <Philip`> | hsivonen: Don't know |
| 16:04 | <Philip`> | hsivonen: The spec is too confusing for me to know whether oninput is meant to exist |
| 16:04 | <Philip`> | (It doesn't talk about oninput anywhere, but does talk about input events in various places) |
| 16:36 | <Philip`> | Hmm, it looks like http://ma.tt/2009/03/downloadable-web-fonts/ linked to me and gave me a thousand visitors |
| 16:38 | <Dashiva> | You crashed my opera :( |
| 16:38 | <Philip`> | Oh |
| 16:38 | <Philip`> | http://bugs.opera.com/ ;-) |
| 16:38 | <Dashiva> | It was an outdated build anyhow. Good chance to upgrade. |
| 17:06 | <jgraham> | Philip`: You try to help people and all you get is grief |
| 17:12 | <Philip`> | Hmm, my html-to-xhtml page is almost accepted by the validator, so I'm not sure why Dr Hoffmann thinks it's "completely corrupted" and "nonsense" |
| 17:12 | <gsnedders> | What page? |
| 17:12 | <Philip`> | http://services.philip.html5.org/html-to-xhtml/ |
| 17:12 | <gsnedders> | That's nonsense. |
| 17:13 | <Philip`> | If someone is going to complain about the page, they should complain about how it refers to itself as "http://whatever-this-site-is-called/html-to-xhtml/"; |
| 17:13 | <Philip`> | which is because it didn't have a domain name at all until earlier today, and I didn't bother editing the HTML |
| 17:13 | <Dashiva> | hehe |
| 17:15 | <jgraham> | There is always something suspicious about people to refer to themselves using their academic qualifications |
| 17:15 | <Dashiva> | I'm a bit confused here. Isn't the main use for svg in html handcoding? |
| 17:16 | <jgraham> | And about people who refer to adademic qualifications when they are quoting people |
| 17:16 | <Dashiva> | If you use a tool, wouldn't you have a separate file to begin with? |
| 17:19 | <Philip`> | Indeed, Dr Graham |
| 17:21 | jgraham | notes he has not actually graduated yet |
| 17:21 | <Philip`> | Dashiva: The problem is that you're ignoring the mythical authoring tools that will let you edit SVG that's embedded inside HTML files |
| 17:21 | gsnedders | cracks his whip |
| 17:21 | <gsnedders> | Graduate, bitch! |
| 17:23 | jgraham | isn't sure he likes being gsnedders bitch |
| 17:23 | <Dashiva> | Philip`: I suppose so. I don't see what they offer over existing standalone tools though. |
| 17:24 | <Philip`> | Dashiva: They offer supporting evidence for arguments which assume they exist |
| 17:24 | <jgraham> | Dashiva: You're not an Amaya user? Heretic! |
| 17:24 | <Dashiva> | It's my sense of logic, it just won't let go :( |
| 17:25 | <jgraham> | Also I saw a presentaion once about another similar tool, so I think they exist for a literal value of exist |
| 17:26 | <Dashiva> | I am convinced some people implement things just to confound multiple choice exams |
| 17:26 | <Dashiva> | "SMTP uses TCP/IP" --- "No, I know for a fact you can run SMTP using monkeys and bongo drums" |
| 17:29 | <Philip`> | "What colour are elephants? (a) Grey; (b) Pink; (c) Fluorescent orange so they don't lose each other in the dark" -- "They're definitely pink" |
| 17:29 | <Philip`> | I blame http://news.bbc.co.uk/1/hi/sci/tech/7951331.stm |
| 17:55 | <scherkus> | quick question: are there specs for applying CSS styles to form elements? |
| 17:57 | <jgraham> | scherkus: AFAIK, no |
| 17:57 | <jgraham> | But IANACE |
| 17:57 | <scherkus> | help me out here.. what's CE stand for :) |
| 17:58 | <jgraham> | CSS Expert :) |
| 17:58 | <scherkus> | gotcha :D |
| 17:58 | <scherkus> | I attended a panel at SXSW last week and someone had a presentation at all the inconsistencies |
| 17:58 | <gsnedders> | jgraham: Why not? |
| 17:59 | <scherkus> | its even broken for the same browser on different platforms |
| 17:59 | <jgraham> | scherkus: I believe they are considered replaced elements from a CSS-theory standpoint |
| 17:59 | <jgraham> | But of course that is not true in practice |
| 18:00 | <jgraham> | (the idea was that users would prefer native-looking widgets on all web pages. This turns out to not quite be true) |
| 18:00 | <scherkus> | well for example, safari 4 beta applies "width" differently |
| 18:01 | <scherkus> | mac: http://www.flickr.com/photos/teleject/3353399474/ (left aligned) |
| 18:01 | <scherkus> | win: http://www.flickr.com/photos/teleject/3352574769/ (centered) |
| 18:01 | <scherkus> | however this was a panel by and for web designers |
| 18:01 | <scherkus> | people who typically style form elements heavily |
| 18:03 | jgraham | notes that HTML has error correction and seems to be doing OK' |
| 18:04 | <jgraham> | scherkus: I think it is a known problem. But it is hard to solve |
| 18:04 | Philip` | supposes people's definitions of "OK" may differ |
| 18:04 | <scherkus> | jgraham: cool.. thanks for the help |
| 18:06 | jgraham | thinks that anyone whos definition of OK excludes (probably) the world'smost popular document format needs a new definition of "doing OK" |
| 18:07 | <Philip`> | "common", not "popular" |
| 18:08 | <jgraham> | pedant |
| 18:08 | <jgraham> | Alhough I think that probably depends on the definition of "popular" |
| 18:08 | <Philip`> | I suppose you think influenza is a good design that should be copied too, since it's been wildly successful at becoming widespread |
| 18:09 | <Dashiva> | It _is_ copied all the time :P |
| 18:10 | <jgraham> | If you are trying to make a virus, that is true |
| 18:10 | <Dashiva> | Not to mention viral marketing |
| 18:10 | <Philip`> | Dashiva: I said "should be", not "is" :-) |
| 18:11 | <Dashiva> | Doesn't vaccine development require copying it a lot? :) |
| 18:12 | <Philip`> | Bah, that's not what I mean :-p |
| 18:14 | <Philip`> | If you're trying to make a virus that kills millions then it may be good to copy the characteristics of existing viruses, and if you're trying to make a new HTML then it may be good to copy the characteristics of HTML, but that's all largely irrelevant if you're trying to make an embeddable vector graphics language or a car or anything else |
| 18:15 | <Philip`> | e.g. maybe your goal is to let advanced users do useful things as easily as possible, rather than making it possible for everybody in the world to publish their own thoughts |
| 18:16 | <Philip`> | so it's not particularly helpful to compare it to HTML |
| 18:16 | <Philip`> | (and can be actively misleading) |
| 18:16 | <Philip`> | or, uh, something like that |
| 18:17 | <Dashiva> | Let's make a compiler that explodes if you make an error |
| 18:22 | <Philip`> | That would be better than some compilers I use, which explode even when I haven't made an error |
| 18:27 | <Philip`> | rubys: s/OATH/OAUTH/ on the meeting minutes? (I would edit it myself but I'm too lazy to register on the wiki so I won't) |
| 18:38 | <rubys> | fixed. Thanks! |
| 19:19 | <jgraham> | rubys1: s/Power/Powder/ ? |
| 19:26 | <rubys1> | jgraham: do you have a link to powder? I don't know what this is, or who the Aaron is that they were referring to. |
| 19:27 | <rubys1> | http://www.w3.org/2007/02/powder_charter |
| 19:29 | <jgraham> | rubys1: Yeah that's what I was thinking of |
| 19:29 | <rubys1> | any idea on aaron's last name? |
| 19:31 | <jgraham> | Nope |
| 19:32 | <jgraham> | All I know is that it is something chaals is involved with |
| 19:37 | <danbri> | powder is an attempt to do the bit of PICS that RDF never handled: writing metadata labels for whole batches of pages who share eg common URI structure |
| 19:38 | <danbri> | "everything on pics.playboy.com is a rude jpeg" etc |
| 19:38 | <danbri> | different sense of pics there, sorry :) |
| 19:45 | <Hixie> | rubys1: i'm guessing aaron is eran? |
| 19:46 | <rubys1> | that makes more sense |
| 19:47 | <rubys1> | Hammer-Lahav presumably? |
| 19:49 | <danbri> | yeah, my guess too |
| 19:49 | <rubys1> | updated |
| 19:58 | <rubys> | hixie: any other errata or significant omissions? |
| 20:59 | gsnedders | sighs |
| 21:00 | <gsnedders> | At lunch today, group of friends concluded that I was the girl of the group. The only girl (biologically speaking, that is) agreed. |
| 21:01 | <Philip`> | Biologicality is overrated |
| 21:02 | <gsnedders> | Also, I'm apparently a lesbian. |
| 21:20 | <olliej> | gsnedders: heheh |
| 21:21 | <gsnedders> | Maybe I really should take up takkaria's suggestion of wearing a corset, kilt, and heels to the ball… |
| 21:22 | <gsnedders> | There again, I did tell my date to the ball of his suggestion, and she said she'd desert me! :'( |
| 21:23 | <takkaria> | being a lesbian isn't so bad |
| 21:25 | <gsnedders> | Knowing how pissed off I am at guys in general at the moment I guess I could pass as a lesbian… |
| 21:30 | <dglazkov> | ping Hixie |
| 21:35 | <Hixie> | dglazkov: here |
| 21:35 | <dglazkov> | I came here to cry about test 48 |
| 21:36 | <dglazkov> | and the definition of "visited" |
| 21:37 | <dglazkov> | Hixie: did you have a specific part of CSS spec in mind that this test tests? |
| 21:43 | <dglazkov> | Hixie: acid3 test 48, obviously |
| 21:43 | Hixie | looks |
| 21:43 | <gsnedders> | Because Hixie has never written any other test suite with 48 tests in it :P |
| 21:44 | <Hixie> | dglazkov: i think the test just checks the definition at http://www.w3.org/TR/css3-selectors/#link |
| 21:46 | <dglazkov> | Hixie: yep. but with arrival of ubiquitous iframes-at-load (ads, tracking, etc.) the definition of "visited" seems to have been somewhat tainted |
| 21:46 | <dglazkov> | in other words, _I_ haven't visited those frames. They visited me |
| 21:47 | <dglazkov> | so both FF and WebKit had to come up with a separate class of history items, visited links |
| 21:48 | <dglazkov> | the items that are technically in your history, just not shown to you as a user |
| 21:48 | <dglazkov> | because you haven't really visited them |
| 21:49 | <dglazkov> | I've been wrangling with this dichotomy and can't come up with a good answer |
| 21:49 | <dglazkov> | other than doing the same for chromium. |
| 21:49 | <dglazkov> | but it just seems kinda wrong |
| 21:49 | <dglazkov> | somehow |
| 21:50 | <Hixie> | sorry, had a phone call |
| 21:50 | Hixie | reads up |
| 21:50 | <Hixie> | yeah, i see the issue |
| 21:50 | <Hixie> | i would just do what they do, to be honest |
| 21:50 | <Hixie> | the problem is that sometimes there are sites that do show ocntent in iframes |
| 21:50 | <Hixie> | and you want those to handle :link :visited too |
| 21:50 | <Dashiva> | "Using your favorite web browser (Firefox by default), visit wp-admin/install.php within the directory into which you just installed WordPress on your web site." |
| 21:50 | <Hixie> | even though the user never sees the url bar |
| 21:50 | <Hixie> | consider e.g. a forum |
| 21:51 | <Hixie> | where each forum page is its own iframed page |
| 21:51 | <Hixie> | you'd want the links to become :visited |
| 21:51 | <Hixie> | even though you'd never want to offer them in the autocomplete, and probably never directly in the history |
| 21:51 | <Hixie> | (though you might offer the frame state in the history if you are some kind of god) |
| 21:52 | <dglazkov> | Hixie: but this sucks |
| 21:52 | <dglazkov> | Web sucks |
| 21:52 | <Hixie> | known bug |
| 21:52 | <dglazkov> | :) |
| 21:52 | <Hixie> | :-) |
| 21:52 | <Hixie> | evolved platforms tend to suck |
| 21:52 | <ojan> | dglazkov: in soviet russia, frames visit you! |
| 21:52 | <Hixie> | they also tend to be successful :-) |
| 21:53 | <dglazkov> | ojan: ain't that the truth |
| 23:52 | <Hixie> | making web workers work with application caches is going to be a huge editorial pain in the ass |
| 23:55 | <doodlewarrior> | why does the header tag require an h1...h6 descendant? |
| 23:55 | <doodlewarrior> | shouldnt <header>Welcome</header> be valid? |
| 23:55 | <Hixie> | because there's no header in that header |
| 23:56 | <doodlewarrior> | it seems a bit crufty |
| 23:56 | <doodlewarrior> | to require an <h1> in a <header> |
| 23:56 | <doodlewarrior> | i could use <h1> instead of <header> |
| 23:56 | <Hixie> | yes |
| 23:56 | <doodlewarrior> | but i dont see why i would need to |
| 23:56 | <Hixie> | you should, in fact |
| 23:56 | <Hixie> | <header> is only useful if you are trying to wrap multiple <h1>s or something together into one header |
| 23:57 | <Hixie> | like <header><h1>The Reality Dysfunction</h1><h2>Space is not the only void</h2></header> |
| 23:57 | <doodlewarrior> | i guess it just seems silly to me |
| 23:57 | <Hixie> | a header and subheader acting as a joint single header for a book |
| 23:57 | <doodlewarrior> | after all the fuss about semantic tags |
| 23:57 | <Hixie> | then just ignore <header> and use <h1> :-) |
| 23:58 | <doodlewarrior> | to have h1 be a header and header be used to group multiple headers |
| 23:58 | <doodlewarrior> | esp. since <footer> is valid without any particular children |
| 23:58 | <Philip`> | Rename it to <headergroup> |
| 23:58 | <doodlewarrior> | or hgroup, for that matter |
| 23:59 | <Niictar> | Can I put <nav> in a <header>? |
| 23:59 | <doodlewarrior> | as long as it has an h1...6, i believe so |