2007-09-01 [17:01:00.0000] it's actually kind of vague [17:01:01.0000] see the RFC [17:01:02.0000] http://www.ietf.org/rfc/rfc2854.txt [17:01:03.0000] "The text/html media type is now defined by W3C Recommendations; [17:01:04.0000] the latest published version is [HTML401]. In addition, [XHTML1] [17:01:05.0000] defines a profile of use of XHTML which is compatible with HTML [17:01:06.0000] 4.01 and which may also be labeled as text/html." [17:02:00.0000] HTML5 would update the definition of the text/html media type to include the HTML serialization of HTML5, but not the XML serialization [17:02:01.0000] I'm not sure whether it could meaningfully remove old allowed serializations from what is legal for the media type [17:02:02.0000] because I don't know how media type registration rules work [17:03:00.0000] othermaciej: I guess that's my point. It sounds like that would be a bureaucratic hassle with no real-world advantages. [17:03:01.0000] it does look like the existing RFC does not allow HTML 3.2 or HTML 2 [17:04:00.0000] othermaciej: ? which bit to you gather that from? [17:04:01.0000] *do you [17:04:02.0000] there's not a whole lot of MUST there [17:05:00.0000] indeed [17:05:01.0000] webben: the "Published specification:" section [17:05:02.0000] it's written sloppily [17:05:03.0000] it's unclear what other specifications, if any, alsocover the media type [17:05:04.0000] I'd have though if they meant to exclude 3.2 they would have said "The text/html media type is now defined by the latest version of the W3C Recommendation for HTML" (or something) [17:06:00.0000] rather than talking about "Recommendations" (plural) [17:06:01.0000] or actually said: "You should not serve 3.2 ..." [17:06:02.0000] in any case it's standard practice for W3C specifications to define their associated MIME type [17:06:03.0000] well, it does mention two recommendations, HTML 4.01 and XHTML 1.0 [17:07:00.0000] /me resists the urge to comment that it's a bad habit of W3C specs to allude to MIME types that they might register someday... [17:07:01.0000] othermaciej: ah, I see how your reading works. [17:08:00.0000] othermaciej: i think it's a bit of a stretch on such vague text though [17:08:01.0000] webben: well, it's pretty clear what this registration allows, but it's totally unclear what, if anything, it disallows [17:09:00.0000] the only actual MUST is related to line breaks [17:09:01.0000] and here's the worst abuse of must in a spec ever: [17:09:02.0000] "User agents executing such [17:09:03.0000] scripts or programs must be extremely careful to insure that [17:09:04.0000] untrusted software is executed in a protected environment." [17:45:00.0000] Hmm, I never knew Thunderbird automatically strips out " (was: ...)" from subjects when replying [18:41:00.0000] hmm. I wonder why people are wasting their time discussing "HTML 4.02" and widely unsupported SGML features. [18:58:00.0000] how many NETs can dance on the head of a pin? [01:45:00.0000] hmm, wonder if we can limit text/html sniffing to "text/html", "text/html; charset=iso-8859-1" and "text/html; charset=ISO-8859-1" so that when you declare utf-8 it's not sniffed [01:46:00.0000] or are there feeds in the wild that are declared as "text/html; charset=utf-8"? [06:45:00.0000] Hixie: "Error loading the folder list: Internal Server Error. Let Hixie know." [06:46:00.0000] oh, now it worked (3rd time) [09:13:00.0000] Hixie: i fail to parse this sentence (in a note in #writing): "It is suggested that newlines be inserted after the DOCTYPE and any comments that aren't in the root element." [09:58:00.0000] zcorpan: "We suggest that you insert newlines after the DOCTYPE and after each comment that isn't in the root element." [10:01:00.0000] aha [10:02:00.0000] i read "it is suggested" as in "we have heard" [10:03:00.0000] but doing so for comments after the root will append LF characters to the html element each time it is parsed and serialized [10:05:00.0000] not a biggie but the rest of the section goes to great lengths to ensure that things round-trip [10:15:00.0000] hmm, they are appended to the body -- not html [10:37:00.0000] hmm, i should remember to take out my own address from the To field when replying to myself :| [10:52:00.0000] not sure i'm fond of the ambigous ampersand thing [10:52:01.0000] it's more straightforward to require it to be escaped 2007-09-02 [19:53:00.0000] hi [20:13:00.0000] hi tibberio [20:42:00.0000] hi Lachy_, I'm asleep, night [05:42:00.0000] I implemented the text/html content sniffing algorithm in JavaScript http://html5.lachy.id.au/content-sniffing/ [05:42:01.0000] also checked it into google code [05:44:00.0000] Lachy: should the [invalid] sequence '' really be text/html? [05:45:00.0000] that's what the algorithm says [05:45:01.0000] in the spec [05:45:02.0000] I wondered about that too, but I figured it would be better to implement exactly what the spec says and then provide feedback about it [05:46:00.0000] wait, why wouldn't be text/html? [05:47:00.0000] if anything, it should always be text/html, since it's not well formed XML and thus can't be a feed [05:47:01.0000] i.e. probably shouldn't be Atom [06:05:00.0000] Lachy: that's a good environment to run testcases [06:05:01.0000] for browsers [06:05:02.0000] yep [06:06:00.0000] it would probably be a good idea to integrate it into the live dom viewer later [06:06:01.0000] similar to http://html5.org/parsing-tests/testrunner.htm and http://simon.html5.org/test/html/serializing/001.htm [06:07:00.0000] though you could use json as the format for the tests [06:08:00.0000] right, that shouldn't be too hard to do [06:08:01.0000] though i wonder if you have to have a server involved to actually test what browsers do [06:09:00.0000] i.e. the relevant sniffing only occurs for resources served over the wire [06:09:01.0000] yeah, but that's not too hard to do [06:09:02.0000] true [06:10:00.0000] I tried a few simple test cases with data: URIs and Firefox didn't sniff them [06:10:01.0000] right, data: are never sniffed [06:10:02.0000] afaik [06:11:00.0000] so we'd probably need to load from the server into an iframe or something [06:11:01.0000] exactly [06:11:02.0000] slightly annoying but certainly doable :) [06:12:00.0000] I think I have a PHP script somewhere that will echo whatever data you pass, using whatever MIME specified, [06:12:01.0000] we could perhaps also use Philip`'s appoach with creating tests [06:13:00.0000] hmm [06:13:01.0000] might be better to just find all the edge cases [06:18:00.0000] http://canvex.lazyilluminati.com/misc/dom-viewer/reflect.cgi?%3Cscript%3Ealert('Hello%20world')%3C/script%3E echoes stuff but it's probably not so useful since it always does text/html [06:19:00.0000] (There's a reflect-xhtml.cgi for XHTML) [06:19:01.0000] I've got one that does the same thing, though I'm going to make it check for a content-type parameter in the query string and use that [06:21:00.0000] http://canvex.lazyilluminati.com/misc/dom-viewer/reflect.c - I don't really fancy modifying that to actually parse the query string at all, so it's probably more sensible to use PHP :-) [06:21:01.0000] the only problem with something like that is that it creates a huge security risk [06:21:02.0000] Lachy: the spec doesn't define text/plain when it lacks angle brackets? [06:21:03.0000] virtuelv, no, why would it? [06:22:00.0000] /me goes back and reads the proposal again [06:22:01.0000] virtuelv_, it's the algorithm that sniffs content delivered as text/html to see if it's a feed. If it's not a known feed format, it's text/html [06:22:02.0000] seel http://www.whatwg.org/specs/web-apps/current-work/multipage/section-content-type-sniffing.html#content-type3 [06:38:00.0000] Philip`, zcorpan, http://html5.lachy.id.au/output [06:41:00.0000] Lachy: you need to do stipslashes() [06:42:00.0000] oh crap, that should be off by default on my server. [06:43:00.0000] hmm. I'll have to add stripslashes for now, since I can't remember how to do it in .htaccess or php.ini or whatever it is [06:43:01.0000] i can't get "" to be echoed [06:44:00.0000] in firefox [06:45:00.0000] and safari [06:46:00.0000] no wait it works in safari [06:47:00.0000] Lachy: you also need an id on the :) [06:47:01.0000] oops :-) [06:51:00.0000] where can I find an Atom test file? [06:52:00.0000] in the rfc :) [06:52:01.0000] never mind, found one [06:53:00.0000] hmm, ie doesn't add attributes to the root node from misplaced tags [06:53:01.0000] it appears that Firefox will only sniff as the result of a GET request, not a POST [06:56:00.0000] Firefox needs to see the Atom xmlns and there needs to be at least a title and link element in it for it to be sniffed as a feed. [06:56:01.0000] http://html5.lachy.id.au/output?type=text%2Fhtml&data=%3Cfeed+xmlns%3D%22http%3A%2F%2Fwww.w3.org%2F2005%2FAtom%22%3E%0D%0A++%3Ctitle%3EExample+Feed%3C%2Ftitle%3E%0D%0A++%3Clink+href%3D%22http%3A%2F%2Fexample.org%2F%22%2F%3E%0D%0A%3C%2Ffeed%3E%0D%0A [06:58:00.0000] doesn't opera sniff for feeds? [06:58:01.0000] it doesn't appear to [06:58:02.0000] maybe it needs more content in it [06:59:00.0000] nope, it doesn't (at least not for Atom) [07:00:00.0000] not for rss 2.0 either [07:01:00.0000] hmm. How do I turn off AddDefaultCharset for just that one file? [07:02:00.0000] I've tried putting it in a directive, but it doesn't seem to to work [07:03:00.0000] AddDefaultCharset Off in the same folder should work (for all files in that folder) [07:03:01.0000] yeah, but it's in the root directory, so I don't want it to do that [07:04:00.0000] dunno [07:04:01.0000] perhaps opera only does sniffing for following s [07:05:00.0000] oh well, I'll just turn it off for everything now and fix it later [07:06:00.0000] maybe, try it [07:06:01.0000] but it's a php file, and AddDefaultCharset doesn't apply to php [07:06:02.0000] php sends its own headers [07:07:00.0000] http://html5.lachy.id.au/output?type=text%2Fhtml&data=%3Clink+rel%3D%22alternate%22+type%3D%22application%2Fatom%2Bxml%22+href%3D%22http%3A%2F%2Fhtml5.lachy.id.au%2Foutput%3Ftype%3Dtext%252Fhtml%26data%3D%253Cfeed%2Bxmlns%253D%2522http%253A%252F%252Fwww.w3.org%252F2005%252FAtom%2522%253E%250D%250A%2B%2B%253Ctitle%253EExample%2BFeed%253C%252Ftitle%253E%250D%250A%2B%2B%253Clink%2Bhref%253D%2522http% [07:07:01.0000] 253A%252F%252Fexample.org%252F%2522%252F%253E%250D%250A%253C%252Ffeed%253E%250D%250A%22%3E%0D%0A%3Cp%3ETest [07:07:02.0000] AddDefaultCharset applies based on the MIME type I think [07:08:00.0000] so all text/* responses get ; charset=... [07:08:01.0000] yeah but php will still send its own headers [07:08:02.0000] php overrides apache [07:09:00.0000] indeed, but it's not. I'm using headers("Content-Type: $type"); in the php, so it should send exactly what's in the text box, but it doesn't [07:09:01.0000] it does afaict [07:10:00.0000] yeah, it does now, cause I turned off AddDefaultCharset already [07:10:01.0000] but if I turn it on again, it won't [07:11:00.0000] can you turn it on again, please? [07:12:00.0000] done [07:12:01.0000] oh, indeed. that's not how my server behaves though [07:13:00.0000] perhaps a different version of php or different settings [07:13:01.0000] which version of Apache does your server run? [07:13:02.0000] 1.3 iirc [07:13:03.0000] I'm using PHP 5, Apache 1.3 I think [07:13:04.0000] php 4 [07:13:05.0000] ok, so then php 5 pays attention to apache settings [07:13:06.0000] I'll try it with PHP4... [07:15:00.0000] hmm, perhaps i was wrong [07:16:00.0000] nope, still does it [07:16:01.0000] I've turn default charset off again [07:30:00.0000] zcorpan, for the HTML status updating, I think it would be easiest if we just use HTTP basic authentication and using a .htpasswd file, which grants access to the confirmation script for Hixie (or whoever else is authorised) [07:32:00.0000] Lachy: ok [07:33:00.0000] we just need to write the script that commits the changes to the DB, and then make a few minor changes to the existing update script and we're done [07:53:00.0000] Lachy: great [07:56:00.0000] Lachy: it's all the magic_quotes settings [07:56:01.0000] gsnedders, thanks, I do have the instructions for it written down somewhere. I just have to find them and set it up [07:57:00.0000] http://uk2.php.net/magic_quotes [07:57:01.0000] (a "feature" that's gone in PHP6) [07:58:00.0000] interpreter/compiler-level settings that change programming language semantics suck big time [07:59:00.0000] yeah, PHP already sucks and fixing PHP by turning off silly "features" just needs to be done [08:03:00.0000] according to the PHP devs, is invalid HTML. [08:04:00.0000] the only way I could get them to fix the bug in strip_tags() which claims to only work with valid HTML anyway was to quote both HTML5, HTML 4.01, and SGML. [08:06:00.0000] PHP devs often don't know HTML very well at all [08:06:01.0000] try to get them to fix the ampersand bug, for instance! [08:07:00.0000] the one where they generate and output "...&PHPSESSID=..." instead of using & [08:08:00.0000] Or just get them to handle multiple selects/checkboxes without hacking the names [08:09:00.0000] yeah, that too [08:09:01.0000] Although that's not html-specific, they mangle it the same way with parsing ini files [08:09:02.0000] Lachy: oh, that's a config option. duh :P [08:10:00.0000] gsnedders, I know, but it should be default! [08:10:01.0000] but backwards compatibility!11!!!!elevnty! [08:10:02.0000] they should have fixed it for PHP5, which isn't backwards compatible anyway [08:10:03.0000] or even PHP6 [08:10:04.0000] yes, they're backwards compatible!11!!1! [08:10:05.0000] It's stupid. [08:11:00.0000] They refuse to fix things in the name of BC, yet totally break it anyway. [08:11:01.0000] yeah, I know. I should just learn python and how to get it running on my server [08:13:00.0000] yeah, I've been meaning to learn python too. [08:14:00.0000] far more general use, too [08:23:00.0000] Python's going to break backward compatibility too, but at least they're intending to fix lots of small mistakes at the same time, which seems good as long as you only do it once a decade [08:40:00.0000] woo hoo! My first python script running on my server :-) http://lachy.id.au/temp/test 2007-09-03 [22:33:00.0000] I added some more features to the content sniffing page. It will now load data from the query string (like the live dom viewer) and shows the browser output in an iframe [22:34:00.0000] http://html5.lachy.id.au/content-sniffing/?%3Cfeed%3Etest2%3C/feed%3E [02:18:00.0000] Philip`: running a copy of the validator.nu software is not hard if your host allows permanantly memory-resident processess and has mod_jk [02:19:00.0000] what happened with the validator.nu outage was that the shared host was taken down for scheduled maintenance one day [02:19:01.0000] then something went wrong and the outage took 2 days instead of planned 5 hours [02:20:00.0000] since the maintenance went long, I couldn't make sure validator.nu was back up before I flew out to Budapest [02:21:00.0000] and then mod_jk was missing [02:22:00.0000] and then giving my password to a random internet cafe computer was unacceptable security-wise and GPRS roaming didn't work in Romania, so it took a while to find working wifi connectivity to get everything back up [02:22:01.0000] I'm going to investigate moving away from the shared host to avoid maintenance breaks when I happen to be traveling [02:23:00.0000] my apoligies to everyone who tried to use the service during the outage [02:23:01.0000] apologies even [03:07:00.0000] Re: accesskey UI: BBEdit/TextWrangler show shortcut keys on dialog buttons when the command key is held down. [03:38:00.0000] I'm going to implement POSTing documents to validator.nu. Is Content-Location on POST a good way to indicate the base URI of the POSTed document? (the RFC leaves Content-Location on POST undefined) [04:44:00.0000] hsivonen: are there any alternatives? :) [04:44:01.0000] why do you need the base uri? [04:51:00.0000] zcorpan: the alternative would be along the lines of X-Content-Location [04:51:01.0000] zcorpan: the base URI is needed for the DTD-loading mode [04:52:00.0000] zcorpan: also, if Hixie really goes down the road of requiring conformance checkers to load images, then the base URI will be needed for HTML5 as well [04:52:01.0000] (although I'm kinda hoping Hixie won't go there :-) [05:04:00.0000] hsivonen: you could opt to not support loading DTDs in such cases [05:08:00.0000] zcorpan: I could. but picking a HTTP header name and sticking its value in a SAX InputSource is so embarrassingly easy that I feel I should do it for completeness [05:11:00.0000] the harder part is checking that the pre-existing code that expects to see sane URIs doesn't break too much when a URI is missing [05:32:00.0000] hi hsivonen, how was your holiday? [05:45:00.0000] Lachy: the holiday was fine [05:45:01.0000] cool :-) [05:47:00.0000] re POST and Content-Location header, I don't understand how you intend it to be used. Are you expecting clients to send the Content-Location header with their POST request? [05:49:00.0000] and is the HTML file supposed to be posted in the body of the request? [05:49:01.0000] Lachy: yes and yes [05:50:00.0000] ok, but regular web browsers wouldn't send that header. Which clients would you expect to do so? [05:50:01.0000] Lachy: blogging systems, command line tools and such non-browser Web service clients [05:51:00.0000] ok [05:51:01.0000] Lachy: I intend to implement back end support for form POST, too [05:52:00.0000] (but I don't have a satisfactory design for form POST front end UI yet) [06:25:00.0000] the omit-alt comments are fun [06:25:01.0000] hey annevk_zeist [06:26:00.0000] alles goed? [06:28:00.0000] ja, op zich wel [06:28:01.0000] beetje dingen opruimen van vakantie enzo [06:28:02.0000] oke [06:29:00.0000] en nog steeds geen baggage uit Madrid... [06:29:01.0000] still? (or have I misunderstood) [06:29:02.0000] yeah, still no luggage [06:30:00.0000] just not been there to get it, or…? [06:30:01.0000] neh, they're messing it up [06:30:02.0000] /me sighs [06:30:03.0000] it's been too long since I've read such languages [06:30:04.0000] supposedly it would go to my parents, but they haven't received anything yet and trying to reach KLM luggage service on the phone is not really possible [06:31:00.0000] ergh. [06:36:00.0000] anything interesting happened? [06:37:00.0000] annevk_zeist: oh, just one or two flamewars on public-html. not really interesting by WG standards, though [06:38:00.0000] Lachy, yt? [06:38:01.0000] gsnedders, heh [06:38:02.0000] yo [06:38:03.0000] annevk_zeist: mainly on the subject of making @alt optional [06:38:04.0000] Lachy, I think it would be nice if the FAQ became a wiki page [06:39:00.0000] ok. If you make the page, I can set up redirection [06:40:00.0000] k, I'll look into it when I have my laptop again [06:40:01.0000] (it's in Utrecht, not lost with the luggage) [06:45:00.0000] /me really needs some sort of hosting on a stable domain for all the browser tests I'm going to write [06:47:00.0000] you can get firstname.html5.org [06:47:01.0000] meh. geoffrey is a horrible name :) [06:48:00.0000] annevk_zeist: gsnedders would be preferable, but I wouldn't complain with geoffrey [06:48:01.0000] I suppose I could break the convention... [06:48:02.0000] /me ponders [06:51:00.0000] /me has got annevk_zeist pondering. is that safe? :P [06:53:00.0000] as long as you're not here :p [06:58:00.0000] I'll get you gsnedders.html5.org later; need my laptop first [06:58:01.0000] annevk_zeist: yeah, having your own computer would be useful; thanks, though :) [07:11:00.0000] Hixie: Re: offline: Isn't if more Web-like for each app view to have its own URI instead of having one top-level URI and a lot of Ajaxy dynamism on the one "page"? [07:49:00.0000] ah, 2033 new e-mails [08:31:00.0000] hello [08:31:01.0000] everyone having a good labor day? [10:22:00.0000] labour day> [10:23:00.0000] s/>/?/ [10:24:00.0000] gsnedders, apparently it's labor day in the US. I think it's a public holiday [12:30:00.0000] hmm, one angry e-mail to KLM and I suddenly get a call from Spain [12:30:01.0000] /me wonders if they're connected [12:43:00.0000] annevk, what's KLM? [12:43:01.0000] btw, this article seems to be nothing more than FUD http://www.gnucitizen.org/blog/i-dont-think-that-you-understand-firefox3-vulnerable-by-design [12:46:00.0000] the port scanning script is the only one that seems reasonable, but it depends on whether or not readystate ever fires for 3 LOADING when the request hasn't been allowed [12:46:01.0000] airline company [12:47:00.0000] it doesn't go to readyState 3 obviously (that would also allow more than just port scanning) [12:47:01.0000] yeah, I didn't think it would [12:48:00.0000] but restricting browsers from setting the Content-Access-Control header in the request seems reasonable, if over the top [12:49:00.0000] virtuelv_, why would it matter if it was sent as a request header? [12:52:00.0000] Lachy: You're right. It shouldn't [12:52:01.0000] The TRACE method is used to invoke a remote, application-layer loop- [12:52:02.0000] back of the request message. The final recipient of the request [12:52:03.0000] SHOULD reflect the message received back to the client as the [12:52:04.0000] entity-body of a 200 (OK) response. [12:53:00.0000] but, understanding that correctly, offsite TRACE would mostly be impossible anyway [12:53:01.0000] or rather, not [12:53:02.0000] the point being: [12:54:00.0000] (You can theoretically get content back, but the only thing you could possibly steal were the cookies sent to the server in the first place) [12:55:00.0000] IIRC, cookies aren't sent with XHR requests to other domains, are they? [12:58:00.0000] There is exactly one mention: "User agents which implement this specification should take care not to expose other trusted data (cookies, HTTP header data) inappropriately." [12:58:01.0000] whatever that means [12:59:00.0000] if the meaning is that cookies are not to be sent, then I'd say the spec ought to state this explicitly [12:59:01.0000] hmm. I'm not sure, but it would seem crazy if browsers did, since it would open up a huge security risk [12:59:02.0000] headers and content body and such is still a big vague atm I'm afraid [13:05:00.0000] now everyone can fix bugs: http://wiki.whatwg.org/wiki/FAQ [13:06:00.0000] Lachy, you were ok with this, right? [13:07:00.0000] yeah [13:09:00.0000] I'll set up redirection from /faq/ on the blog later [13:10:00.0000] k [13:41:00.0000] jgraham_, found a bug in your tool: http://wordsandpictures.dyndns.org/cgi-bin/tables/table_inspector.py?uri=http%3A%2F%2Fannevankesteren.nl%2F2007%2F09%2Ftmb-overview&algorithm=html4&scope=1&headers=1 [13:41:01.0000] jgraham_, it doesn't respect [13:41:02.0000] jgraham_, or maybe a bug in scope="rowgroup", dunno [13:42:00.0000] jgraham_, it also has encoding issues [13:50:00.0000] It's fascinating how RB claims to be implemented in WebKit and Trident [13:50:01.0000] to see* [14:09:00.0000] I think the section should give some advice on smileys [14:12:00.0000] :-)

? [14:12:01.0000] for instance [14:12:02.0000] /me always wonders whether ASCII art is appropriate there [14:13:00.0000] annevk: What do you expect the behaviour to be? [14:13:01.0000] /me doesn't want to fix the "wrong" problem [14:14:00.0000] jgraham_, I expect stuff like "Day 1" not be repeated until the end of the table, it solely applies to the it's in [14:14:01.0000] annevk: i think it's appropriate. it's what you would use if you didn't use an image, and it's quite common in "plain text" [14:17:00.0000] Oh, I see what you mean [14:25:00.0000] good :) [14:25:01.0000] I was actually wondering if the algorithm could be made even more nice for the lazy author so I could omit scope= there too [14:26:00.0000] scope=rowgroup is implicit if the first row of a rowgroup contains a single header cell [14:28:00.0000] zcorpan, well, there are Unicode smileys... [14:29:00.0000] do homework, or write test cases? hmmm… [14:29:01.0000] ☹☺☻ [14:36:00.0000] annevk: yeah, but they aren't very common in "plain text", probably wasn't what the author typed in, and there are only 3 of them [14:37:00.0000] forums often have dozens of smileys [14:38:00.0000] using exactly what was typed in as alt has the advantage that it round trips copy-and-paste [14:48:00.0000] hmm, specialcasing is ugly [14:48:01.0000] doable, but ugly [14:48:02.0000] indeed [14:48:03.0000] ie already specialcases it, it allows hidden inputs in head [14:48:04.0000] oh [14:49:00.0000] doesn't make it less ugly but proves that it is doable :) [14:49:01.0000] I was thinking that maybe form.elements could be filled up during parsing [14:49:02.0000] yeah, i was pondering about alternative approaches too [16:04:00.0000] annevk: I think the heading information for your table is almost right now [16:05:00.0000] The finalk thing is that the Day 2 has "Day 1" as one of its headings (and so on down the page) [16:05:01.0000] This is because these headers aren't assigned any heading information from the scope algorithm [16:06:00.0000] and HTML 4 says "In the absence of header information from either the scope or headers attribute, user agents may construct header information according to the following algorithm" [16:06:01.0000] how would Day 2 get Day 1? Day 1 has an explicit scope... [16:07:00.0000] From the implicit algorithm. [16:07:01.0000] I do get how Day x gets Location, estimated time, etc. although that's not desired [16:07:02.0000] I think it's an incorrect reading of the HTML 4 spec, but I don't know what the right reading is. [16:08:00.0000] I'm not sure why you think there's absense of scope or headers attribute [16:09:00.0000] I suppose I need something like Day x to override the Location, Height, ... headings [16:09:01.0000] Because, at the moment, my implementation assumes that ""In the absence of header information from either the scope or headers attribute" applies to the cell we are trying to assign a heading to [16:09:02.0000] assuming works like that [16:10:00.0000] jgraham_, yeah, so you look up and find a , but that can't be the header of the header because it has a different scope... [16:10:01.0000] in the case of the cell containing "Day 2" there is no information from any scope or headers attribute that assigns heading information to that cell [16:11:00.0000] annevk: But does HTML 4 actually say that? [16:11:01.0000] not sure [16:11:02.0000] it also seems that the headers given in are not applied to the four columns... [16:12:00.0000] My reading is that if a cell has no heading information supplied from headers or scope you then do the row/column search ignoring scope or headers attributes [16:12:01.0000] this is the case in the HTML5 algorithm though [16:12:02.0000] annevk: agreed [16:12:03.0000] HTML5 actually has exactly how I want it [16:12:04.0000] My imp. of the HTML 5 algorithm seems to get that right [16:12:05.0000] including the scope=rowgroup headers not getting the headers [16:13:00.0000] /me wonders if that always makes sense [16:13:01.0000] Probably not. [16:13:02.0000] My conclusion from playing with a few tables on the net is that nothing always makes sense [16:14:00.0000] You could have a header cell for a rowgroup in a row that contained summary information/totals for the rowgroup [16:14:01.0000] I think I've seen tables like that in print. [16:14:02.0000] I'm not sure if that's correct usage though [16:15:00.0000] correct usage of what? [16:15:01.0000] of HTML tables [16:15:02.0000] Why aren't you? [16:16:00.0000] because it doesn't make much sense to me to put summaries in headers [16:16:01.0000] annevk: I don't see why not. A good example would be demographics. [16:16:02.0000] say you have area then pop in the thead [16:17:00.0000] then you have each rowgroup as a continent with total pop [16:17:01.0000] then you have countries in each rowgroup row [16:17:02.0000] with their respective pop [16:18:00.0000] Don't forget that a td can be a header simultaneously with being a data cell. [16:20:00.0000] that's not actually clear from HTML4 [16:20:01.0000] annevk: yes it is. it's stated in the DTD [16:20:02.0000] is there anything that contradicts that? [16:21:00.0000] the prose doesn't really support that comment [16:22:00.0000] annevk: I think trying to read the comments against the prose doesn't make much sense. Specifications aren't meant to be read to be self-contradicting. :) (They might be self-contradicting by accident, but you don't seem to be arguing that's the case here.) [16:24:00.0000] I'm not sure why it doesn't make sense. Comments, notes, examples, etc. are all non-normative. (Although in this specific case nothing much is normative and the whole thing is rather vague.) [16:26:00.0000] annevk: Actually the prose does say: "Note that it's not always possible to make a clean division of cells into headers or data. You should use the TD element for such cells together with the id or scope attributes as appropriate." Which seems a relatively clear restatement of the principle. [16:26:01.0000] annevk: Yes. But they are all intended to help explain the spec, not contradict it. [16:27:00.0000] therefore trying to read them as a source of contradictions, like two alternative sources for the same thing, is strange [16:27:01.0000] (that quote is from 11.4.1) [16:28:00.0000] annevk: also note the example intended to illustrate the scope attribute which features td with the scope="row" [16:28:01.0000] followed by the text in the prose: "Although the first cell in each row contains data, not header information, the scope attribute makes the data cell behave like a row header cell." [16:29:00.0000] /me is unsure what the benefit is of trying to figure out what the intent of html4 is [16:29:01.0000] isn't it better to focus on figuring out an algorithm for html5 that works with existing tables on the web? [16:29:02.0000] /me is not going to add scope="row" to all his s [16:29:03.0000] euh, s [16:29:04.0000] zcorpan: I'm interested insofar as it it necessary to show people that vauge = bad + probably illogical [16:30:00.0000] (illogical in non-trivial cases, that is) [16:30:01.0000] jgraham_: ok [16:31:00.0000] it's also good to know what use cases HTML4 catered for [16:31:01.0000] annevk: Actually looking through the prose throughout supports that comment. [16:31:02.0000] I hadn't realised just how supported it was. [16:32:00.0000] ? [16:33:00.0000] I can't see how one could construct anything contradicting the comment. [16:33:01.0000] oh [16:34:00.0000] zcorpan: Depends on whether "figuring out an algorithm ... that works with existing tables" without consulting the spec that data table creators used is a realistic task. I suspect it isn't. [16:36:00.0000] (Although I'd certainly say it would be also worth looking at representations of the spec in major unlearning resources like w3schools.) [16:37:00.0000] I'm not sure why it's unrealistic to study actual tables out there [16:39:00.0000] annevk: I didn't say it was. I think that's a very important part of the process too. [16:40:00.0000] annevk: What I am saying is it's unrealistic to /only/ do that. [16:42:00.0000] (because the spec is a guide to well-authored tables that aren't sampled and unlearning resources are a guide to badly authored ones) [16:43:00.0000] /me shrugs and goes to bed [16:43:01.0000] speaking of w3schools... http://w3schools.invisionzone.com/index.php?showtopic=15018 *grin* [16:50:00.0000] I'm hoping that doesn't count as a legitimate use case for [16:50:01.0000] (It's good to see the author has stopped using - now they just need to go a little further and stop using too...) [16:57:00.0000] Given the way how HTML5 defines that browsers should present an image (inside an HTML document) you can theoretically style that page using the Link: HTTP header... [16:57:01.0000] /me adds it to a list of silly things to test 2007-09-04 [17:01:00.0000] ", and to the #mrt crew, the #mrt.no crew, and the cabal for their ideas and support." [17:01:01.0000] lol, I'm pretty sure that read #whatwg before... [20:35:00.0000] jgraham, is the source code for your table inspector available yet, or are you planning to release it soon? [01:12:00.0000]