| 00:23 | <sicking> | Hixie, so what are the features of websockets that you are concerned about changing? |
| 00:24 | <Hixie> | the handshake |
| 00:24 | <Hixie> | specifically the processing of websocket-origin and websocket-location |
| 04:13 | <othermaciej> | Hixie: are you about? |
| 04:13 | <Hixie> | vaguely |
| 04:15 | <othermaciej> | Hixie: there's a new (highres and otherwise improved) timer API being discussed on webkit-dev - should I propose it for HTML5 or as a standalone spec in Web Apps WG? |
| 04:16 | <Hixie> | webapps |
| 04:16 | <Hixie> | i have an editor lined up for it |
| 04:16 | <takkaria> | wow |
| 04:17 | <othermaciej> | I might even be able to edit or co-edit myself. |
| 04:17 | <othermaciej> | it seems pretty simple |
| 04:17 | <othermaciej> | only complexity is that it kind of conceptually ties into HTML5's event loop |
| 04:18 | <Hixie> | afk, bbiab |
| 04:19 | <othermaciej> | later |
| 05:35 | <Hixie> | http://www.w3.org/2001/tag/2008/09/23-minutes.html -- the tag appears to be so far removed from what we're trying to do here that our meeting at the tpac ought to be quite interesting |
| 06:59 | <othermaciej> | Hixie: reading that belatedly... I see what you mean |
| 07:03 | <othermaciej> | how does one run for TAG? |
| 07:03 | <othermaciej> | I wonder how much I would hate that |
| 07:06 | <Hixie> | othermaciej: an AC rep has to nominate you and you have to be picked |
| 07:07 | <othermaciej> | it was mentioned that there are "elected" members in those minutes |
| 07:07 | <Hixie> | someone offered to nominate me once |
| 07:07 | <Hixie> | i told them in no uncertain terms not to bother |
| 07:07 | <othermaciej> | 5 selected by the AC |
| 07:07 | <othermaciej> | er |
| 07:08 | <othermaciej> | elected |
| 07:08 | <othermaciej> | 3 nominated by director |
| 07:08 | <othermaciej> | I'm wondering if having someone there with a real-world relationship to the web would help |
| 07:08 | <othermaciej> | probably not worth it |
| 07:09 | <Hixie> | you have to ask: what is the purpose of the tag |
| 07:09 | <Hixie> | what effect can it have on the web |
| 07:09 | <othermaciej> | "stewardship of the Web architecture" |
| 07:09 | <Hixie> | o_O |
| 07:09 | <othermaciej> | I guess participating in a group where my goal would be preventing damage rather than contributing something positive would be a waste of time |
| 07:12 | <othermaciej> | reading the process document makes me wince |
| 07:12 | <Hixie> | it's not clear to me that the TAG has ever caused harm, other than a few cases where people for some reason felt compelled to somehow change the tag's mind (usually fruitlessly, and later just ignoring them instead) |
| 07:12 | <othermaciej> | so much silly process and bureaucracy, which is the ignored when inconvenient |
| 07:15 | <Hixie> | (and then the harm was just wasting time) |
| 07:17 | <othermaciej> | they rarely have managed to get up the energy to waste major amounts of time |
| 07:20 | <hsivonen> | are those TAG minutes meant to become public? |
| 07:24 | <othermaciej> | I think so... |
| 07:37 | <Hixie> | why would they not be? |
| 08:22 | <hsivonen> | Hixie: they aren't public now. but I guess that's just a review period thing |
| 08:28 | <Hixie> | i wish people would get over this secrecy nonsense |
| 08:29 | <hsivonen> | the people who'd be most offended or advantaged by knowing the secret stuff that might get removed already have access... |
| 08:29 | <Hixie> | i know |
| 08:29 | <Hixie> | that's what boggles my mind |
| 08:34 | <hsivonen> | Hixie: I see you didn't address my pending feedback about line breaks in input type=text. should I start sending comments on the form stuff or assume that there's still pending feedback for stuff that has been migrated to html5 spec? |
| 08:34 | <Hixie> | i haven't replied to a single WF2 e-mail yet |
| 08:34 | <hsivonen> | ok |
| 08:34 | <Hixie> | i'd hold back if you have feedback that i haven't yet dealt with |
| 08:34 | <hsivonen> | ok |
| 08:35 | <Hixie> | i'm only about half-done on WF2 integration so far |
| 08:35 | <Hixie> | it's proving at least as hard as i imagined it would |
| 08:36 | <Hixie> | only one more type="" to do! |
| 08:36 | <Hixie> | type=image |
| 08:40 | <hsivonen> | I'm glad to see that the CSS WG said no to versioning syntax: http://www.w3.org/mid/48E4132F.9060201⊙in |
| 08:41 | <Hixie> | css is where i first realised the value of no versioning |
| 09:06 | <hsivonen> | the specialness of lang&xml:lang continues to result in insane amounts of code... :-( |
| 09:12 | <Hixie> | more code than prose? |
| 09:12 | <hsivonen> | Hixie: more code |
| 09:13 | <hsivonen> | Hixie: in the Java version of the parser, I want to have a non-browser sanity mode that magically turns lang on HTML elements to xml:lang in the infoset |
| 09:14 | <hsivonen> | Hixie: but in the C++ version, I don't want to pay the memory footprint that the feature would take if latent |
| 09:14 | <hsivonen> | (and in the Java version not paying for it as RAM footprint would make the code slower) |
| 09:15 | <hsivonen> | on top of that, there are all sorts of checks on the validation layer |
| 09:15 | <gpy> | morning laddies |
| 09:15 | <hsivonen> | different ones for HTML and XHTML |
| 09:15 | <Hixie> | sounds like a pain |
| 09:15 | <Hixie> | i recommend not supporting xhtml. :-D |
| 09:15 | <Hixie> | (only kidding) |
| 09:16 | <hsivonen> | If I could go back in time, I'd ask the definers of XML get rid of the xml: stuff and to grandfather lang and id. (and zap xml:space) |
| 09:16 | <Hixie> | you'd be ignored like everyone who _did_ ask them to doso |
| 09:17 | <hsivonen> | at least I had the foresight to arrange my data structures so that I can just clip my arrays by one element when generating the C++ version |
| 09:22 | <hsivonen> | I wonder what would break if we defined lang in HTML to become an attribute whose local name and qualified name are "lang" but the namespace is the XML namespace |
| 09:23 | <hsivonen> | on HTML nodes, that is |
| 09:23 | <hsivonen> | not on SVG and MathML nodes |
| 09:25 | <Hixie> | getAttribute('lang'), for one, if i understand you right |
| 09:26 | <Hixie> | also you'd still have a syntax issue, except now lang="" in HTML would be equal to xml:lang="" in XML and lang="" in XML and xml:lang="" in HTML would both be null-namespace attributes with no semantics, leading to no end of extra confusion |
| 09:26 | <hsivonen> | Hixie: why? it matches on qualified name, which would be "lang" |
| 09:26 | <Hixie> | i thought it matches on name + namespace=null |
| 09:26 | <hsivonen> | IIRC, no |
| 09:26 | <Hixie> | so how are clashes resolved? |
| 09:26 | <Hixie> | oh, qname |
| 09:27 | <Hixie> | man that would be so confusing |
| 09:27 | <hsivonen> | Hixie: however, it would break setAttribute("lang", 'en') |
| 09:27 | <hsivonen> | unless setAttribute were made magic |
| 09:27 | <Hixie> | i don't care about lang enough to do this |
| 09:27 | <Hixie> | :-) |
| 09:29 | <hsivonen> | well, in general, namespaces result in an insane amount of code that doesn't do anything useful but has to be there for correctness and compat |
| 09:29 | <Hixie> | yes |
| 11:29 | <annevk2> | why is multiple file upload missing? multiple file upload is crucial |
| 11:50 | Philip` | now has a streaming HTML-to-XHTML web service, but it takes about 6.5 seconds to convert the 3MB HTML5 spec, which is better than html5lib but not great :-( |
| 11:51 | <Philip`> | (It's CPU-bound now, and I can't think of any obvious ways to improve it that don't involve getting a faster CPU) |
| 11:53 | <annevk2> | do you use Validator.nu? |
| 11:53 | <olliej> | yoyo annevk2 |
| 11:53 | <Philip`> | I use its parser and serialiser |
| 11:54 | <annevk2> | olliej, hey |
| 11:54 | annevk2 | should really go to work; woke up midday due to jetlag |
| 11:57 | <olliej> | annevk2: wow, you need an excuse for that? |
| 11:57 | olliej | got to work at 2 today |
| 11:58 | <erlehmann> | Philip`: HTML-to-XHTML ? why ? |
| 11:58 | <jgraham> | erlehmann: Because takkaria is slow ;) |
| 11:59 | <jgraham> | (or rather because html5lib is slow and the validator.nu parser doen't have a way of binding to python and takkaria doesn't have a stable hubbub release yet) |
| 11:59 | <erlehmann> | ??? |
| 12:00 | <jmb> | jgraham: we're working on that (slowly) |
| 12:00 | <jgraham> | Which means that parsing html in python is slow whereas parsing xml is fast |
| 12:01 | <jgraham> | So html->xml using validator.nu followed by xml parsing in python might be faster than html5lib |
| 12:02 | <erlehmann> | yay |
| 12:02 | <jgraham> | jmb: I was being tounge in cheek abot blaming you :) Although I am very much looking forward to hubbub |
| 12:02 | <Philip`> | erlehmann: What jgraham said :-) |
| 12:02 | <jmb> | jgraham: aye |
| 12:06 | <annevk2> | olliej, well, usually when I'm in Oslo I try to work 10-6 since I'm not here that often and most people work those hours |
| 12:06 | <olliej> | annevk2: it's 4am and i'm still alive |
| 12:06 | <olliej> | s/alive/at the office ;) |
| 12:07 | <annevk2> | ouch |
| 12:07 | <annevk2> | also, friday beer and general shutdown today is at 5PM |
| 12:07 | <olliej> | annevk2: ah |
| 12:07 | <olliej> | annevk2: i've just been trapped in MSVC hell |
| 12:07 | <annevk2> | anyway, got to go |
| 12:15 | <hsivonen> | Philip`: what was the problem with HTML2XML and escapes? |
| 12:16 | <hsivonen> | whoa! |
| 12:16 | <hsivonen> | that's a weird bug |
| 12:21 | <hsivonen> | just a break statement missing |
| 12:22 | <hsivonen> | I'm amazed the bug was there for so long without getting noticed |
| 12:23 | <hsivonen> | Philip`: fixed in SVN |
| 12:24 | <hsivonen> | Philip`: thanks |
| 12:25 | <hsivonen> | well, it didn't render output ill-formed, so I guess an XML serializer can do something very bogus and still not get caught by an XML parser |
| 12:25 | <hsivonen> | so now Philip` has found a bug in my XML serializer but still not a bug causing ill-formedness :-) |
| 12:38 | <annevk2> | the spec even says "There must be no more than one file in the list of selected files." :/ |
| 12:44 | <hsivonen> | Philip`: javadocs about default policy also fixed in SVN. thanks |
| 12:46 | <annevk2> | Hixie, fwiw, I agree with you on string comparison; also note that it's not exactly a same origin check as you need an additional check for the path component which sicking neglected to mention |
| 12:47 | <annevk2> | maybe s/neglected/forgot/ |
| 12:48 | <annevk2> | (that is, making sure the path component is not specified at all so that Access-Control-Allow-Origin: http://example.com/ fails) |
| 13:06 | <Philip`> | hsivonen: If your criteria is limited to ill-formedness bugs and doesn't count other bugs, then the implementation is trivial, since you could write "String serialize(Document doc) { return "<foo/>"; }" and it would never be ill-formed :-) |
| 13:08 | <Philip`> | hsivonen: Is there an easy way I can compile a new htmlparser-....jar with the fixed code? |
| 13:10 | <hsivonen> | Philip`: no. I'll build a new jar. |
| 13:33 | <hsivonen> | Philip`: new version released: http://about.validator.nu/htmlparser/ |
| 13:34 | <hsivonen> | (wow. now I've actually had to make a new release because of an XML serializer bug Philip` found...) |
| 13:34 | <annevk2> | is there some way to instantiate Validator.nu and get a Python bound libxml tree as output? |
| 13:36 | <hsivonen> | annevk2: locally, you can invoke java -cp htmlparser-1.1.1.jar nu.validator.htmlparser.tools.HTML2XML and pipe the input HTML to it, pipe the XML out and parse it with libxml in Python |
| 13:37 | <annevk2> | it seems that DreamHost has JDK so it might be worth looking into that |
| 13:37 | <hsivonen> | hopefully later you could use the C++ translator with a type mapper that maps to the Python C API types instead of NSPR/XPCOM types |
| 13:37 | <annevk2> | for Anolis, that is |
| 13:38 | <hsivonen> | annevk2: Didn't Philip` just make a network service that is equivalent to HTML2XML? |
| 13:38 | <hsivonen> | (and that saves the JVM startup overhead) |
| 13:39 | <annevk2> | it does involve downloading 3MB in case of HTML5 |
| 13:39 | <annevk2> | twice |
| 13:39 | <hsivonen> | good point |
| 13:39 | <Philip`> | hsivonen: Thanks! |
| 13:39 | <Philip`> | annevk2: You really don't want to start up a new JVM each time you want to parse a document |
| 13:39 | <annevk2> | though maybe Philip` is working on Anolis now in which case I don't have to :) |
| 13:40 | <hsivonen> | you could also try compiling nu.validator.htmlparser.tools.HTML2XML with GCJ to avoid the JVM overhead |
| 13:40 | <Philip`> | but you could run the web service thing (which is just a trivial wrapper around Jetty and the parser/serialiser) locally |
| 13:41 | <hsivonen> | Philip`: on dreamhost, doing so requires dreamhost PS |
| 13:41 | <hsivonen> | Philip`: not allowed on regular accounts |
| 13:41 | <Philip`> | hsivonen: Oh, that could be a problem |
| 13:41 | <Philip`> | annevk2: It's much less than 3MB if you compress it :-) |
| 13:41 | <Philip`> | but then it'll get CPU-bound more easily |
| 13:42 | <annevk2> | why would the JVM not be on by default btw? |
| 13:42 | <Philip`> | Without compression, it takes about 6 seconds (on my very limited CPU) which is 500KB/s which isn't really much |
| 13:43 | <Philip`> | so you shouldn't really need to worry about bandwidth usage |
| 13:43 | <hsivonen> | annevk2: IIRC, Dreamhost default accounts allow you to run Java programs by default but you aren't allowed to leave the processes sitting in the RAM |
| 13:44 | <hsivonen> | and leaving the process waiting is the whole point of avoiding the JVM startup overhead |
| 13:44 | <Philip`> | (Hmm, 25% of the runtime is in sun.nio.cs.UTF_8$Encoder.encodeArrayLoop) |
| 13:45 | <hsivonen> | Philip`: do you have ICU4J in classpath? |
| 13:45 | <annevk2> | ok, Java support indeed there |
| 13:45 | <Philip`> | There's the HotSpot overhead too, since it's really slow the first time you try to pass a document through the process |
| 13:45 | <Philip`> | hsivonen: Hmm... No |
| 13:45 | <annevk2> | but I suppose I can use the service from Philip` |
| 13:45 | <hsivonen> | Philip`: just wondering if the IBM UTF-8 impl is any faster |
| 13:46 | hsivonen | wonders how many distributed moving parts the spec generation now has... |
| 13:46 | <annevk2> | but is Validator.nu much faster than the libxml HTML parser? that's currently being used and already it's slow :/ |
| 13:47 | <Philip`> | hsivonen: It still says it's spending lots of time in sun.nio.cs.UTF_8$Encoder.encodeArrayLoop |
| 13:47 | <hsivonen> | Philip`: ok. I guess it's not picking up the ICU4J code automatically :-( |
| 13:47 | <Philip`> | I presume it is actually using ICU4J now, since it's got some calls to com.ibm.icu.charset.UConverterDataReader.readMBCSTable too |
| 13:48 | <hsivonen> | Philip`: from the name, that looks like a static method that runs on initial class loading |
| 13:48 | <Philip`> | hsivonen: Oh, okay |
| 13:49 | <hsivonen> | annevk2: I haven't benchmarked the Validator.nu parser against libxml2 |
| 13:56 | <Philip`> | annevk2: On my computer, parsing/reserialising the spec with libxml2 (via a Python script) takes about 0.55 seconds and with validator.nu (via (local) HTTP) takes about 0.75 seconds |
| 13:58 | <annevk2> | so it won't go faster |
| 13:58 | <annevk2> | but since libxml2 html can't be used... |
| 14:00 | <Philip`> | If the UTF-8 encoding took zero time then it'd be faster than libxml2 :-) |
| 14:01 | <annevk2> | I should also figure out how to make Python run faster on DreamHost |
| 14:13 | <Philip`> | hsivonen: http://about.validator.nu/htmlparser/ - "Fixed JavaDocs about XML violatation policy defaults." - s/violatation/violation/ |
| 14:14 | <hsivonen> | Philip`: thanks. fixed |
| 16:18 | <gsnedders> | id=draft-recommendation-—-date:-01-jan-1901 — that really isn't nice. |
| 16:19 | <annevk2> | do IDs after dates? |
| 16:20 | <Philip`> | Why does it need an id at all? |
| 16:20 | <gsnedders> | Philip`: Currently all headers get ids |
| 16:21 | <gsnedders> | annevk2: Then it changes every time the date changes |
| 16:21 | <Philip`> | gsnedders: Why? It seems pointless if nothing is linking to it |
| 16:21 | gsnedders | can't remember |
| 16:21 | <gsnedders> | There was a reason |
| 16:22 | <Philip`> | and for the headings where you might actually want to externally link to, like the status and abstract, they already have explicit non-auto-generated id values in the source |
| 16:22 | <gsnedders> | Yeah, that isn't changed. |
| 17:02 | <erlehmann> | why does XHTML 5 not attempt to provide a replacement for the modularization scheme defined and used by XHTML 1.1 ? |
| 17:03 | <Philip`> | erlehmann: Because the people developing XHTML 5 do not see any value in defining a modularisation scheme, as far as I'm aware |
| 17:04 | <Philip`> | (No value to the web, in particular) |
| 17:04 | <erlehmann> | Philip`: i used the modularisation to allow specific XHTML tags in comments. |
| 17:04 | <hsivonen> | erlehmann: couldn't you have done that with a monolithic schema by deleting stuff from it? |
| 17:05 | <Philip`> | erlehmann: Why does modularisation help, rather than just you writing a list of the elements and attributes you want to permit? |
| 17:05 | <erlehmann> | yeah, but i'd have to write the DTDs myself ;) |
| 17:05 | <hsivonen> | erlehmann: what happens when you and the XHTML2 WG disagree on the grouping of elements? |
| 17:06 | <erlehmann> | hsivonen: they'll probably win |
| 17:06 | <hsivonen> | erlehmann: have you ever wanted to allow one but disallow another element within one module (as defined by the XHTML 2 WG)? |
| 17:06 | <erlehmann> | i'm no match for the internet alite |
| 17:06 | <erlehmann> | no. |
| 17:17 | <gsnedders> | BenMillard: 3:19.594, with a large mistake going into the Porsche Curves on RM/RM |
| 18:14 | <Philip`> | Why does no normal web server with reverse proxying cope with streaming transfers? :-( |
| 18:15 | <Philip`> | All I want it to do is redirect the request to another server, and redirect the response back, without buffering everything in the middle |
| 18:16 | <Philip`> | (If I remember correctly, Apache and Lighttpd buffer the whole response, and Nginx also buffers the whole request) |
| 20:20 | <Philip`> | Aha, Squid does reverse proxying properly |
| 20:20 | <Philip`> | though it only supports HTTP 1.0 |
| 20:21 | <Philip`> | so if I use curl with default options then it gets an error message, because curl tries sending Expect: 100-continue which Squid doesn't like |
| 20:22 | <Philip`> | HTTP is so nice |
| 20:26 | <Hixie> | annevk2: i want to do multiple file uploads with something other than min/max |
| 20:47 | <weinig|food> | Hixie: what were you thinking about besides min/max? |
| 20:47 | weinig | is about to implement multiple file uploads |
| 20:50 | <Hixie> | i dunno |
| 20:50 | <Hixie> | count="" or something maybe |
| 20:50 | <Hixie> | or multiple="" |
| 20:51 | <Hixie> | like on <select> |
| 20:53 | weinig | nods |
| 20:53 | weinig | thinks min/max is awkward for it |
| 20:53 | <aboodman> | i think multiple would be better than min/max fwiw |
| 20:54 | <Hixie> | yeah it's totally overloading min max in a silly way |
| 20:54 | <Hixie> | i like multiple |
| 20:54 | <Hixie> | but i only just thought of it |
| 20:54 | <Hixie> | i haven't really looked at this carefully |
| 20:54 | <aboodman> | cprince brought this up a long time ago, but i forgot to ever mention it. sorry. |
| 21:14 | <annevk2> | hmm, I don't care strongly either way on the method I suppose, but it's needed |
| 21:15 | <annevk2> | just having a simple boolean attribute might be enough |
| 21:24 | <erlehmann> | Hixie: wouldn't it be multiple="multiple" ? |
| 21:24 | <Hixie> | it would be a boolean attribute called "multiple" |
| 21:24 | <Hixie> | i just use ="" to mean "attribute" |
| 21:24 | <Hixie> | anyway, afk, bbiab |
| 21:24 | <annevk2> | also, boolean attributes in HTML5 are allowed to have the empty string as value |
| 21:28 | <erlehmann> | empty means what exactly ? true or false ? |
| 21:28 | <erlehmann> | also, a tagcloud is a nav element, amirite ? |
| 21:29 | <annevk2> | if the attribute is present, true, false otherwise |
| 22:33 | <Hixie> | tag clouds are a list |
| 22:34 | <Hixie> | probably an unordered list with an attribute (e.g. data-freq="") giving the frequency (there's no semantic in html for marking up the frequency really) |
| 22:37 | <annevk2> | nested <em> |
| 22:38 | <annevk2> | glad to hear multiple files will be fixed |
| 22:39 | <annevk2> | i'm ok with it not being max=n I think, it doesn't really matter that much |
| 22:39 | <annevk2> | though we might have to support both for a couple of releases |
| 23:04 | <BenMillard> | Hixie, I've seen <a href title="30 entries"> and similar used in tag clouds to indicate frequency to users. |
| 23:05 | <Hixie> | annevk2: nested <em> would be wrong per html5. :-) |
| 23:05 | <Hixie> | BenMillard: that can work |
| 23:05 | <BenMillard> | sometimes it's in the content, like <li><a href>foo</a> (30)</a> or <li><a href>bar (30 entries)</a> |
| 23:05 | <BenMillard> | oops |
| 23:05 | <BenMillard> | <li><a href>Foo</a> (30)</li> |
| 23:10 | <BenMillard> | I might use <ol> if the items were ordered by their frequency |
| 23:11 | <BenMillard> | using <li value> to give the frequency makes sense in the markup but looks to weird in the rendering :) |
| 23:12 | <Hixie> | not that i'm anywhere near being able to do anything about it what with all this web forms work, but, anyone know if the svg group made any progress on the svg in text/html thing? |
| 23:12 | <Hixie> | just curious |
| 23:12 | <Hixie> | BenMillard: yeah, the order is more usually alphabetic or something (for which <ul> probably makes sense) |
| 23:13 | <othermaciej> | Hixie: I believe they are waiting for feedback from you on their proposal |
| 23:13 | <Hixie> | no, i sent them that feedback |
| 23:14 | <BenMillard> | Hixie, yeah that's what I see |
| 23:15 | <BenMillard> | Hixie, you'll be lucky to find many tag clouds using list markup today, though. :) |
| 23:15 | <Hixie> | othermaciej: http://lists.w3.org/Archives/Public/public-html/2008Aug/0852.html |
| 23:16 | <Hixie> | BenMillard: yeah really |
| 23:16 | <BenMillard> | Hixie, I've seen <div class> in the place of <ul> and <div style> in the place of <li> |
| 23:17 | <BenMillard> | for tag clouds which are just a big clump of links, <p class><a href> <a href> happens |
| 23:17 | <Hixie> | yeah |
| 23:17 | <Hixie> | would be good to convey the semantics somehow |
| 23:18 | <BenMillard> | Hixie, it's trivial to make lists of links display like a clump so I think <ul> is fine |
| 23:19 | <BenMillard> | <a href style> is the most common way to set the frequencies, AFAICT |
| 23:19 | <BenMillard> | where style contains different values of font-size |
| 23:19 | <othermaciej> | Hixie: something chaals said gave me the impression that SVG WG didn't think the ball was in their court |
| 23:20 | <othermaciej> | Hixie: but I can't find the specific email remark that made me think that |
| 23:20 | <Hixie> | ah |
| 23:21 | <Hixie> | i thought last i spoke with shepazu he agreed that it was in their court |
| 23:21 | <Hixie> | i guess i'll speak to them at the TPAC |
| 23:21 | <Hixie> | hopefully i'll be done with wf2 by then |
| 23:21 | <BenMillard> | Hixie, <ul><li><a href title></li>...</ul> seems the best way to me (swapping to <ol> if you order the items in some significant way) |
| 23:22 | <BenMillard> | (typos aside...) |
| 23:23 | <smedero> | SVG in HTML is in the SVG WG |
| 23:23 | <smedero> | ... court: http://lists.w3.org/Archives/Public/public-html/2008Aug/0861.html |
| 23:24 | <Hixie> | BenMillard: the main thing the visual tag cloud has is an indication of the top terms, which you lose without semantics in audio |
| 23:24 | <erlehmann> | Hixie: i don't really get why you are pushing this hard for continuing SGML markup. |
| 23:24 | Hixie | checks to make sure he's still in the same universe |
| 23:25 | <Hixie> | come again? |
| 23:25 | <erlehmann> | i mean, there are things it can't do. why not just usi XML then ? |
| 23:25 | <Hixie> | who's using SGML? |
| 23:25 | <BenMillard> | Hixie, that's what I'd use title for (as well as making the visual experience more specific on-demand) |
| 23:25 | <erlehmann> | Hixie: HTML is SGML. |
| 23:25 | <Hixie> | erlehmann: HTML5 is not SGML |
| 23:26 | <Hixie> | see section 8.2 |
| 23:26 | <Hixie> | in particular the sentence that says "While the HTML form of HTML5 bears a close resemblance to SGML and XML, it is a separate language with its own parsing rules. |
| 23:26 | <Hixie> | [...] This version of HTML thus returns to a non-SGML basis." |
| 23:28 | <erlehmann> | okay, yeah, sorry. |
| 23:29 | <BenMillard> | Hixie, the right-hand column here uses title in the way I describe under the "Tags" heading: http://wakeless.net/archive/2006/12/dom-inspector-release |
| 23:30 | <erlehmann> | still don't get it. migration of existing sites will be a huge task anyway, closing a few <br> elements may not add much |
| 23:31 | <BenMillard> | Hixie, it has <a href class title rel style> for every entry, which seems excessive :) |
| 23:31 | <BenMillard> | Hixie, but the title values are good, imho |
| 23:32 | <Philip`> | erlehmann: Migration of existing sites is trivial - you just write a <input type=email> or something, and now your site is making use of HTML 5 |
| 23:32 | <smedero> | erlehmann: hsivonen did a good job summarizing the history of the "HTML is SGML" confusion in his master's thesis: http://hsivonen.iki.fi/thesis/html5-conformance-checker#history |
| 23:32 | <Philip`> | erlehmann: (Migration to XHTML is far harder, because there are hundreds of obscure corner cases you have to cope with, and nobody ever gets them all right) |
| 23:33 | <erlehmann> | Philip`: if i really want to be semantic, it could be a tedious task replacing all the <div>s with <header>, <aside> and whatnot |
| 23:33 | <Philip`> | erlehmann: You're free to waste time if you want, but there's no reason you should do so |
| 23:36 | <erlehmann> | back to tagclouds. i currently assign custom classes. what exactly would a value attribute's value [sic] say ? |
| 23:44 | <erlehmann> | i'd be all for it being a value from 0 to 1 on a log scale :) |
| 23:45 | <gsnedders> | hiho |
| 23:51 | <erlehmann> | huhu |
| 23:51 | <gsnedders> | Can I blame jgraham for the speed of Anolis? He bullied me into keeping with Python when I was going to give up and use C++? |
| 23:53 | <Philip`> | Is Python the bottleneck? |
| 23:54 | <gsnedders> | Philip`: For the sub process, more or less |
| 23:56 | <Philip`> | It would be nice to try to make it work in Jython |
| 23:57 | <Philip`> | (and then you can use Java to optimise bottlenecks) |
| 23:58 | <gsnedders> | Philip`: http://hg.gsnedders.com/anolis/file/d0260bea7c0f/anolislib/processes/sub.py#l133 — can you see anyway to optimize that loop? |
| 23:59 | <Dashiva> | Do you have many nodes with no text, tail or attributes? |