| 00:02 | <annevk> | Hixie, how does the UI for peer-to-peer work? |
| 00:14 | <Hixie> | annevk: you mean the UA UI? |
| 00:14 | <Hixie> | there is no UA UI defined for peer-to-peer... it would be the script that sets that up |
| 00:16 | <annevk> | yeah |
| 00:16 | <annevk> | shouldn't I agree to connect my computer to some computer elsewhere? |
| 00:17 | <annevk> | maybe the implications are not too bad |
| 00:22 | <Hixie> | annevk: it doesn't seem to be any different than XHR or WebSocket |
| 00:51 | <othermaciej> | preventing exfiltration isn't really practical, so there's probably not much more reason to worry about peer-to-peer |
| 01:16 | <MikeSmithX> | the more interesting number for rb would be to measure how many sites that use ruby have rb as part of their ruby |
| 01:52 | <MikeSmith> | Hixie: so it would seem we are probably going to need for somebody to write an actual spec for SRT |
| 02:09 | <Hixie> | MikeSmith: yeah i figured as much. i can do that at some point, it's a pretty simple format. I wrote a draft already, actually, it's in some e-mail somewhere |
| 02:11 | <MikeSmith> | OK, that's good news at least |
| 02:11 | MikeSmith | looks at Hixie's "valid non-empty URL" change |
| 02:13 | <MikeSmith> | ah hey, yeah, you already had valid non-empty URL as a datatype for other attribute values |
| 02:13 | MikeSmith | goes to look at v.nu schema |
| 02:15 | <Hixie> | really? |
| 02:16 | <MikeSmith> | oh, maybe not |
| 02:17 | <MikeSmith> | if not, this might require adding a new datatype to the v.nu HTML5 datatype library |
| 02:18 | <MikeSmith> | hmm, yeah, @manifest was previously allowed to be empty but must be non-empty now too |
| 02:18 | <MikeSmith> | etc. |
| 02:21 | <MikeSmith> | yeah, it seems this will require a new datatype |
| 02:21 | MikeSmith | files a v.nu bug |
| 02:22 | <MikeSmith> | I wish there were some other contributors working on v.nu |
| 02:22 | <MikeSmith> | this would be an easy bug with which to get familiar with that part of the code |
| 02:23 | <MikeSmith> | Hixie: btw, about http://www.w3.org/Bugs/Public/show_bug.cgi?id=9198 |
| 02:23 | <MikeSmith> | since hspace and vspace are already now allowed on embed, I assume that bug must be asking that they be explicitly listed in the obsolete-and-nonconforming section |
| 02:23 | <MikeSmith> | *not allowed |
| 02:25 | MikeSmith | wonders if he can manage to get his spellchecker to forget the words "now" and "not" so that they will always get highlighted and I can get alerted to double-check that I'm using the right one |
| 06:01 | <MikeSmith> | Hixie: I notice there was never any response on the hybi list to John Fallows's "WebSocket API - proposal for binary frame support in JavaScript" |
| 06:02 | <MikeSmith> | http://lists.w3.org/Archives/Public/public-html-comments/2010Jan/0009.html |
| 06:02 | <MikeSmith> | oh, sorry |
| 06:02 | <MikeSmith> | I see it was on the public-html-comments list |
| 06:02 | <MikeSmith> | (which makes sense because it's about the API..) |
| 06:03 | <MikeSmith> | Hixie: should I maybe open a bugzilla bug for it? |
| 06:03 | MikeSmith | really does need to get some food, will step out for a bit |
| 06:04 | <MikeSmith> | hmm, I see that you did actually respond.. I missed the reply because I was looking in the online archives |
| 06:05 | <MikeSmith> | and it got split across the month boundary |
| 06:05 | <wirepair> | mike, you have any more information on that proposal for binary frame support? |
| 06:05 | <wirepair> | that seems kinda iffy |
| 06:05 | <MikeSmith> | wirepair: I know no more than what John posted in the message above |
| 06:06 | <wirepair> | gotcha |
| 06:06 | <wirepair> | ps. don't catch norovirus |
| 06:06 | <MikeSmith> | hai |
| 06:06 | <wirepair> | i've been out for 3 days ;< |
| 06:06 | <wirepair> | things killing me |
| 06:06 | <MikeSmith> | I gargle podine iodine to ward off the bugs |
| 06:07 | <wirepair> | good deal. heh |
| 06:07 | <MikeSmith> | the brown stuff with the hippo character |
| 06:08 | MikeSmith | grabs his umbrella and heads out |
| 06:55 | <othermaciej> | MikeSmith: shouldn't it be on the webapi wg list? |
| 08:22 | <othermaciej> | there have been 666 bugs filed in the HTML5 spec components since TPAC |
| 08:25 | <annevk> | nice number |
| 08:28 | <annevk> | with the peer-to-peer thingie it seems that everyone who gets your address can just connect with you |
| 08:29 | <annevk> | depending on what chat system is running on the site you might get tricked into giving information away to third parties you did not mean to |
| 08:29 | <othermaciej> | I just calculated our total incoming and outgoing bugs and issues since TPAC |
| 08:29 | <othermaciej> | neither number is really trending to 0 |
| 08:30 | <othermaciej> | where is the peer-to-peer thingie spec'd? |
| 08:30 | <annevk> | http://html5.org/tools/web-apps-tracker?from=4828&to=4829 is all there is |
| 08:30 | <annevk> | so there is some speculation from me is to how it works since it's not all defined |
| 08:34 | <othermaciej> | the hard part of peer-to-peer is discovering peers |
| 08:34 | <othermaciej> | most systems start with a well-known host that does the initial coordination |
| 08:35 | <othermaciej> | also not clear what protocol these proposed interfaces would use |
| 08:35 | <othermaciej> | also this probably won't work very well for users behind firewalls if it works in any kind of obvious way |
| 08:36 | <othermaciej> | to get to users on network segments using NAT or otherwise firewalled, you need to tunnel through a central server |
| 08:36 | <othermaciej> | like I gather Opera Unite does |
| 08:38 | <annevk> | the coordination goes through some kind of server in typical usage (non-typical would be to coordinate it over the phone) |
| 08:39 | <othermaciej> | telling someone your IP address over the phone is not an 80% use case |
| 08:42 | <annevk> | yeah |
| 08:46 | <othermaciej> | not even sure if peer-to-peer in general is an 80% use case |
| 08:46 | <othermaciej> | I'd like to see WebSocket implemented and deployed first |
| 08:46 | <othermaciej> | which reminds me that I need to review the new handshake and also get abarth to do so |
| 08:47 | <annevk> | IM is not 80%? |
| 08:50 | <othermaciej> | you don't need peer-to-peer for IM |
| 08:51 | <othermaciej> | Jabber goes through a server, so does IRC, to the best of my knowledge AIM/MSN/ICQ/etc also go through a central server |
| 08:52 | <othermaciej> | the only true peer-to-peer chat I get on my Mac, as far as I know, is Bonjour chat in iChat, which is only a neat trick because it autodiscovers people on the local network segment |
| 08:52 | <othermaciej> | but the interfaces you linked don't seem to have any provision for autodiscovery |
| 08:53 | <othermaciej> | autodiscovery is the hard part |
| 08:55 | <othermaciej> | (IRC does have the possibility of DCC which is directly peer-to-peer but it doesn't seem to be used that often) |
| 08:58 | <othermaciej> | I would guess audio or video chat probably does go peer-to-peer usually |
| 08:58 | <othermaciej> | but not always |
| 08:59 | <othermaciej> | there is group audio chat like Ventrillo which definitely uses a central server |
| 09:01 | <annevk> | mkay |
| 09:05 | <annevk> | autodiscovery would be nice |
| 09:08 | <virtuelv> | othermaciej: in terms of Opera unite it does one of two things |
| 09:08 | <virtuelv> | router configuration (uPnP) and/or a central proxy |
| 09:09 | <othermaciej> | oh, I think MobileMe also has name-based peer-to-peer WAN autodiscovery by name through a central server |
| 09:09 | <virtuelv> | there is also local network discovery |
| 09:09 | <othermaciej> | discovery is the hard part |
| 09:10 | roc | spams webkit-dev |
| 09:10 | <othermaciej> | and letting a web page advertise your host for autodiscovery without permission is the potentially scary security bit |
| 09:13 | <annevk> | ohuh |
| 09:13 | <Hixie> | othermaciej: iChat is peer-to-peer also once you've established the link, as i understand it. Certainly Skype is. |
| 09:13 | <Hixie> | annevk: there's no reason the address couldn't be unique per PeerToPeerServer -- indeed, that seems like a good idea |
| 09:13 | <othermaciej> | Hixie: iChat is an app not a protocol - which protocol do you mean? |
| 09:14 | <Hixie> | othermaciej: the one used for video chat, sorry |
| 09:14 | <othermaciej> | I believe its use of AIM and Jabber is not peer-to-peer, but I believe audio and video chat and Bonjour chat are peer-to-peer |
| 09:14 | <Hixie> | sounds right |
| 09:14 | <othermaciej> | (audio and video peer discovery being done via the central server initially) |
| 09:14 | <virtuelv> | skype uses STUN, afaict |
| 09:14 | <virtuelv> | http://www.ietf.org/rfc/rfc3489.txt |
| 09:14 | <Hixie> | video conferencing is the main use case i'm imagining this stuff being used for |
| 09:14 | <Hixie> | anyway this is all highly experimental so far |
| 09:15 | <virtuelv> | plus http://tools.ietf.org/html/draft-ietf-behave-turn-16 |
| 09:15 | <Hixie> | i'm just poking around with proposals because it comes up every now and then |
| 09:15 | <virtuelv> | http://adamfisk.wordpress.com/2006/08/16/cringely-skype-open-infrastructure/ |
| 09:16 | <annevk> | Hixie, yeah, I was thinking it had to be tied to the origin or something |
| 09:16 | <annevk> | but unique would be even better |
| 09:17 | <Hixie> | annevk: i hope that we don't have to define the network-level stuff but can instead just defer to a separate spec, the way the websocket api defers to a websocket protocol spec, except without me having to write the network side :-) |
| 09:18 | <annevk> | I wonder what the quality of such a spec would be |
| 09:18 | Philip` | thought the point of Skype was that it tried loads of connection methods, which is necessary in order to make something that works for all users |
| 09:19 | <Philip`> | e.g. if one user can accept external network connections, they get used as a proxy between users that can't |
| 09:19 | <othermaciej> | like I said - discovery is the hard part |
| 09:20 | <Philip`> | (so it'll silently use up your bandwidth) |
| 09:20 | <Philip`> | and it does various tricks like trying to use port 80 to get around firewalls |
| 09:20 | <Philip`> | (as well as all of the NAT traversal) |
| 09:22 | <Philip`> | This isn't the discovery part, it's the part where you figure out how to send traffic to someone after you've discovered them |
| 09:30 | <othermaciej> | when I say "discovery" I mean including the ability to actually establish a connection to the peer |
| 09:30 | <othermaciej> | I guess that's really discovery + routing |
| 09:34 | <Philip`> | Ah |
| 09:46 | <Hixie> | othermaciej: the change proposal link in ISSUE-66 in the issue-status list links to something other than it claims |
| 09:46 | <othermaciej> | Hixie: probably my falt, let me see |
| 09:46 | <othermaciej> | Hixie: ugh - it looks like the forwarding links from the old wiki are not working |
| 09:46 | <othermaciej> | MikeSmithX: are you around? |
| 09:47 | <othermaciej> | Hixie: probably all other links to the esw wiki are broken |
| 09:47 | <MikeSmithX> | othermaciej: I am now |
| 09:47 | <othermaciej> | MikeSmithX: esw wiki doesn't seem to be forwarding to the new wiki |
| 09:47 | <othermaciej> | or rather, all old links to it forward to the main page |
| 09:48 | <othermaciej> | is that expected? |
| 09:50 | <MikeSmithX> | it's expected but we are working on fixing |
| 09:50 | <MikeSmithX> | iy |
| 09:50 | <MikeSmithX> | it |
| 09:50 | <othermaciej> | ok |
| 09:50 | <othermaciej> | I should probably fix all the links to the old wiki on the issue status page |
| 09:52 | <othermaciej> | would it work to just global replace http://esw.w3.org/topic/HTML/ with http://www.w3.org/html/wg/wiki/ |
| 09:52 | <MikeSmithX> | yeah, it should |
| 09:53 | <othermaciej> | Hixie: fix't |
| 09:56 | <Hixie> | thanks |
| 09:57 | <othermaciej> | MikeSmith: by the way, your "age of oldest unresolved bug" number doesn't look so hot (from my recent email) - I hope you get a chance to do a pass over H:TML bugs soon |
| 09:58 | <MikeSmith> | othermaciej: yeah, there are a couple of old ones I know I can close out |
| 09:58 | <MikeSmith> | because I filed them myself |
| 09:58 | <othermaciej> | heh |
| 09:59 | <othermaciej> | I will forgive you if you skip the standard boilerplate on self-filed bugs (assuming no one else is going to be looking to escalate them or anything) |
| 10:00 | <MikeSmith> | ok |
| 10:02 | <othermaciej> | this article sounds amazingly authoritative given that the author clearly has no idea what he's talking about: http://www.betanews.com/article/Latest-HTML5-working-draft-published-despite-claims-of-sabotage/1268084800 |
| 10:09 | <MikeSmith> | aha, authoritative indeed |
| 10:09 | <MikeSmith> | 'By "accessibility," Le Hégaret was using a keyword in the argument in favor of breaking Canvas out from HTML5' |
| 10:09 | <othermaciej> | I am just amazed how may simple declarative statements there are false |
| 10:10 | <MikeSmith> | sharp character, this Scott M. Fulton, III is |
| 10:10 | <MikeSmith> | he understands all the secret code words |
| 10:10 | <Hixie> | i couldn't read it all the way through, it was too painful |
| 10:17 | <nessy> | what a twisted world view |
| 10:17 | <nessy> | some people can make plots out of anything! |
| 10:26 | <annevk> | oh hey, my blog is linked |
| 10:26 | <annevk> | good times |
| 10:27 | <hsivonen> | http://twitter.com/jdowdell/status/10187314697 |
| 10:28 | <hsivonen> | I guess the "range of expectation" is narrow |
| 10:29 | <Hixie> | i'm sure glad people are tweeting rather than filing bugs |
| 10:29 | <annevk> | the more important question is of course, do you care about SWF? |
| 10:29 | <Hixie> | makes my life a lot easier |
| 10:30 | <hsivonen> | annevk: at least on the desktop market, it would be considered a bug if CSS and SWF colors didn't look the same for a given RGB value |
| 10:31 | <hsivonen> | on a related note, neither swfdec nor gnash manages to show me any .swf files on Ubuntu (64-bit) |
| 10:32 | <hsivonen> | the context menus suggests the plug-ins do load |
| 10:43 | <othermaciej> | one of the top two reasons we can't colormatch images in HTML in Safari is because SWF doesn't colormatch vector drawing |
| 10:45 | <othermaciej> | I almost @replied to jdowdell but I regained my senses |
| 10:45 | <othermaciej> | because one of the top two reasons we can't colormatch everything in Safari is SWFs! |
| 10:59 | <MikeSmith> | hsivonen: thanks for the comment on the iri-ref v.nu bug I raise |
| 10:59 | <MikeSmith> | *raised |
| 11:00 | <MikeSmith> | hsivonen: the solution of changing the iri-ref datatype to never accept the empty string and changing the allowed value for attributes that do accept either the empty string or iri-ref does sound like the best way to handle it |
| 11:00 | <MikeSmith> | shall I go ahead and make that change? |
| 11:10 | <MikeSmith> | othermaciej, annevk - I created a new bugzilla HTML WG component for the HTML5 differences from HTML4 doc |
| 11:29 | <hsivonen> | MikeSmith: go ahead |
| 11:30 | <hsivonen> | I'm getting really close to point where the HTML5 parser in Gecko can be turned on by default |
| 11:30 | <hsivonen> | I hope to have some cycles for V.nu hacking when that's done |
| 11:32 | <MikeSmith> | hsivonen: cool |
| 14:24 | <Philip`> | "Exactly what is the purpose for this change proposal? Are you saying that you want to change the section in the specification?" - I thought it was fairly clear that the purpose of a change proposal is to propose a change to the specification |
| 14:27 | <Philip`> | (Maybe the confusion is that Shelley thinks it necessarily reflects the personal opinion of the person who is writing the proposal, rather than just being intended as an option for the WG to consider and potentially vote on) |
| 14:30 | <hsivonen> | Did Safari 4 change WebKit's doctype sniffing inspired by HTML5 as Leif claims? |
| 14:41 | <annevk> | i somewhat doubt it |
| 14:44 | <annevk> | doesn't look like it looking at trac |
| 14:44 | <gsnedders> | hsivonen: hyatt rewrote it around the Saf4 time, IIRC |
| 14:44 | <gsnedders> | (hyatt definitely rewrote it to match HTML 5 sometime) |
| 14:44 | <annevk> | really? |
| 14:45 | <annevk> | the doctype string file hasn't been touched for four years |
| 14:45 | <annevk> | and to match HTML5 that would need changing |
| 14:45 | <annevk> | furthermore the HTML parser code doesn't seem to be changed around doctypes either |
| 14:46 | <annevk> | meanwhile I noticed Safari implements <hgroup> among other new elements |
| 14:46 | <annevk> | well, WebKit |
| 14:48 | <annevk> | gsnedders, I can't find evidence for your claim |
| 14:48 | <annevk> | would be nice though if he did |
| 14:48 | <gsnedders> | annevk: He certainly changed it a fair bit to get it pretty close to HTML 5 |
| 14:49 | <annevk> | gsnedders, not per SVN |
| 14:49 | <gsnedders> | https://bugs.webkit.org/show_bug.cgi?id=17639 is one bug |
| 14:50 | <annevk> | hmm, that bug is marked WFM |
| 14:50 | <annevk> | and the file it talks about hasn't been touched for four years |
| 14:50 | <gsnedders> | Indeed |
| 14:51 | <gsnedders> | Was the list in HTML 5 not originally based upon WebKit? |
| 14:51 | <annevk> | yes |
| 15:07 | <Philip`> | Shelley's view of change proposals seems to make compromise impossible, if everybody only proposes what they personally think is best (given that the result must be the acceptance of exactly one change proposal) |
| 15:09 | annevk | wonders if everyone has the experience outlined here: http://camendesign.com/blog/choice |
| 15:10 | <annevk> | balot screen kind of fails if that's true |
| 15:11 | <gsnedders> | annevk: Apparently happens when you upgrade to IE8 and ballot screen at the same time |
| 15:13 | <meledin> | Those pictures are pure win |
| 15:14 | <Philip`> | At least the IE8 upgrade won't happen on Win7 (since it ships with that version) |
| 15:15 | <meledin> | Even if it |
| 15:15 | <meledin> | is accidental, I'd have to give props to Microsoft for ingeniously circumventing the ballot |
| 15:18 | <Philip`> | Forcing users to respond to an unsolicited prompt that expects them to make an important decision is always going to be bad UI design |
| 15:18 | <Philip`> | and I don't see an obvious way that Microsoft could have made it much better |
| 15:21 | <Philip`> | (It seems they only do these prompts for IE8 (which is important for security and compatibility, and is a major user-visible change so it shouldn't be done silently, so it's worth annoying the user for it) and browser choice (which is legally required), and otherwise successfully avoid it) |
| 15:22 | <hsivonen> | considering that the ballot is legally required, I'm surprised they haven't done more code review and QA on it |
| 15:42 | <gsnedders> | Anyone know of any Python WebIDL parser? |
| 15:46 | <Philip`> | gsnedders: http://suika.fam.cx/www/markup/html/whatpm/readme has a Perl one, which is close |
| 15:46 | gsnedders | slaps Philip` |
| 15:47 | <Philip`> | :-( |
| 15:47 | Philip` | often chooses what language to use based on available libraries rather than on properties of the language itself |
| 15:47 | <gsnedders> | So, should I go for one and use GHC, and use that as a Python extension? :P |
| 15:47 | <gsnedders> | *use GHC for it |
| 15:48 | <gsnedders> | Philip`: That decision has already been made |
| 15:48 | <Philip`> | Do you mean write your own? |
| 15:48 | <Philip`> | Also, do you mean GHC as in Haskell? |
| 15:48 | <gsnedders> | Yes |
| 15:49 | <gsnedders> | http://hackage.haskell.org/package/webidl |
| 15:50 | <Philip`> | That sounds more complex than using one written in a sensible language like Perl, and making it emit the parse tree as JSON so you can load it into Python |
| 15:50 | <gsnedders> | Maybe I should just rewrite Anolis in Haskell :P |
| 15:50 | <Philip`> | Alternatively, maybe you shouldn't :-p |
| 15:52 | <Philip`> | In any case, it's probably best to treat the IDL parser as a standalone tool that emits an easy-to-parse text stream, instead of tightly coupling your Python code to it |
| 15:56 | <boblet> | There’s a new doctor in the house — me ;-) |
| 15:56 | <boblet> | http://html5doctor.com/i-b-em-strong-element/ |
| 15:56 | boblet | really hopes no one spots an obvious mistake in like 5 seconds (gulp) |
| 15:57 | <paul_irish_> | very nice post. |
| 15:57 | <gsnedders> | Philip`: That means trying to write Perl ;P |
| 15:58 | <boblet> | paul_irish_: thanks! your powers of speed reading are … impressive! |
| 15:58 | <Philip`> | "i-b-em" - that sounds familiar |
| 15:58 | <paul_irish_> | haha |
| 15:58 | <boblet> | hehe |
| 15:59 | <gsnedders> | boblet: Would the examples for the b element not normally be ::first-line? |
| 15:59 | <boblet> | it just worked out that way, honest guv’nor! |
| 15:59 | <Philip`> | gsnedders: Probably only a little bit of Perl |
| 16:01 | <MikeSmith> | boblet: great stuff |
| 16:01 | <boblet> | gsnedders: one would, and the other wouldn’t — the :first-line example is the ”use :first-line not b to do this” one |
| 16:01 | <boblet> | thanks for your help MikeSmith |
| 16:01 | <MikeSmith> | boblet: I especially like that way you curl your capital letter Is |
| 16:01 | <MikeSmith> | or at least that one I |
| 16:02 | <boblet> | I am but your humble padawan ;-) |
| 16:02 | <MikeSmith> | do more of that |
| 16:02 | <MikeSmith> | oh my god |
| 16:02 | <MikeSmith> | that O outdoes the I |
| 16:02 | <MikeSmith> | that is audacious O |
| 16:02 | <MikeSmith> | that is like a "Story of O" O |
| 16:02 | <boblet> | versals in the house ;-) @font-face of course |
| 16:03 | <MikeSmith> | boblet: in future articles, you :first-letter examples should just successively use bigger and bigger letters, until the get up to 320px or so |
| 16:04 | <gsnedders> | boblet: But do you really often do such things with less than the opening line? My memory could alwasy be wrong… :P |
| 16:05 | <boblet> | MikeSmith: click through to the example page — it makes more sense there (the “O” is the first drop cap, the “I” is the size of subsequent drop caps ;-) ) |
| 16:05 | <MikeSmith> | blb |
| 16:05 | <MikeSmith> | boblet: sorry, I got distracted |
| 16:05 | <boblet> | gsnedders: in traditional typography the first word or phrase is common |
| 16:06 | <MikeSmith> | reading "Story of O" |
| 16:06 | <boblet> | traditional = scribes in brown habits, early years of the printing press etc |
| 16:07 | <gsnedders> | Yet of course it remained primarily stylisitc with print |
| 16:08 | <gsnedders> | And what I was taught was one very specific style of one printer |
| 16:08 | <MikeSmith> | boblet: anyway, seriously, very nice article |
| 16:08 | <gsnedders> | Anyone know off-hand what Ada's full name in "Ada or Ardor" is? |
| 16:09 | <boblet> | MikeSmith: I raise my mug of green tea to your good health :) |
| 16:09 | <gsnedders> | (for some obscure memonic reason) |
| 16:09 | <MikeSmith> | boblet: you have set a higher bar for the other HTML doctors |
| 16:09 | <MikeSmith> | you have made them look like HTML interns |
| 16:10 | <MikeSmith> | or HTML candy-stripe nurses |
| 16:10 | gsnedders | cheats and takes advantage of the fact that the text is available online |
| 16:10 | <gsnedders> | "A girl was born on July 21, 1872, at Ardis, her putative father's seat in Ladore County, and for some obscure mnemonic reason was registered as Adelaida. |
| 16:10 | <MikeSmith> | http://html5candystripenurses.com/ |
| 16:10 | <boblet> | gsnedders: I’ll have to ask John from ILT about it |
| 16:11 | <gsnedders> | ILT? |
| 16:11 | <boblet> | http://ilovetypography.com/ |
| 16:12 | <boblet> | MikeSmith: you’re too kind. But yeah I worry I may have got a little carried away. Hopefully they’ll forgive an enthusiastic locum |
| 16:12 | <Philip`> | http://ihatetypography.com/ - someone needs to register that |
| 16:12 | boblet | is very disappointed that nurses site isn’t registered |
| 16:13 | gsnedders | hates the opening chapters of Ada |
| 16:20 | <lazni> | in windows 7, ie8 still pops up that modal dialog at first load |
| 16:22 | <Philip`> | The upgrade-from-IE7 one? |
| 16:24 | <lazni> | yes, "Welcome to IE8" |
| 17:10 | <AryehGregor> | Hixie, could you say that if a URL resolves to the current resource, and the user agent knows that a text/html response will cause an error, it can decline to fetch the resource even if it's not actually in cache, on the assumption that it will be text/html? (Similar for XML MIME types.) |
| 17:12 | <AryehGregor> | As for validation, surely something somewhere already says that script/stylesheet/etc. links must point to documents of the correct type. E.g., if <script src=""></script> points to a text/html file, then that violates the sentence "If the language is not that described by "text/javascript", then the type attribute must be present, as described below." |
| 17:12 | <AryehGregor> | So a validator could recognize that "" is not a text/javascript resource and raise an error. |
| 17:15 | <Philip`> | AryehGregor: Some people (like hsivonen, I think) don't like the validity of a document depending on external resources |
| 17:16 | <Philip`> | It makes it really hard to check that a document is valid and that it stays valid |
| 17:16 | <AryehGregor> | You can't check all validity requirements anyway. |
| 17:16 | <Philip`> | (and you'd get different results when copying-and-pasting into a text input box vs pointing at the URL) |
| 17:16 | <AryehGregor> | Hmm, true. |
| 17:16 | <Philip`> | That's not a reason to add more requirements that can't be checked |
| 17:17 | <Philip`> | particularly since it's a lot of complexity for something that nobody is going to do on purpose |
| 17:17 | <AryehGregor> | The requirement already exists, I'm not suggesting it be added. :) |
| 17:17 | <Philip`> | s/particularly since/Also,/ |
| 17:17 | <Philip`> | Oh, okay |
| 17:17 | <AryehGregor> | Validators that don't want to depend on whether you copy-paste or point at the URL can implement only special cases of this check -- like saying it's an error only if the URL resolves to the current document, but HTML is the wrong type for the link. |
| 17:18 | <AryehGregor> | Which is the current case. |
| 17:19 | <Philip`> | Actually, <script src=""> doesn't point to a text/html file |
| 17:20 | <Philip`> | because (per the spec) it doesn't resolve to anything |
| 17:20 | <AryehGregor> | It doesn't? |
| 17:20 | <Philip`> | It hasn't done for at least 16 hours |
| 17:21 | <Philip`> | So I think that's not a problematic case, because it points at nothing and is clearly an error |
| 17:23 | <Philip`> | The problematic case was like <link rel="stylesheet index" href=""> |
| 17:23 | <Philip`> | where the stylesheet link doesn't point to anything, while the index link points to the base URL |
| 17:24 | <AryehGregor> | Well, I was making suggestions to Hixie based on the previous version of the spec. |
| 17:24 | <AryehGregor> | As an alternative fix. |
| 17:25 | <Philip`> | The previous version of the spec would download the current page again if you wrote <script src=""> |
| 17:25 | <Philip`> | which was the problem |
| 17:25 | <Philip`> | so the fix would have to involve changing it to not download the current page again |
| 17:25 | <Philip`> | and so it couldn't then define validity in terms of the downloaded resource |
| 17:26 | <Philip`> | although I think I forgot to read what you actually said |
| 17:27 | <AryehGregor> | Right. My suggestion was that the spec say that if a URL is known to resolve to the current resource, then if that URL is used for something that's not supposed to be HTML, the UA can raise an error without bothering to actually retrieve the resource. |
| 17:27 | <AryehGregor> | Or just ignore it, or something. |
| 17:28 | <AryehGregor> | And validators can say it's an error if they know it resolves to the current resource. |
| 17:28 | <AryehGregor> | I guess that's messier than just adding some special cases, but it seems less hackish. |
| 17:29 | <Philip`> | That seems like a lot of complexity for something that nobody is going to do on purpose |
| 17:31 | <Philip`> | It doesn't seem non-hackish to me, because it's still going to do weird things in hypothetical cases like a POST response page saying <script src=""> where the GET returns a different content-type |
| 17:32 | <Philip`> | and it'll be hard for authors to predict behaviour because it'll depend on whether the UA cached enough information about the page |
| 17:35 | <AryehGregor> | I guess. |
| 19:00 | <Dashiva> | It's so easy to forget that we are currently living in an era of low public-html traffic |
| 19:04 | <knowtheory> | what? |
| 19:13 | <Philip`> | Dashiva: We are? |
| 19:13 | <Philip`> | Monthly post counts: http://chart.apis.google.com/chart?cht=lc&chd=t:1773,1307,1099,1313,1258,576,398,491,324,307,441,370,802,767,462,455,965,515,301,592,263,694,818,734,288,657,857,954,1477,1220,1144,681,518,1459,908&chds=0,2000&chs=300x150 |
| 19:15 | <othermaciej> | Philip`: what is that counting? |
| 19:15 | <Dashiva> | The graph is deceitful, it doesn't compensate for christmas |
| 19:16 | <Philip`> | othermaciej: Monthly posts :-) |
| 19:16 | <Philip`> | The numbers from http://lists.w3.org/Archives/Public/public-html/ in particular |
| 19:16 | <othermaciej> | monthly posts to what? |
| 19:16 | <Philip`> | (excluding the periods that aren't months) |
| 19:16 | <othermaciej> | public-html? |
| 19:16 | <Philip`> | public-html |
| 19:19 | <knowtheory> | Dashiva: how is it decietful? |
| 19:38 | <othermaciej> | since we're throwing stats around - I made some monthly bug stats: http://spreadsheets.google.com/ccc?key=0AoCAfo_LQ5_kdFFWWmpCMWxsLUN2TW9VYi1uNEJGenc&hl=en |
| 19:41 | <othermaciej> | it's interesting that find rate and fix rate seem to roughly track each other, yet the resulting net incoming rate seems to oscillate quite a bit |
| 22:00 | <Hixie> | AryehGregor: that would fail to handle this http://www.hixie.ch/tests/evil/html/external-resources/001.test |
| 22:00 | <Hixie> | AryehGregor: (note no content-type headers) |
| 22:01 | <Hixie> | AryehGregor: (also note that that test is now out of date) |