| 02:29 | <MikeSmith> | annevk: fwiw, off-the-shelf highlighting in vim is quite good and pretty easily configurable, and it's undo/redo mechanism is significantly better than any other editor I've used |
| 02:30 | <MikeSmith> | it has unlimited undo, and a command, "earlier", that take a time argument - to take you back to the state of the buffer that point in time |
| 02:30 | <MikeSmith> | http://vimdoc.sourceforge.net/htmldoc/undo.html#:earlier |
| 02:34 | <heycam> | annevk, done the web idl abstract rewording you asked for earlier |
| 09:38 | hendry | wonders if there is a campaign to get rid of XSLT from browsers |
| 09:39 | Philip` | hopes not, since he has some code that uses it |
| 09:41 | <Dashiva> | What use is client-side XSLT apart from outsourcing the processing cycles to the browser? :) |
| 09:42 | <Philip`> | The XML data could come from a remote server you don't have control over, so you can't do the XSLT on the server side |
| 09:43 | <Philip`> | or you could want to transform a single XML file in several different ways, and not want to have the client download several slightly different XML files instead of just one |
| 09:46 | <annevk> | The remote server argument is slightly bogus without Access Control |
| 09:46 | <Philip`> | But it's less bogus with it :-) |
| 09:48 | <Dashiva> | Besides, if you want XSLT on remote server content you can pull it to your local server once and do XSLT there, leaving you with local requests too. Double win :D |
| 09:49 | <annevk> | Dashiva, it wouldn't be user-specific anymore in that case |
| 09:49 | <Philip`> | Dashiva: That's kind of nasty for performance and cost and security |
| 09:50 | <Philip`> | Also it requires too much cleverness on the server - it's much nicer when you can just serve static files |
| 10:02 | <Philip`> | hendry: About http://natalian.org/archives/2008/05/15/10-years-of-xml/ "Near impossible to diff (and nicely version control)" - there's research like http://www.seas.upenn.edu/~harmony/ that seems kind of relevant for that (though I've no idea how well it'd work in practice) |
| 10:04 | <hendry> | well I still think XSLT is unneccessary bloat in the browser |
| 10:04 | <hendry> | i've personally since very few websites use it. i wonder if there are some stats of how much of the "web" uses that publishing technique |
| 10:04 | <MikeSmith> | hendry: It's a moot point now |
| 10:05 | <MikeSmith> | and I reckon we've got a lot of other examples of unnecessary bloat in browsers |
| 10:06 | <MikeSmith> | anyway, it gives developers another choice |
| 10:12 | <MikeSmith> | hendry: a rant about XML can't be complete without specific mention of Draconian error handling |
| 10:13 | <hendry> | MikeSmith: that's Hixie's rant :) |
| 10:13 | <MikeSmith> | dunno if you've seen this: |
| 10:13 | <MikeSmith> | http://esw.w3.org/topic/HTML/DraconianErrorHandling |
| 10:15 | <hendry> | MikeSmith: ok ok, i'll link it :) |
| 10:16 | <hsivonen> | In the early days of Atom, I disagreed with markp on XML error handling. |
| 10:16 | <hsivonen> | I was wrong. He was right. |
| 10:17 | <MikeSmith> | hendry: fwiw, judging popularity of XML by how many people attended one session at XTech 2008 doesn't seem like any kind of real indicator at all (since there were only 100 or so total attendees at XTech anyway) |
| 10:19 | <MikeSmith> | hsivonen: yeah, Mark was the real pioneer in stating very clearly and emphatically and publicly about why it didn't and couldn't work |
| 10:19 | <hsivonen> | (although writing a liberal feed parser isn't the full answer. We still need a spec on how to be liberal in an interoperable way.) |
| 10:19 | <MikeSmith> | hsivonen: true |
| 10:19 | <Philip`> | The only reliable way to measure popularity is Google search result counts, and then you'll find that XML is 98% as popular as God |
| 10:19 | <MikeSmith> | heh |
| 10:20 | MikeSmith | heads off to write a rant about God |
| 10:20 | <MikeSmith> | speaking of God, the real rant I'd write if I had the time is, Please don't get religous |
| 10:21 | <MikeSmith> | especially don't get religious about technologies |
| 10:21 | <MikeSmith> | either way, good or bad |
| 10:21 | <MikeSmith> | because nobody really cares about your religious beliefs |
| 10:22 | <Philip`> | What do you mean by "religious"? |
| 10:22 | <Lachy_> | MikeSmith, it would be better if people didn't get religious about anything at all, not just technology. Even religion |
| 10:22 | <MikeSmith> | Lachy_: amen |
| 10:22 | <zcorpan> | Lachy: what if being religious makes one happier? |
| 10:22 | Philip` | assumes people tend not to literally worship technologies |
| 10:23 | <toolskyn> | Philip`: according to googlefight.com, xml is actually more popular than god |
| 10:23 | <MikeSmith> | Philip`: religous = dogmatic support/rejection of particular ideas/technologies/whatever |
| 10:23 | <Lachy> | zcorpan, getting high makes some people happier. That doesn't make it a good thing to do. |
| 10:24 | <Philip`> | What if you being religious makes everyone else happier (because e.g. it discourages you from harming them)? |
| 10:24 | <Dashiva> | Crushing your enemies, having them driven before you, and hearing the lamentations of their women also makes some people happy |
| 10:24 | <zcorpan> | Lachy: getting high has long-term physical damage, i don't see being religious having such consequences |
| 10:24 | <MikeSmith> | Lachy: getting high is a good thing to do, actually |
| 10:24 | <Lachy> | Philip`, the argument from morality you're partially using is bogus |
| 10:24 | <MikeSmith> | more people should do it |
| 10:25 | <MikeSmith> | I have personal empirical evidence that getting high is a good thing |
| 10:25 | <jgraham> | Being religious might have long term damage to society |
| 10:26 | <hsivonen> | there seems to be a clash between screen readers and hand ergonomics |
| 10:26 | <Lachy> | zcorpan, religion gets in the way of progress, particularly science and education, causes more wars than anything else in history, and is overall a big waste of time. |
| 10:26 | <hsivonen> | the default key bindings of both Orca and NVDA suck on a keyboard that doesn't have an Insert key |
| 10:26 | <MikeSmith> | Lachy: religion unites people toward common goald |
| 10:26 | <MikeSmith> | goal |
| 10:26 | <MikeSmith> | goalsssss |
| 10:26 | <MikeSmith> | and against common enemies |
| 10:26 | <Dashiva> | MikeSmith: Uniting with the common goal of not uniting with other people? ;) |
| 10:27 | <MikeSmith> | e.g. against XML, or against "bad HTML" |
| 10:28 | <MikeSmith> | or against "lazy authors" and in support of... whatever the alternative is ("educating authors"?) |
| 10:28 | <Philip`> | Lachy: It seems to have encouraged progress in architecture, given how most major building projects in the past have either been for war or religion |
| 10:28 | <Dashiva> | In support of requiring a HTML author's licence |
| 10:29 | <MikeSmith> | Dashiva: great idea |
| 10:29 | <MikeSmith> | we should issue those |
| 10:29 | <MikeSmith> | like 007 license to kill |
| 10:30 | <MikeSmith> | instead, license to write HTML that actually works in browsers |
| 10:30 | <MikeSmith> | and that parses interoperably |
| 10:30 | <Dashiva> | With licence to parse unambiguously |
| 10:30 | <Philip`> | If you get caught with overlapping tags then you get 3 points on your license |
| 10:31 | <hsivonen> | Philip`: you mean like <i <b> >? :-) |
| 10:31 | <Dashiva> | That would be an interesting mental exercise, write up penalty guidelines for a HTML licence |
| 10:31 | <toolskyn> | we should have one of these IE 'friendly warnings' about code written by a unlicenced author |
| 10:31 | <Dashiva> | How many points for not quoting an attribute containing spaces? |
| 10:32 | <Philip`> | hsivonen: No :-p |
| 10:32 | <Philip`> | hsivonen: (but it's not overlapping elements, since elements are just part of a tree and there's no concept of overlapping, and I can't remember a better term) |
| 10:32 | <Dashiva> | interleaved? |
| 10:32 | <Dashiva> | incestuous? |
| 10:33 | <MikeSmith> | "incestuous" is a much better term -- more intriguing |
| 10:33 | <Philip`> | I suppose "misnested" works |
| 10:33 | <Dashiva> | It does cover the "My parent is my sibling" thing |
| 10:34 | <hsivonen> | hmm. I getting tree builder test failures again |
| 10:34 | <Philip`> | ("... [this algorithm] has been dubbed the "adoption agency algorithm" (in contrast with other possibly algorithms for dealing with misnested content, which included the "incest algorithm", the "secret affair algorithm", and the "Heisenberg algorithm").") |
| 10:34 | <hsivonen> | presumably, there has been a reason for the tests to change |
| 10:35 | <hsivonen> | daytime soap opera algorithm |
| 10:35 | <MikeSmith> | the "red-headed stepchild" algorithm |
| 10:37 | <MikeSmith> | the red-headed stepchild metaphor is maybe a good metaphor for how some view the work on spec'ing text/html and Web apps |
| 10:38 | hsivonen | doesn't like hair-colorist algorithm names |
| 10:38 | <hsivonen> | http://pastebin.ca/1022593 |
| 10:38 | <hsivonen> | those are the errors I'm getting |
| 10:43 | <hsivonen> | hmm. I've made marquee scoping... |
| 10:43 | <hsivonen> | but still </div> closes the marquee... |
| 10:43 | <hsivonen> | hmm... |
| 10:45 | <MikeSmith> | I have a request that if somebody here (Philip`? or of jgraham is listening) who has commit access to the html5lib sources could consider, I'd appreciate it |
| 10:46 | <MikeSmith> | some time last year, Mark P. commented out this line in html5parser.py |
| 10:47 | <MikeSmith> | sys.stderr.write("Warning: Undefined behaviour for start tag %s"%name) |
| 10:47 | <MikeSmith> | but for whatever reason didn't comment out this one: |
| 10:47 | <MikeSmith> | sys.stderr.write("Warning: Undefined behaviour for end tag %s"%name) |
| 10:48 | <MikeSmith> | wondering of somebody could comment that out in the source |
| 10:48 | <MikeSmith> | or at least put at "\n" on the end of it if it's kept in |
| 10:53 | <Philip`> | MikeSmith: Is there an easy way to make that message get printed? |
| 10:54 | <MikeSmith> | Philip`: put a <section> element into your source |
| 10:54 | <MikeSmith> | or any other new element that Hixie hasn't yet added to the parsing algorithm |
| 10:54 | <MikeSmith> | e.g., <header> |
| 10:55 | <Philip`> | Ah, yes |
| 10:56 | <Philip`> | MikeSmith: Committed |
| 10:56 | <MikeSmith> | Philip`: cheers |
| 10:56 | <Philip`> | (since it looks like http://canvex.lazyilluminati.com/html5lib/changeset/988 just forgot about the end tag case) |
| 10:57 | <MikeSmith> | Philip`: I guess I can actually see some value in having that message emitted, even for non-debugging purposes |
| 10:58 | <MikeSmith> | but for consistency at least, if it's already commented out for the start-tag case, seems it is right to have it out for the end-tag case |
| 10:58 | <hsivonen> | hmm. I have a string interning bug with special legacy names like marquee and listing |
| 10:58 | <Philip`> | MikeSmith: I'm not sure I see the value, since there's nothing a user of the library can do about the issue |
| 10:59 | <Philip`> | (and since it parses pretty much like the user would expect) |
| 10:59 | <MikeSmith> | Philip`: OK |
| 11:00 | <Philip`> | But someone could point out the value to me :-) |
| 11:01 | <hsivonen> | so is there a one-stop place for finding all the parsing algorithm-sensitive obsolete elements? |
| 11:05 | <Philip`> | hsivonen: Like extracting all the element name strings listed in the spec? |
| 11:05 | <hsivonen> | Philip`: yeah |
| 11:05 | <hsivonen> | Philip`: I extracted all the names that the non-parsing part of the spec defines as conforming |
| 11:06 | <hsivonen> | now I manually added marquee, nobr, wbr, listing and plaintext because those failed test cases |
| 11:09 | hsivonen | landed a huge pile of refactored and new parser code |
| 11:10 | <Philip`> | hsivonen: applet, basefont, bgsound, big, center, dir, font, frame, frameset, hidden, image, isindex, listing, marquee, nobr, noembed, noframes, plaintext, s, spacer, strike, tt, u, wbr, xmp |
| 11:10 | <hsivonen> | Philip`: thank you |
| 11:11 | <Philip`> | are the things I see as /"[a-z0-9]+"/ in the tree constructor spec as of some months ago that are not in the list of conforming elements |
| 11:11 | <Philip`> | (though I'm not totally positive I didn't accidentally miss any) |
| 11:12 | <hsivonen> | "hidden"??? |
| 11:12 | <Philip`> | Uh |
| 11:12 | <hsivonen> | is that a false positive on the attribute? |
| 11:12 | <Philip`> | Yes |
| 11:13 | <Philip`> | I removed most of the false positives but missed that one |
| 11:13 | <hsivonen> | ok. thanks |
| 11:14 | <hsivonen> | Philip`: thanks. my list lacked bgsound and noembed still |
| 11:15 | <hsivonen> | fortunately, the additions didn't break my magic hash function |
| 11:16 | Philip` | sees that the grass in his garden is now significantly taller than a cat lying down |
| 11:16 | jwalden | wonders why freenode just dropped like a rock |
| 11:16 | hsivonen | didn't notice freenode dropping |
| 11:16 | <Philip`> | jwalden: Just looked like a minor netsplit to me |
| 11:16 | <Philip`> | 11:16 -!- Netsplit wolfe.freenode.net <-> irc.freenode.net quits: mpt, jwalden, weinig |
| 11:16 | <jwalden> | interesting |
| 11:17 | <jwalden> | from my pov everyone else dropped but those three :-) |
| 11:17 | <mpt> | we're special |
| 11:18 | Philip` | wonders if IRC networks all use a spanning tree between servers, rather than something more robust |
| 12:05 | <zcorpan> | othermaciej: http://software.hixie.ch/utilities/js/live-dom-viewer/?%3Chtml%20test%3Da%3E%3Cscript%3Edocument.documentElement.setAttributeNS(''%2C'test'%2C'b')%3C%2Fscript%3E |
| 12:05 | <zcorpan> | othermaciej: results in duplicate attributes in webkit |
| 12:07 | <zcorpan> | for some reason i thought the dom didn't check for duplicate attributes with different prefixes |
| 12:09 | <othermaciej_> | zcorpan: if you make the namespace parameter null it works, so the bug seems to be treating empty string and null differently |
| 12:09 | <othermaciej> | "works" == adds the attribute only once |
| 12:18 | <Philip`> | Hmm, I have a site which uses TinyMCE so non-technical users can edit bits of content, and I've been asked to remove the part of the TinyMCE image upload dialog that asks for alt text because it's too confusing for users |
| 12:20 | <hsivonen> | argh. my token name list is lacking the ruby stuff |
| 12:21 | <jmb> | Philip`: is that merely an interface issue, or is the concept of alt text causing the confusion? |
| 12:22 | <Philip`> | jmb: I'm not quite sure what the distinction is - do you mean could the problem be fixed by changing the UI to still ask for alt text but in a different way (as opposed to fixing the interface to simply not ask for it at all)? |
| 12:23 | <jmb> | that's the distinction I'm interested in, yes |
| 12:25 | <zcorpan> | Philip`: perhaps there should be a button "No images view" which replaces all images with editable boxes |
| 12:25 | <Philip`> | I can't think of any way the UI could ask for alt text without having to explain to users the concepts like the purpose of it and why they should bother, and I can't think of any way to explain that without confusing or scaring the users who are just wanting to stick an image in there |
| 12:25 | <zcorpan> | with an explanation that search engines and blind people will see that text instead of the images |
| 12:26 | <Philip`> | zcorpan: That sounds too hard to implement - I can cope with changing dialog boxes, but I don't want to fiddle with the TinyMCE code in any non-trivial way |
| 12:26 | <hsivonen> | Philip`: I think a big problem with the whole alt debate is wanting to build tool UI for it but not wanting to explain to users what it is about |
| 12:27 | <hsivonen> | which is why the V.nu Image Report actually mentions accessibility in the UI text |
| 12:27 | <zcorpan> | Philip`: but i think it would be better UI than dialog boxes |
| 12:28 | <Philip`> | (In this case, it's unlikely that many people will ever view the content that was created, and the creators usually won't care about search engines or blind people - they just get the satisfaction of making their own text appear on the site) |
| 12:29 | <zcorpan> | if they don't care then forcing them will produce bad alt text anyway |
| 12:29 | <hsivonen> | argh. the SVG stuff requires storing both the lowercase name and the case-corrected name on the stack |
| 12:29 | <hsivonen> | bad |
| 12:29 | <Philip`> | zcorpan: My definition of "better" includes "possible to implement without excessive effort" :-) |
| 12:29 | <Philip`> | hsivonen: Can you store just the case-corrected name and use case-insensitive comparison operators in appropriate places? |
| 12:29 | <hsivonen> | Philip`: that would kill perf |
| 12:30 | <hsivonen> | Philip`: actual tree builders won't need to keep the case-corrected string around, but to implement streaming SAX, it has to stay around |
| 12:31 | <hsivonen> | so another round of changing interfaces it is... |
| 12:31 | <MikeSmith> | hsivonen: while recognizing the general principle of trying to avoid additional options/pragmas whenever possible, I continue to think a pretty effective UI for the specific alt case (and perhaps only the alt case) would be an option saying, "Don't show me error messages about missing alt attributes." |
| 12:31 | <zcorpan> | hsivonen: would making the tokenizer case-preserving and having the parser do case-insensitive checks kill perf? |
| 12:32 | <hsivonen> | zcorpan: I'm not sure |
| 12:32 | <hsivonen> | zcorpan: probably not if the tree builder dispatches on magic enums instead of strings |
| 12:33 | <hsivonen> | currently, mine dispatches on interned string identity |
| 12:33 | <hsivonen> | but I want to change it to use magic numbers instead |
| 12:33 | <hsivonen> | I'm just unsure if today is the right time to do it |
| 12:33 | <hsivonen> | that is, will Hixie change stuff in a way that requires me to reassign the magic numbers |
| 12:34 | <zcorpan> | probably :P |
| 12:38 | <Philip`> | That sounds like a sensible answer whenever the question is "will Hixie cause pain to an implementor like me" |
| 12:39 | <hsivonen> | that is, are there elements that are safe to coalesce into one enum value? |
| 12:40 | <hsivonen> | td and th seem to always go together |
| 12:40 | <hsivonen> | and thead, tbody and tfoot |
| 12:40 | <hsivonen> | but are those all? |
| 12:40 | <hsivonen> | oh and h1-h6 |
| 12:40 | <Philip`> | What about things like b and i? |
| 12:41 | <hsivonen> | now I have to run some regexps on my code to find out |
| 12:41 | <Philip`> | Or: If you coalesce h1-h6, how do you handle <h1>foo</h6> ? |
| 12:41 | <hsivonen> | Philip`: I'm thinking about coalescing what the spec always has under one case |
| 12:42 | <Philip`> | Ah |
| 12:42 | Philip` | could probably derive that automatically from his OCamlified version of the spec, except it's old and incomplete and currently broken :-( |
| 12:43 | <hsivonen> | I could derive it almost automatically if I hadn't forgotten Perl as much as I have :-( |
| 12:44 | <Philip`> | TIMTOWTDI, so if you forget several ways then there's usually still another way to solve any problem with Perl |
| 12:47 | <Dashiva> | Philip`: And then you have two problems. |
| 12:49 | <Philip`> | Dashiva: That's what one-liners are for - once you've solved your original problem using Perl, you just discard the Perl into the murky depths of your .bash_history and it's no longer a problem |
| 12:50 | <Lachy> | LOL, it's funny when comment spammers try to post on-topic comments to the whatwg blog, yet fail miserably :-) |
| 12:50 | <Lachy> | this is in the moderation queue for the Vim article. "Your article is very interesting, Checking means verify the file. Thank you for giving this information." |
| 12:53 | <Philip`> | Blogs should just not accept URLs in comments at all |
| 12:53 | <Dashiva> | It's too bad spam bot authors have no sense of humor, or they could use "Your ideas are intriguing and I wish to subscribe to your newsletter." |
| 12:53 | <Philip`> | (including not having an explicit "website" field for commenters) |
| 12:58 | <zcorpan> | Philip`: then spammers will just write plain text URLs |
| 12:59 | <zcorpan> | Philip`: and people who actually want to link to relevant things it will just be annoying to deal with plain text URLs |
| 12:59 | <Dashiva> | We'll avoid the whole target/window.open spectacle, though :) |
| 13:12 | <zcorpan> | what Julian Reschke suggests for xhr seems to go against the priority of constituencies principle |
| 13:13 | <zcorpan> | http://www.w3.org/mid/48314BE6.9060708⊙gd |
| 13:13 | <hsivonen> | sigh |
| 13:20 | <hsivonen> | Hixie: any particular reason why pre and listing don't go together when breaking out of foreign lands? |
| 13:21 | <zcorpan> | hsivonen: because pre is common and listing is not? |
| 13:21 | <hsivonen> | yeah. I'm going to send email |
| 13:23 | <Lachy> | Philip`, URLs and email addresses provide a convenient mechanism for giving automatic approval or rejection of repeat commenters, based on their prior comment history |
| 13:23 | <Lachy> | it also allows commenters to be reliably identified and help them become known if they're seen to consistently write good comments |
| 13:24 | <Philip`> | Lachy: But I could trivially write a comment and use someone else's email address and URL |
| 13:24 | <Lachy> | surprisingly, few spammers actually seem to do that |
| 13:24 | <Philip`> | so "reliably identified" seems untrue |
| 13:25 | <Lachy> | although there are times where trolls will falsely post as someone else, it doesn't happen frequently enough to be of great concern |
| 13:25 | <Lachy> | I only know it to have happened once to me in all my years of reading and posting comments on blogs |
| 13:26 | <zcorpan> | readers should be able to report bad comments |
| 13:26 | <zcorpan> | i think that results in less overhead to deal with comments |
| 13:26 | <Philip`> | Then spammers will start spamming the 'report bad comment' feature |
| 13:27 | <Lachy> | rating systems like that are used on slashdot, youtube, digg, etc. |
| 13:27 | <Lachy> | They're not entirely successful |
| 13:27 | <Philip`> | Some much less successful than others |
| 13:27 | <zcorpan> | ok |
| 13:27 | <Philip`> | (The quality of comments on Slashdot does seem somewhat higher than on Youtube) |
| 13:28 | <Lachy> | I'm surprised slashdot has done little to actually filter out the goatse spam. |
| 13:28 | <Lachy> | Philip`, that's because the intelligence of slashdot readers is probably, on average, greater than that of your average youtube user |
| 13:30 | <Philip`> | Lachy: It seems to have pretty much died out by itself, just by being modded to -1 all the time so normal users will never see it |
| 13:31 | <Lachy> | I always have all comments shown. I don't like comments being automatically hidden, since it messes up threading and makes me wonder what I'm missing out on |
| 13:32 | <Philip`> | You're missing out on the goatse spam :-p |
| 13:32 | <Lachy> | sometimes the troll comments are funny |
| 13:38 | hsivonen | identified 52 magic enum items for the tree builder |
| 13:38 | <hsivonen> | of which 14 group more than one element name |
| 13:46 | <annevk> | implementing SVG for demonstration purposes? |
| 13:46 | <hsivonen> | annevk: I'm preparing to |
| 13:47 | <hsivonen> | annevk: I'm doing architectural changes that make the SVG case tweaking stuff not suck quite that badly for perf |
| 13:52 | <gsnedders> | Internet at home is broken :( |
| 13:53 | <Philip`> | Try your neighbours' internets |
| 13:54 | <gsnedders> | All require passwords (in terms of wireless, at least) |
| 13:54 | gsnedders | is in café in town |
| 13:55 | <Philip`> | They might be using WEP, which doesn't really count as a password ;-) |
| 13:55 | <gsnedders> | :) |
| 13:55 | <Lachy> | gsnedders, most people use very simple passwords. It shouldn't be too hard to guess, or even crack |
| 13:55 | <gsnedders> | Lachy: too much effort :( |
| 13:56 | <gsnedders> | it makes no sense, because I get to SHOWTIME, before it fails to get any further. |
| 13:56 | <Lachy> | is it less effort to walk to the cafe, than it is to take 5 minutes to crack a password? |
| 13:56 | <gsnedders> | And BT being typically BT are refusing to do anything about it and are blaming us. |
| 13:56 | <gsnedders> | Lachy: It means trying to crack the password. :P |
| 13:57 | <gsnedders> | Lachy: and it's only 50m away from school |
| 13:58 | <Philip`> | A couple of wireless routers around my house are unsecured and have login/password "admin"/"admin" |