| 00:51 | <Hixie> | annevk: cool, thanks |
| 01:40 | Hixie | summons zcorpan |
| 01:41 | <Hixie> | anyone have an opinion on whether <ol> is appropriate for lists that happen to be ordered as opposed to lists where the order is significant? |
| 01:41 | <Hixie> | e.g. should a list of place names, ordered alphabetically, be <ol> or <ul>? |
| 01:42 | <tantek> | Hixie, example of each? |
| 01:42 | <tantek> | <ol> - since "alphabetically" could mean any number of a different canonical sorting orders given variations in language/culture/punctuation treatment. |
| 01:43 | <Hixie> | hm, interesting |
| 01:43 | <tantek> | no way to really automatically discern how such ordering was done if in a <ul> and thus if it has significance or not. |
| 01:43 | <Hixie> | you used the opposite argument to zcorpan in his feedback in 2006 |
| 01:43 | <Hixie> | :-) |
| 01:43 | <tantek> | I've learned more about alphabetic sorting since then. |
| 01:44 | <Hixie> | no i mean he argued the opposite of you just now, in his mail from 2006 |
| 01:44 | <Hixie> | he says: |
| 01:44 | <Hixie> | > I think <ol> is a list where the order is significant to the meaning; |
| 01:44 | <Hixie> | > where the order is emphasized. For lists that happen to be ordered but |
| 01:44 | <Hixie> | > the order isn't really of a big significance or isn't of higher |
| 01:44 | <Hixie> | > significance than the global order of the document, <ol> shouldn't be |
| 01:44 | <Hixie> | > used IMHO. |
| 01:44 | <Hixie> | > Otherwise people might use <ol> whenever a list happens to be in order, e.g. |
| 01:44 | <Hixie> | > an A-Z list... |
| 01:44 | <tantek> | One way to think about it is, if the author considers the ordering to be something they did deliberately (whether by hand in the markup), then the way to communicate "deliberate ordering" is by using an <ol> |
| 01:45 | <tantek> | (whether by hand in the markup, or by selection of some backend sorting option in a query) |
| 01:45 | <Hixie> | so <ul> elements must be in effectively random order? |
| 01:45 | <tantek> | not necessarily |
| 01:45 | <tantek> | randomness isn't necessary |
| 01:46 | <Hixie> | right, i mean effectively random, as opposed to necessarily truly random |
| 01:46 | <tantek> | If the author considers the order to be arbitrary (not necessarily random), or doesn't care about the order, then a <ul> makes sense. |
| 01:46 | <tantek> | <ol> = author cared about the order. <ul> = author didn't care about the order. That simple. I think. |
| 01:46 | <Hixie> | i suppose a list of place names that were arbitrarily ordered alphabetically as opposed to ordered alphabetically for any particular reason would mean <ul> then |
| 01:48 | <tantek> | again, that assumes that a universal canonical alphabetic ordering exists that everyone implicitly understands, which is a false assumption. |
| 01:48 | <Hixie> | how do you mean? |
| 01:49 | <tantek> | alphabetic ordering is not just by characters per unicode code points for example |
| 01:50 | <Hixie> | sure. i mean, the ordering could be anything, i was just using "alphabetical" as an arbitrary ordering |
| 01:50 | <tantek> | e.g. depending on the context of the data, say it were English names of works of art (like movies or songs) |
| 01:51 | <Hixie> | for example the list could be in order of the population size per place, but if that was just an arbitrary order selected by the author simply for presentational reasons, that would be a <ul> |
| 01:51 | <tantek> | often a leading word "The" is omitted from the canonical sorting |
| 01:51 | <tantek> | there are numerous other such rules. and that's just for English. names of works of art. |
| 01:51 | <tantek> | so there are language specific rules. domain specific rules. etc. |
| 01:51 | <Hixie> | whereas if the list was a list of places, ordered by population size to show the differences in population sizes, <ol> would be correct |
| 01:52 | <tantek> | I think for whatever reason, if the author intends an order, then it should be marked up with <ol> |
| 01:52 | <tantek> | One thought experiment you could do is, if the browser scrambled the set of <li>s inside a <ul>, how upset would the author be when viewing the page? |
| 01:53 | <Hixie> | right |
| 01:53 | <tantek> | so a list of places, ordered by name (or population), that would be bad |
| 01:53 | <tantek> | but a grocery shopping list, not a problem |
| 01:53 | <Hixie> | so if i'm just listing places i've been, and i happen to have put them in order of population size, i don't care if the browser reorders them to be by order of how much rainfall they get per year |
| 01:54 | <tantek> | right, if the author didn't take care with the order, then a <ul> - hence my reasoning above |
| 01:54 | <tantek> | "for whatever reason, if the author intends an order, then it should be marked up with <ol>" |
| 01:54 | <tantek> | otherwise use <ul> |
| 01:54 | <Hixie> | but if i have a grocery list in order of importance, or in order of which aisle the stuff is found in, then i'd want <ol> |
| 01:54 | <Hixie> | well |
| 01:54 | <Hixie> | i'm saying the author might "intend an order" but not care about it at the same time |
| 01:54 | <tantek> | right, but I have yet to see such a grocery list ;) |
| 01:55 | <tantek> | then intend an order wins out |
| 01:55 | <tantek> | the intent is enough to justify an <ol> |
| 01:55 | Hixie | often puts grocery lists in order of importance, so that he can make sure to get the important stuff first, since he is limited by volume when cycling from the shop :-) |
| 01:55 | <Hixie> | hmm |
| 01:55 | <tantek> | the intent doesn't have to be "big" or "significant" - because who decides that? |
| 01:55 | <Hixie> | so zcorpan is saying that intent doesn't win out, but that care does |
| 01:56 | <Hixie> | i wonder which is more useful |
| 01:56 | <tantek> | given that authors are the ones writing the markup, I find that going with respecting/modeling author intent usually produces better results. ;) |
| 01:56 | <tantek> | rather than some academic argument about how "significant" the ordering is |
| 01:57 | <Hixie> | well, the default rendering of <ul> is bullets, and i'd want bullets if i gave a list of places i'd been to, even if i ordered them (arbitrarily) by population size or lexically or by rainfall |
| 01:57 | <Hixie> | so that would argue against what you're saying :-) |
| 01:57 | tantek | puts shopping lists into clusters by type of store. two levels of unordered lists. |
| 01:58 | <tantek> | the English "i'd want bullets if i gave a list of places i'd been to, even if i ordered them" translates to the CSS: ol.places { list-style:disc } |
| 01:58 | <tantek> | because that "want" is presentational |
| 01:58 | <Hixie> | sure, but authors aren't going to think that way |
| 01:58 | <Hixie> | they'll just use <ul> |
| 01:59 | <Hixie> | and pretend that they didn't order the items |
| 01:59 | <tantek> | that's fine, then they are saying the order doesn't matter |
| 01:59 | <Hixie> | and that they just _happen_ to be in alphabetical order |
| 01:59 | <Hixie> | right -- the order doesn't matter, even though there is one intended |
| 01:59 | <Hixie> | they don't care about the order, even though they intended one |
| 02:00 | <tantek> | no. if they are pretending they didn't order the items, then they didn't order them. |
| 02:00 | <tantek> | we can't really assume otherwise. |
| 02:00 | <Hixie> | <ul> <li> Apple <li> Banana <li> Cherry </ul> |
| 02:00 | <Hixie> | i ordered that list. |
| 02:00 | <tantek> | in this case no care = no intent. |
| 02:00 | <Hixie> | that i'm using <ul> doesn't change the fact that i did indeed give an order |
| 02:01 | <Hixie> | now i might not care if the browser changes the order |
| 02:01 | <Hixie> | but i still intended one |
| 02:02 | <tantek> | so is the distinction then an intent of default presentational order vs. a semantic order? |
| 02:02 | <tantek> | because one could argue your ordering of that list was purely for presentational purposes. that is, you only intended it for presentational purposes, not for anything semantic. |
| 02:03 | <Hixie> | right |
| 02:03 | <Hixie> | i ordered it just to make it easier to scan |
| 02:04 | <tantek> | ok easy enough to modify my previous statement to take that into account: |
| 02:04 | <tantek> | "for whatever reason, if the author intends an order with some meaning behind it beyond just the presentation, then it should be marked up with <ol>" |
| 02:12 | <Hixie> | ok, i've tried to update the spec to explain this, along with some examples. http://www.whatwg.org/specs/web-apps/current-work/multipage/section-lists0.html |
| 02:13 | tantek | clicks and hope his browser doesn't lock up. |
| 02:14 | <tantek> | BTW, really like the *zero* or more. About darn time. |
| 02:14 | <tantek> | do you need to say "change the meaning of the document"? |
| 02:14 | <tantek> | wouldn't just "change the meaning." be sufficient ? |
| 02:14 | <Hixie> | change the meaning of what? |
| 02:15 | <tantek> | per English implied object resolution rules |
| 02:15 | <tantek> | the previous clause |
| 02:15 | <tantek> | "changing the order" |
| 02:15 | <tantek> | also implied object |
| 02:15 | <tantek> | see previous clause |
| 02:15 | <tantek> | "where the items have been intentionally ordered" |
| 02:15 | <tantek> | also implied object |
| 02:15 | <tantek> | see previous clause |
| 02:16 | <tantek> | "a list of items" |
| 02:16 | <tantek> | resolution complete |
| 02:16 | <tantek> | no need to bring "the document" into it. |
| 02:16 | <Hixie> | hmmm |
| 02:16 | <Hixie> | i _would_ like to remove the document from that setence |
| 02:16 | <tantek> | doesn't everyone read/write English like code? |
| 02:17 | <tantek> | ;) |
| 02:17 | <othermaciej> | the way I think of it is <ol> --> numbered list, <ul> --> bulleted list |
| 02:17 | <othermaciej> | I would expect that is the typical author's operational understanding as well |
| 02:17 | tantek | smacks othermaciej for presentational-major thinking. |
| 02:18 | <Hixie> | othermaciej: i agree, but i think the current definition pretty much matches that |
| 02:18 | <tantek> | what do bullets sound like? |
| 02:18 | <Hixie> | about 100ms of silence |
| 02:18 | <othermaciej> | in normal English speech, nothing, they are just separators |
| 02:18 | <othermaciej> | to a screen reader, I dunno |
| 02:18 | <tantek> | othermaciej, that's why presentational-major thinking = FAIL. |
| 02:19 | <tantek> | Hixie, I suggested ditching the "... of the document." clause from both the <ul> and <ol> sections. |
| 02:19 | <Hixie> | "the meaning" on its own just doesn't sound right |
| 02:19 | <tantek> | Hixie, "sounding right" is irrelevant. It's a spec. The question is, does it parse? |
| 02:19 | <othermaciej> | tantek: I don't see how your argument addresses presentational thinking - it might be an argument against ever using bulleted lists, but I wouldn't buy that conclusion |
| 02:19 | <Hixie> | i'm leaving "of the document" until I (or someone else) can think of something better to replace it with |
| 02:19 | <othermaciej> | anyway, let me give a practical example |
| 02:19 | <tantek> | Hixie, empty space. |
| 02:19 | <Hixie> | tantek: actually i consider it sounding right to be of paramount importance to me :-) |
| 02:19 | <othermaciej> | sometimes, on the webkit blog, I post lists of new features or bug fixes |
| 02:20 | <othermaciej> | usually, I use <ul> for this since they are in no particular order, and mentally speaking, bullets are appropriate |
| 02:20 | <othermaciej> | on one occasion, I made a point of the fact that these were 10 features of particular interest |
| 02:20 | <othermaciej> | and in that case I used <ol> |
| 02:20 | <tantek> | Hixie, if you prefer to place a redundant object resolution place that's ok too, and English depends on redundancy for error correction. |
| 02:20 | <tantek> | e.g. s/of the document/of the list of items |
| 02:20 | <othermaciej> | I think those were both correct choices but I'd be hard-pressed to argue there is a deep semantic difference |
| 02:21 | <tantek> | since that "list of items" is what the object resolution demonstrated above |
| 02:21 | <tantek> | that should address the "sounding right" concern. |
| 02:21 | <Hixie> | othermaciej: well, if you have any suggested improvement to the text of the spec as it was just changed, i'm certainly open to it |
| 02:22 | <Hixie> | tantek: i'm not convinced people would understand what it meant if i did that |
| 02:22 | <Hixie> | it's not just the list that would change meaning |
| 02:22 | <Hixie> | it's the list plus any content referring to that list |
| 02:22 | <Hixie> | e.g. see the examples |
| 02:22 | <Hixie> | especially for <ol> |
| 02:22 | <othermaciej> | to be honest I didn't read the spec, I just wanted to point out that something based on subtle distinctions of how important or essential an order is would not match the mental model of the typical author |
| 02:23 | <Hixie> | othermaciej: i'm not sure defining them as being "when you want bullets" or "when you want numbers" would work either, though |
| 02:24 | <othermaciej> | I think the distinction between <ol> and <ul> is not intrinsic to the data, it's about whether the author wants to specify and emphasize the order |
| 02:24 | <othermaciej> | depending on authorial intent, you could meaningfully do either for the same list in the same context |
| 02:24 | <Hixie> | othermaciej: in your example of numbering them to emphasise the number of items, you are not emphasising the order. |
| 02:24 | <othermaciej> | for example, "top 10" lists are pretty arbitrary in their order, many could just as easily be "10 things" lists |
| 02:25 | <Hixie> | indeed |
| 02:25 | <Hixie> | so that's not emphasising the order |
| 02:25 | <othermaciej> | fair enough |
| 02:25 | <othermaciej> | although if the spec says that in that situation you should use numbers in the text content or <ul> with CSS counter styling, I am not sure it will be followed |
| 02:26 | <othermaciej> | (also not sure that it would ever be an improvement to do that) |
| 02:26 | <Hixie> | i would say that in those contexts the current definition works better |
| 02:26 | <Hixie> | the current definition is that the order matters |
| 02:26 | <Hixie> | and if you refer to "number 5", clearly the order matters |
| 02:26 | <Hixie> | since changing the order would change what each one was numbered |
| 02:26 | <othermaciej> | that's true |
| 02:26 | <Hixie> | <ol> is now defined as "The ol element represents a list of items, where the items have been intentionally ordered, such that changing the order would change the meaning of the document." |
| 02:26 | <othermaciej> | sometimes the numbers aren't there for order, but to provide referents for external references |
| 02:26 | <Hixie> | right |
| 02:27 | <othermaciej> | (which might not even be in the same document) |
| 02:27 | <othermaciej> | that seems like a good enough definition |
| 02:27 | <Hixie> | the current definition just says they've been intentionally ordered, not that they are ordered according to any grand scheme |
| 02:27 | <Hixie> | i like it |
| 02:27 | <Hixie> | yeah |
| 02:27 | <Hixie> | it has withstood your attempt at disproving it :-P |
| 02:28 | <othermaciej> | I wasn't trying to disprove it |
| 02:28 | <othermaciej> | I hadn't even read it at the time I commented |
| 02:28 | <othermaciej> | I was just reading scrollback and thought it sounded a bit abstract |
| 02:28 | <Hixie> | i know, i'm just teasing :-) |
| 02:30 | <othermaciej> | if I were trying to disprove something, you'd know it :-) |
| 02:30 | <Hixie> | hah |
| 03:25 | Hixie | adds more examples and their hidden meanings to the spec |
| 04:02 | <Hixie> | hm |
| 04:02 | <Hixie> | <ol reversed> is an interesting proposal |
| 04:02 | <Hixie> | i wonder how the browser vendors would react to it |
| 04:24 | Hixie | bcc's 50-odd people |
| 04:25 | <Hixie> | ok that was a reply to 100+ e-mails |
| 04:25 | <Hixie> | yay |
| 04:26 | <Hixie> | what's next... maybe section/p/hr stuff... maybe em/strong stuff... |
| 04:26 | <Hixie> | hm |
| 04:27 | <tantek> | Hixie, rather than reversed, why not allow a list-increment? |
| 04:27 | <tantek> | default: +1 |
| 04:28 | <tantek> | -1 to achieve reverse |
| 04:29 | <Hixie> | reversed does more than just change the step increment |
| 04:29 | <Hixie> | it also changes the start value |
| 04:29 | <Hixie> | we can add step later if there's a use case for it |
| 04:29 | <tantek> | sure but there already is a start attr |
| 04:29 | <Hixie> | right but reversed changes the default start value |
| 04:29 | <tantek> | the use case is lists that are numbered in large increments |
| 04:29 | <tantek> | see patent docs |
| 04:29 | <tantek> | which often use increments of 2 or 10 |
| 04:29 | <Hixie> | they can use value="" |
| 04:29 | <tantek> | for paragraphs etc |
| 04:30 | <Hixie> | decrementing to 1 is a much more common case |
| 04:30 | <tantek> | they can't often use "value" because that requires too much fix-up |
| 04:30 | <tantek> | for the same reason we don't ask all numbered lists to use "value" |
| 04:30 | <tantek> | on every list item |
| 04:30 | <Hixie> | the whole point of large numbers in those cases is that you don't need fixup to add an additional paragraph |
| 04:31 | <Hixie> | hence why value="" is the most appropriate |
| 04:31 | <Hixie> | (if you really want to use a list in those cases) |
| 04:32 | <tantek> | people will just use hacks then to get auto-numbered by +10 lists |
| 04:32 | <tantek> | like <li>...</li> then 9 empty <li> etc. |
| 04:32 | <Hixie> | i haven't seen that happen |
| 04:32 | <tantek> | then CSS to only have every 10th item show up |
| 04:32 | <Hixie> | i have seen peple use hacks for stepping down to -1, though |
| 04:32 | <Hixie> | (namely using scripts to do it) |
| 04:33 | <tantek> | yeah, i've seen them, but can't find any examples offhand |
| 04:33 | <Hixie> | see the e-mail i just sent whatwg for the research that was done to back up reversed="" -- it wasn't as extensive as with other cases like some of the research i've done, but it was eye-opening nonetheless |
| 04:33 | <tantek> | if i do i'll let you know |
| 04:33 | <Hixie> | cool, thanks |
| 04:35 | <Hixie> | lord, microsoft adcenter is trying to hire me |
| 04:35 | Hixie | informs them politely that he's happy working for the company that's whipping their asses already |
| 04:37 | <tantek> | Hixie, I do also remember running into the need to do reverse lists in the past as well, and having to resort to "value" manually to do it in my own markup. |
| 04:45 | <othermaciej> | Hixie: they probably got you from a list of people working for the company that's whipping their asses |
| 04:52 | <tantek> | question: is this a valid URL for this channel: irc://irc.freenode.net/whatwg ? |
| 05:04 | <takkaria> | tantek: I believe so |
| 05:05 | <tantek> | I wasn't able to use that "company that's whipping asses" to find an irc: URL validator. |
| 05:14 | <othermaciej> | tantek: there's two different expired drafts floating around |
| 05:14 | <othermaciej> | tantek: by the latest one, it should be irc://irc.freenode.net/#whatwg |
| 05:15 | <jruderman> | without a # works in more clients |
| 05:15 | <othermaciej> | or irc://irc.freenode.net/whatwg,ischannel |
| 05:16 | <othermaciej> | (not sure if anything implements either of those rules |
| 05:16 | <othermaciej> | ) |
| 07:27 | <Hixie> | ok here goes |
| 07:27 | <Hixie> | 63 elements on sections |
| 07:27 | <Hixie> | and related subjects |
| 07:27 | Hixie | beings replying |
| 07:32 | <othermaciej> | 63 elements? |
| 07:32 | <othermaciej> | or do you mean 63 emails? |
| 07:32 | <othermaciej> | hello roc |
| 07:33 | <Hixie> | er yes |
| 07:33 | <Hixie> | e-mails |
| 07:33 | <roc> | hello |
| 07:34 | <jruderman> | roc! |
| 07:43 | <Hixie> | hey roc |
| 08:28 | <virtuelv> | hsivonen: yt? |
| 08:29 | <hsivonen> | virtuelv: yes |
| 08:29 | <virtuelv> | do you have any plans on releasing the source for your validator? |
| 08:30 | <hsivonen> | virtuelv: it's already Free Software with a publicly readable svn repo |
| 08:30 | <hsivonen> | virtuelv: I don't have immediate plans to cut version numbered release packages |
| 08:30 | <virtuelv> | hsivonen: Hm. I totally missed that |
| 08:31 | <hsivonen> | virtuelv: http://about.validator.nu/#src |
| 08:31 | <virtuelv> | hsivonen: thanks |
| 08:32 | <virtuelv> | Would be helpful if there was a ToC on the about page |
| 08:36 | <hsivonen> | virtuelv: ok. I'll add one in the next update of that page that I have in preparation |
| 08:56 | <Dashiva> | Dmitry wants to hold votes on his proposals, this could get messy |
| 08:57 | <Lachy> | Hixie, the top 10 movie list example contains movies that really shouldn't be in the top 10, and is missing others that should. :-) |
| 08:57 | <Lachy> | also, there's no example of using the start attribute given |
| 08:57 | <Hixie> | pah |
| 08:57 | <Hixie> | (re the movies) |
| 08:57 | <Hixie> | and yeah |
| 08:57 | <Hixie> | i need to add more examples |
| 08:58 | <Hixie> | i just added examples for the things people asked about |
| 08:59 | <Dashiva> | Hixie: Maybe you should take a pointer from FORTRAN and start labeling your step sequences, that way it's easy to say which sequence to abort :) |
| 09:00 | <Lachy> | you need an example that shows <ol start=100 reversed>, and explain why long reversed lists should ideally include a start attribute to help with incremental rendering |
| 09:11 | <Hixie> | Dashiva: maybe |
| 09:11 | <Hixie> | Lachy: if you really think so, send mail :-) |
| 09:12 | Hixie | is well into a big e-mail on sections by now |
| 09:30 | <Lachy> | Hixie, I will later |
| 09:31 | <Hixie> | k |
| 09:34 | <hendry> | what is the (code) name for the IE7 engine? e.g. Firefox -> Gecko |
| 09:35 | <hsivonen> | hendry: Trident |
| 09:35 | <hsivonen> | hendry: since IE 4.0 |
| 09:37 | <hendry> | though aren't there more codenames as the engine in 6 is different to 7, if you know what i mean. i just re-read all that stuff that came up last month. |
| 09:37 | <hendry> | i should just re-read :) |
| 09:37 | <Hixie> | the code name for the IE engine is trident |
| 09:37 | <Hixie> | as far as know it hasn't got version-specific names |
| 09:38 | <madmoose> | Maybe for IE8 they'll rename it "Tridents". |
| 09:39 | <Hixie> | hah |
| 09:41 | jgraham_ | wonders if they'll fulfil nominative-determinism and stop shipping new engines once they reach three |
| 09:42 | <jgraham_> | s/new/more/ |
| 09:42 | <Hixie> | hm? |
| 09:42 | <jgraham_> | Trident == Three pronged, no? |
| 09:45 | <Hixie> | oh, that name |
| 09:45 | <Hixie> | i thought you meant IE8 |
| 09:45 | <Hixie> | and couldn't work out how 8 would mean 3 |
| 09:55 | <annevk> | DOM travelsal is complicated :( |
| 09:55 | <Hixie> | yeah |
| 09:55 | <annevk> | traversal, even |
| 09:55 | <Hixie> | you wanna fix it? |
| 09:55 | <annevk> | heh |
| 09:56 | <Hixie> | it needs an editor |
| 09:56 | <Hixie> | with a big hammer and a bag full of nails |
| 09:56 | <Hixie> | and a lot of duct tape |
| 09:56 | <annevk> | i've got quite a list of specs already |
| 09:57 | <annevk> | can you tell if http://tc.labs.opera.com/dom/traversal/002.htm is valid or not? |
| 09:58 | <annevk> | opera had some minor failures, but i'm not sure if my minimized test is ok |
| 09:58 | <Hixie> | i dunno |
| 09:58 | <Hixie> | it took me days to do the acid3 test that does mutations in the handler |
| 09:58 | <Hixie> | and i'm still not sure it's correct |
| 09:58 | <annevk> | ouch |
| 09:58 | <annevk> | that's the test i'm talking about :) |
| 09:58 | <Hixie> | :-) |
| 09:59 | <Hixie> | someone should write up the traversal spec unambiguously |
| 09:59 | <Hixie> | then we'd know |
| 10:01 | <jgraham_> | Hixie: You should produce a list of specs that you _don't_ think need an editor, as it would be simpler to refer to that than a list of specs that do |
| 10:01 | <jgraham_> | ;) |
| 10:04 | <annevk> | how did the moz call go? |
| 10:04 | <annevk> | cookies? |
| 10:07 | <Hixie> | jgraham_: no, i have a list of specs that need an editor |
| 10:07 | <Hixie> | jgraham_: it's relatively short |
| 10:08 | <Hixie> | jgraham_: (companion specifications on the whatwg wiki) |
| 10:08 | <jgraham_> | Hixie: I wasn't quite being serious :) |
| 10:08 | <Hixie> | :-) |
| 10:12 | Philip` | wondered why his web server had TCP connections with apache2-hixie.hixie.dreamhost.com, before finding that that's just what whatwg.org reverse-resolves into |
| 10:14 | <Philip`> | By the way, http://status.whatwg.org/annotate-web-apps.php doesn't look so good |
| 10:17 | Philip` | wonders if people who carefully make XHTML sites in PHP do something to disable/redirect warning/error messages like that, or if they just assume they're never going to occur |
| 10:22 | <zcorpan> | Philip`: what? php error messages are supposed to be xhtml compatible!! they changed <br> to <br />!1 |
| 10:35 | <Lachy> | Hixie, yt? |
| 10:36 | <Lachy> | Hixie, in the first note in the dl elmement section, you're missing the word "using" |
| 10:37 | <Lachy> | oh, actually, you wrote "using accidentally" instead of "accidentally using" |
| 10:56 | <Hixie> | thanks will fix |
| 11:06 | <Lachy> | Hixie, why do you refer to <dl> as an association list where it's defined, but then refer to description list everywhere else? |
| 11:11 | <Hixie> | carelessness, probably |
| 11:11 | <Hixie> | aw man, henri is asking for a rewrite of the outline algorithm |
| 11:11 | <Hixie> | and there i was thinking i'd get through all this feedback without having to do any work |
| 11:12 | <Hixie> | i guess i'll reply to this tomorrow |
| 11:12 | <annevk> | that must be a subset of his feedback then |
| 11:13 | <annevk> | lots of parsing questions stuff prolly requires work |
| 11:13 | <annevk> | s/stuff// |
| 11:14 | <zcorpan> | annevk: it seems you read "this" as "his" |
| 11:14 | <annevk> | yeah |
| 11:31 | <zcorpan> | the first example on <figure> reminds me of <listing> |
| 11:40 | zcorpan | notes that the 3rd example has <p/><img/><p/> |
| 11:41 | <annevk> | that reminds me of a frontpage commercial |
| 11:41 | <annevk> | http://annevankesteren.nl/2004/02/microsoft-frontpage-and-valid-html |
| 11:42 | <annevk> | "</p>That's right. We said FrontPage.</p>" |
| 12:12 | <zcorpan> | Hixie: you got a 9 too much, so it would be 5000 people, not 500 :) |
| 12:15 | <Philip`> | <pre><code> |
| 12:15 | <Philip`> | interface PrimaryCore { |
| 12:15 | <Philip`> | seems sort of wrong, since it'll have a blank line at the top |
| 12:15 | <Philip`> | (since <pre> just deletes a directly following newline character token, which doesn't happen in this case) |
| 12:16 | <Philip`> | Also it says void sendData(in sequence<byte> data); which looks wrong |
| 12:17 | <zcorpan> | Philip`: what wrong about it? |
| 12:17 | <Philip`> | Oh, whoops, nothing |
| 12:17 | <Philip`> | I forgot it was showing HTML code |
| 12:20 | <annevk> | the <aside> example that's also a self-reference contains a too early </pre> |
| 13:20 | Philip` | wonders if there's a web server anywhere which sends Content-Encoding: deflate, so he can test his code on it |
| 13:30 | <hsivonen> | Philip`: curiously, x-gzip (as opposed to gzip) happens |
| 13:37 | <hsivonen> | Philip`: when I implemented gzip support for V.nu, I figured that clients support gzip so well that supporting deflate and x-gzip is not worthwhile. |
| 13:37 | <hsivonen> | Philip`: it seems to me that deflate is mostly a dead letter in the spec |
| 13:38 | <hsivonen> | Philip`: so my guess is that your software doesn't need to support deflate |
| 14:05 | <Philip`> | Hmm, I suppose I might as well keep untested deflate support since it's only two lines of code and it'll let me see how many servers attempt to send deflate responses |
| 14:06 | <Philip`> | It looks like only Opera sends accept-encoding:x-gzip |
| 14:07 | <Philip`> | (and nothing sends 'compress') |
| 14:08 | <SadEagle> | konq does x-gzip as well. But no compress. |
| 14:08 | Philip` | adds x-gzip to see what happens |
| 14:09 | <Philip`> | SadEagle: Aha, okay |
| 14:09 | <Philip`> | Konqueror 3.5.something sends "x-gzip, x-deflate, gzip, deflate" |
| 14:10 | <Philip`> | but x-deflate seems a bit peculiar since it e.g. isn't mentioned in RFC2616 |
| 14:12 | Philip` | wonders how many people handle accept-encoding:gzip,identity;q=0 |
| 14:26 | <annevk> | If I have to boxes A and B. A is located at 0,0 and B at 10,10. I can now speak about the distance between the top edge of A and the top edge of B (being 10). Now if B is located at -10,-10 this distance should be -10. Is saying "The distance between the top edge of A and the top edge B" enough or should I say "downward distance" |
| 14:26 | <annevk> | suggestions? |
| 14:28 | <zcorpan> | if it can be negative then it's not a distance |
| 14:28 | <Philip`> | "The y coordinate of the top edge of A minus the y coordinate of the top edge of B"? |
| 14:28 | <annevk> | i guess that's best, yes |
| 14:29 | <Philip`> | Or skip the annoyingly verbose English and say A<sub>T</sub><sub>y</sub> - B<sub>T</sub><sub>y</sub> |
| 14:30 | <Philip`> | Actually, skip the annoyingly verbose HTML too and say $A_T_y - B_T_y$ |
| 14:30 | <annevk> | heh |
| 14:31 | <zcorpan> | Hixie: it should be attribute boolean reversed, not attribute long reversed |
| 14:32 | <zcorpan> | Hixie: s/is present/is absent/ in the paragraph defining reversed='' |
| 14:50 | <Philip`> | Oh, it turns out there's at least one site which sends 'delate' content, and it makes my code die with "unknown compression method" |
| 14:53 | <zcorpan> | 'delate'? |
| 14:53 | <Philip`> | Uh |
| 14:54 | <Philip`> | 'deflate' |
| 14:56 | <Philip`> | Aha, it works when I add a second parameter and do new InflaterInputStream(in, new Inflater(true)); |
| 15:04 | <Philip`> | In 1024 pages, I see 2 deflate and 141 gzip |
| 15:04 | <Philip`> | (when sending accept-encoding:gzip,deflate) |
| 15:17 | <Philip`> | In 8192, I see 12 deflate and 1195 gzip |
| 15:21 | <Philip`> | In 16384, 16 and 2439 |
| 15:21 | <Philip`> | deflate is not entirely negligible |
| 15:26 | <Philip`> | http://www.toua-u.ac.jp/ - "Video/X-Flv: .flv" - is that meant to do anything? |
| 15:30 | <Philip`> | http://www.superexpressgonzalez.com/ - P3P: policyref="http://www.tiendavirtual.ws/w3c/p3p.xml"; CP="NOI DSP COR NID PUB NOR" """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""" |
| 15:52 | <Philip`> | <!doctype php public "-//w3c//dtd xphp 1.0 transitional//en" "http://www.w3.org/tr/xphp1/dtd/xphp1-transitional.dtd">; |
| 15:53 | <Philip`> | <!doctype html public "guest house,jersey,haven,guest house,accommodation,jersey,channel islands,saint helier,british "> |
| 15:53 | <Philip`> | <!doctype html public "-//w3c//dtd xhtml 1.0 transitional//en" "http://www.w3.org/tr/xhtml1/dtd/xhtml1-transitional.dtd"; "http://www.w3.org/tr/html4/loose.dtd">; |
| 15:53 | <Philip`> | The world is a crazy place :-( |
| 15:54 | <annevk> | fortunately we have unambigious rules to interpret it |
| 15:55 | <zcorpan> | i've seen "php" doctypes before |
| 15:55 | <Philip`> | http://www.thermaglaze.com/ - <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd"; [ <!ATTLIST a target CDATA #IMPLIED> ]> - that's another way to work around the nonconformance of <a target> |
| 15:55 | <zcorpan> | people change all their files from .html to .php and do a global s/html/php/ |
| 15:56 | <zcorpan> | (which suggests there are some <php> elements out there, too) |
| 15:56 | <Philip`> | http://www.pervasive.com/ - <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 5.0//EN" /> |
| 15:57 | <Philip`> | <!doctype html public "–//w3c//dtd html 3.2//en"> - hmm, bad copy-and-paste |
| 15:57 | <Philip`> | (from Word, presumably) |
| 15:58 | <zcorpan> | or search-replace |
| 15:58 | <Philip`> | Oh, I suppose that's possible |
| 16:13 | <Philip`> | Hmm, IE6 and HTML5 (or non-IE browsers) seem to differ in doctype moding for ~1% of pages |
| 16:15 | <Philip`> | The most significant difference is <!doctype html public "-//w3c//dtd html 4.0 transitional//en" "http://www.w3.org/tr/rec-html40/loose.dtd">; (on ~0.5% of pages) |
| 16:17 | <zcorpan> | Philip`: do you have pointers to the remaining ~0.5%? |
| 16:23 | <Philip`> | 276 – <!doctype html public "-//w3c//dtd html 4.0 transitional//en" "http://www.w3.org/tr/rec-html40/loose.dtd">; |
| 16:24 | <Philip`> | 69 – <!doctype html public "-//"aol hometown//html 3.0 transitional//en"> |
| 16:24 | <Philip`> | 45 – <!doctype html public "-//w3c//dtd html 4.0 transitional//en" "http://www.w3.org/tr/html4/loose.dtd">; |
| 16:24 | <Philip`> | 34 – <!doctype html public "-//w3c//dtd xhtml 1.0 transitional//en" system "http://www.w3.org/tr/xhtml1/dtd/xhtml1-transitional.dtd"/>; |
| 16:24 | <Philip`> | 22 – <!doctype html public "-//w3c//dtd html 4.0 transitional//en" "http://www.w3.org/tr/html40/loose.dtd">; |
| 16:24 | <Philip`> | 19 – <!doctype html public "-//w3c//dtd html 4.0 frameset//en" "http://www.w3.org/tr/rec-html40/frameset.dtd">; |
| 16:24 | <Philip`> | zcorpan: ^ Those are the counts (out of 62592 pages) of the top doctypes that are IE standards and HTML5 quirks |
| 16:24 | Philip` | will upload some page with all the details later |
| 16:24 | <zcorpan> | Philip`: thanks! |
| 16:25 | <Philip`> | (These are all pages from dmoz.org, so there's still an uncorrected bias towards certain sites/domains) |
| 16:26 | <Philip`> | (Also, by "HTML5" I mean HTML5 plus http://lists.w3.org/Archives/Public/public-html/2008Jan/0006.html ) |
| 16:34 | Philip` | wonders how many pages he should download |
| 16:39 | <gsnedders> | how many implementations are there of RDF? |
| 16:41 | <Philip`> | I get quite a few "Corrupt GZIP trailer" errors |
| 16:45 | <Philip`> | <button onlick="..."> doesn't quite do what I want |
| 17:03 | <Philip`> | <!doctype html public "-//w3c//dtd xhtml 1.0 transitional//en" "http://www.w3.org/tr/xhtml1/dtd/xhtml1-transitional.dtd"; /> |
| 17:04 | <Philip`> | HTML5 say quirks, Firefox/Opera say limited-quirks |
| 17:04 | <Philip`> | *says |
| 17:05 | <Philip`> | (IE says limited-quirks too, and I can't be bothered to test Safari) |
| 17:10 | <zcorpan> | um... how come safari doesn't do quirks mode in the live dom viewer? |
| 17:11 | <Philip`> | How are you testing that? |
| 17:11 | <zcorpan> | <style>body{background:f00}</style> |
| 17:13 | <zcorpan> | safari seems to not care about garbage after the system identifier |
| 17:14 | <zcorpan> | so html5 is in disagreement with browsers about trailing /> in the doctype |
| 17:14 | <Philip`> | Safari seems to think "foo<!doctype html>" is non-quirks |
| 17:14 | <zcorpan> | mozilla too |
| 17:15 | <aroben> | zcorpan: Philip`: what version of Safari are you testing? |
| 17:15 | <zcorpan> | probably an old one |
| 17:16 | <zcorpan> | 3.0.2 (522.13.1) |
| 17:16 | <aroben> | ok |
| 17:16 | <aroben> | that's good |
| 17:16 | <aroben> | WebKit trunk just recently changed to match HTML 5's doctype parsing much more closely |
| 17:16 | <aroben> | I'm glad you're testing the behavior of Safari 3 |
| 17:16 | <Philip`> | I see 88 pages with <!doctype .../> out of 109434 |
| 17:16 | <zcorpan> | aroben: would be interesting to know if your changes breaks anything |
| 17:18 | <aroben> | zcorpan: yes, it would ;-) |
| 17:18 | <aroben> | zcorpan: I'm sure we'll find out eventually |
| 17:18 | <zcorpan> | i would also like to know if pages that use something other than //en in the FPI are rendered correctly in firefox/safari |
| 17:19 | gsnedders | is confused. how is the billion laughs attack that many? if 0 is 2 chars, then is repeated twice, it becomes four, then that is repeated twice you get eight, etc. only doubling. |
| 17:20 | <zcorpan> | gsnedders: double it enough times and it will become a billion pretty quickly :) |
| 17:20 | <gsnedders> | but with only 30 it gets to only 120, I thinks |
| 17:20 | <gsnedders> | that can't be right |
| 17:20 | gsnedders | is being dumb |
| 17:22 | <gsnedders> | doubling starting at two goes up in powers of two. |
| 17:22 | <gsnedders> | duh. |
| 17:23 | <gsnedders> | which means laugh30 is $2^{30+1}$ characters |
| 17:24 | <Philip`> | I see 8 pages with <!doctype html> out of 109434 |
| 17:31 | <annevk> | mine mine mine |
| 17:34 | <Philip`> | 135 pages have Server "ZX_Spectrum/1997 (Sinclair_BASIC)" |
| 17:35 | <SadEagle> | coolness. |
| 17:39 | <Philip`> | Oh, they're all subdomains of narod.ru |
| 17:39 | <Philip`> | so there's only one web server running on a Spectrum :-( |
| 17:58 | <Philip`> | http://philip.html5.org/data/doctypes-2.html |
| 17:58 | <Philip`> | (The "Test" thing probably doesn't work in WebKit) |
| 18:00 | <gsnedders> | yeah, it just reportd non-quirks for all |
| 18:01 | <gsnedders> | Philip`: <http://bugs.webkit.org/show_bug.cgi?id=15062>; |
| 18:14 | <hsivonen> | hrm. Google docs reject Firefox 3 and tells me to get Firefox 1.5.0.12 or newer |
| 18:15 | <svl> | Google does lots of bad sniffing for "Firefox" instead of looking at "Gecko" :( |
| 18:17 | <svl> | (And Firefox 3 at present is still "minefield" (I'm pretty certain, though didn't check)) |
| 18:19 | <aroben> | Philip`: here's a test for what mode the page is in that works in WebKit: http://trac.webkit.org/projects/webkit/browser/trunk/LayoutTests/fast/doctypes/resources/TestDoctype.js?rev=30431 |
| 18:29 | <gavin> | svl: nightly builds are "Minefield" |
| 18:29 | <gavin> | betas are "Firefox" |
| 18:29 | <hsivonen> | I tried with a beta, fwiw |
| 18:31 | <gsnedders> | svl: Firefox is a registered trademark of the Mozilla Foundation, and its use is massively limited |
| 18:32 | <gavin> | "massively limited"? limited to "Firefox", sure :) |
| 18:37 | <gsnedders> | gavin: limited to any official build |
| 18:37 | <zcorpan> | Philip`: i think the regular expression could have been <!doctype[^>]*> |
| 18:39 | <gsnedders> | Philip`: seems to have been merged into trunk by now |
| 18:40 | <zcorpan> | <!doctype html public "-//sq//dtd html 2.0 + all extensions//en" "hmpro3.dtd"> |
| 18:40 | <zcorpan> | hmm |
| 18:43 | <gsnedders> | Philip`: only diff in WebKit from HTML 5 is <!doctype html public "-//w3c//dtd html 4.01 transitional//en" ""> is quirks |
| 18:46 | <zcorpan> | opera seems to mostly match html5 on commonly used doctypes, but mostly match ie on uncommonly used doctypes |
| 18:47 | <zcorpan> | what i want to know is if changing our impl will break pages with the uncommonly used doctypes :) |
| 18:48 | <zcorpan> | if they already are broken in safari 3 and firefox, then it might be a good idea to revise html5 |
| 18:49 | <zcorpan> | (i.e. broken because they get the wrong mode) |
| 19:28 | <Philip`> | zcorpan: <!doctype[^>]*> probably would be entirely sensible, given that that's what HTML5 now expects |
| 19:29 | Philip` | will re-run it with that |
| 19:31 | Philip` | saw 39 application/xhtml+xml pages out of 100K (when sending the same Accept header as FF3) |
| 19:38 | <Philip`> | gsnedders: Do you mean the 'Test' button thing works correctly in WebKit now, so I don't need to change anything (like copying aroben's suggestion)? |
| 19:38 | <gsnedders> | Philip`: yeah |
| 19:38 | <Philip`> | gsnedders: Okay, thanks, that makes it easy for me :-) |
| 19:39 | <annevk> | http://tc.labs.opera.com/html/parsing/doctype/001.htm also tests DOCTYPE handling |
| 19:39 | <annevk> | fwiw |
| 19:43 | Philip` | wonders if he should make each doctype link to a list of all the pages which were found to use it |
| 20:21 | <Hixie> | Philip`: how is <!doctype html public "-//"aol hometown//html 3.0 transitional//en"> triggering HTML5 quirks? |
| 20:24 | <Philip`> | Hixie: After the second '"' it's in 'after DOCTYPE public identifier state', and the 'a' puts it into 'bogus DOCTYPE state', which ends up setting the correctness flag to 'incorrect', so it gets treated as quirks |
| 20:28 | <Hixie> | Philip`: oh, yeah, good point |
| 20:30 | <Philip`> | Oh, I just realised why my downloader was failing to terminate for one particular site |
| 20:30 | <Philip`> | That site is an infinitely long radio stream |
| 20:31 | <Hixie> | hah |
| 20:31 | <Hixie> | yeah |
| 20:31 | <Hixie> | i've come across infinite pages before too |
| 20:31 | <Hixie> | that's how i know there's an infinite number of <font> elements on the web |
| 20:32 | <gsnedders> | :D |
| 20:44 | <Philip`> | Why has my /bin/grep become thousands of times slower than it should be? |
| 20:47 | <Philip`> | and only on my XML file of extracted doctypes, not on any other file I've tested? |
| 20:48 | <Philip`> | Oh, it goes thousands of times faster again if I set LANG=en_GB without the .UTF-8 |
| 20:50 | <Philip`> | even though it's perfectly fast on a different computer with LANG=en_GB.UTF-8 |
| 20:51 | <gsnedders> | computers are odd. |
| 20:55 | <Dashiva> | Probably has to do with matching stuff like \w |
| 20:57 | <Philip`> | I was only doing "grep processed foo.xml" |
| 21:00 | <Dashiva> | Maybe it does some conversion to an internal representation or some other premature optimization anyway? |
| 21:01 | <Philip`> | <!DOCTYPE "=""" PUBLIC="PUBLIC" HTML="html" HTTP://WWW.W3.ORG/TR/XHTML1/DTD/XHTML1-STRICT.DTD="http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd"; 1.0="1.0" -//W3C//DTD="-//W3C//DTD" STRICT//EN="Strict//EN" |
| 21:01 | <Philip`> | =" |
| 21:01 | <Philip`> | " XHTML="XHTML"> |
| 21:01 | <Philip`> | Hmm, I can guess how that happened |
| 21:03 | <Dashiva> | haha |
| 21:04 | <Hixie> | wow, what a mess |
| 21:04 | <Dashiva> | Whitespace is our friend, right? |
| 21:13 | <Hixie> | hsivonen: re your request for rewriting the outlining algorithm |
| 21:13 | <Hixie> | you say |
| 21:13 | <Hixie> | > This sweep should probably define what checkpoint data to |
| 21:13 | <Hixie> | > store on each element that implements the HTMLHeaderElement interface to allow |
| 21:13 | <Hixie> | > local recomputation. |
| 21:14 | <Hixie> | can't i just say what i usually say, namely that at any time, the outline must match the results you would get from running the algorithm over the whole tree? |
| 21:14 | <Hixie> | i don't want to be the one defining how you are to do optimisations... |
| 21:15 | <hsivonen> | Hixie: does stating that actually make the said optimization reasonably implementable? |
| 21:15 | <Hixie> | depends what the algorithm is |
| 21:16 | <Philip`> | http://philip.html5.org/data/doctypes-2.html#%3C!doctype_html%3E - now with links to pages, at least for the obscurer doctypes |
| 21:17 | <hsivonen> | Hixie: I don't see how it would hurt to define the outline in terms of forward sweeps from a hn state |
| 21:17 | <Hixie> | oh it wouldn't, and that's just a subset of defining it as part of a forward sweep in general |
| 21:17 | <Hixie> | i'm just saying the spec shouldn't define ny algorithm twice |
| 21:17 | <Hixie> | since the spec shouldn't be optimising |
| 21:18 | <Hixie> | or rather, it should be optimising for lack of ambiguity |
| 21:18 | <Hixie> | and adding two algorithms defeats that :-) |
| 21:18 | <gsnedders> | the current section confuses me |
| 21:18 | <Hixie> | hmm... i guess each sectioning element is effectively atomic in the algorithm... |
| 21:18 | <Hixie> | that is, anything outside a section can't affect what's inside it |
| 21:18 | <Hixie> | and what's inside a section can only affect the relative position of further nested sections... |
| 21:19 | <Hixie> | that's useful |
| 21:19 | <hsivonen> | Hixie: surely a forward sweep algorith can be the only definition |
| 21:19 | <Hixie> | hsivonen: yes |
| 21:19 | <Hixie> | hsivonen: that's what i plan to do |
| 21:19 | <hsivonen> | Hixie: excellent. thanks |
| 21:20 | <Hixie> | but your e-mail asked for one or two other things too, which i'm not sure i should necessarily do |
| 21:20 | <Hixie> | actually i think what i quoted above is a non-issue, and your second requirement might already be met |