| 09:17 | <hsivonen> | hmm. does C# allow switch() {} with any string values or just interned strings? |
| 09:17 | <hsivonen> | just curious after glancing at the Twintsam source link for [imps] |
| 09:17 | <hsivonen> | from [imps] |
| 09:21 | <Lachy> | hsivonen, check the language spec http://msdn2.microsoft.com/en-us/library/Aa645596(VS.71).aspx |
| 09:25 | <hsivonen> | hmm. looks like any string is allowed |
| 09:50 | <zcorpan> | another case where serialization needs to throw: an element in the "http://www.w3.org/2000/xmlns/"; namespace |
| 09:53 | <zcorpan> | and an unclear case: two attributes that have the same namespace and local name |
| 12:12 | zcorpan | looks at http://www.w3.org/TR/xml/#NT-AttValue |
| 12:13 | <zcorpan> | where does it say that attribute have to follow the Char production? |
| 12:52 | <hsivonen> | what's the mediawiki way to say <code>? |
| 12:53 | <hsivonen> | bah. <code> according to http://diberri.dyndns.org/wikipedia/html2wiki/index.html |
| 13:19 | <hsivonen> | Lachy: is there a way to permanently convince the whatwg wiki that I am not a bot and to suppress the math test? |
| 13:21 | <Lachy> | hsivonen, maybe. |
| 13:26 | <Lachy> | hsivonen, it should only ask you that when you attempt to add a URI to the page |
| 13:28 | <hsivonen> | Lachy: ah. |
| 13:28 | <hsivonen> | Lachy: I have namespace URIs |
| 13:28 | <Lachy> | ok |
| 13:29 | <Lachy> | I may be able to turn it off for some additional user groups, I'm just trying to find out what group I can put you in to do it |
| 13:33 | <Lachy> | hsivonen, I made you a Sysop and Bureaucrat so you have more rights now and won't see the captcha |
| 13:37 | <hsivonen> | Lachy: thanks |
| 13:37 | <Lachy> | does anyone know if it's possible to delete users? I want to get rid of known spammers completely |
| 13:55 | <Lachy> | I'm upgrading wordpress on the blog, so it may break a little bit during the process |
| 13:57 | <met_> | hi |
| 13:57 | <met_> | http://blog.whatwg.org/ gives some php errors |
| 13:59 | <Philip`> | met_: http://krijnhoetmer.nl/irc-logs/whatwg/20070911#l-104 |
| 14:00 | <met_> | oh, haven't read yesterday log, ok |
| 14:00 | <met_> | no today, i am blind |
| 14:00 | <met_> | completelly 8-) |
| 14:00 | <Philip`> | It was a minute before you joined :-) |
| 14:14 | <Lachy> | blog is working again |
| 14:23 | <Lachy> | wordpress update is complete, now with a comment preview function |
| 14:30 | <hsivonen> | http://html4all.org/wiki/index.php/List_Rules |
| 14:30 | <hsivonen> | public-html might benefit from adherence to those rules |
| 14:35 | <Philip`> | Do they have a new secret hideout where things like "Greg's draft" of the list rules were circulated? |
| 14:37 | <Lachy> | I finally upgraded my own blog too :-) |
| 14:37 | <Lachy> | I think I was running an obsolete and insecure version of WP for a while now |
| 14:56 | <hsivonen> | http://wiki.whatwg.org/wiki/Validator.nu_XML_Output |
| 14:56 | <hsivonen> | comments welcome before I implement |
| 15:01 | zcorpan | would suggest to rename the uri attribute to url, though doesn't feel strongly about it |
| 15:02 | <hsivonen> | zcorpan: thanks. done. |
| 15:07 | <zcorpan> | hsivonen: is it not allowed to have <code><a> or <a><code> in <message>? |
| 15:07 | <zcorpan> | i.e. nest them |
| 15:07 | <hsivonen> | zcorpan: <a><code> should be allowed by the prose. |
| 15:07 | <zcorpan> | ah, right |
| 15:08 | <zcorpan> | why not the other way around? |
| 15:08 | <hsivonen> | zcorpan: simplicity |
| 15:08 | <zcorpan> | ok |
| 15:09 | <Lachy> | why put the url attribute on each info, error and non-document-error element instead of just once on their parent messages element? |
| 15:09 | <hsivonen> | Lachy: because messages may pertain to schema or DTD files |
| 15:10 | <hsivonen> | Lachy: defining inheritance could save bytes, though |
| 15:10 | <Lachy> | ok |
| 15:10 | <zcorpan> | the m element is in the v.validator.nu... namespace? |
| 15:10 | <Lachy> | yeah, I'd do <messages url=".."> for the default url and then allow url="" on other elements if its different for some |
| 15:11 | <hsivonen> | zcorpan: in the n.validator.nu... |
| 15:11 | <hsivonen> | Lachy: makes sense |
| 15:12 | <Lachy> | so if someone submits an IRI, will the output include the punycode version? |
| 15:12 | <hsivonen> | Lachy: yes |
| 15:12 | <Lachy> | ok. what's the reason for not outputting IRIs? |
| 15:13 | <Lachy> | for better compatibility with legacy software? |
| 15:13 | <hsivonen> | Lachy: 1) I don't have the code to do that. 2) It would be less compatible with IRIless clients. |
| 15:17 | <Lachy> | does this "implementation may count column numbers in terms of UTF-16 code units instead of characters." mean that characters above the basic multilingual plane, which use more than 16 bits, would be counted as 2 characters instead of 1? |
| 15:18 | <zcorpan> | looks good to me |
| 15:19 | <hsivonen> | Lachy: yes |
| 15:20 | <hsivonen> | zcorpan: thanks |
| 15:20 | <Lachy> | hsivonen, would a Content-Type issue be emitted as a <non-document-error>? |
| 15:21 | <hsivonen> | Lachy: hmm. IIRC, not |
| 15:21 | <hsivonen> | Lachy: IIRC, that's an IOException internally |
| 15:22 | <Lachy> | so wouldn't that qualify for <non-document-error type="io">? |
| 15:25 | <hsivonen> | Lachy: doh. yes |
| 15:26 | <hsivonen> | I should have said yes, above. sorry |
| 15:26 | <Lachy> | hsivonen, did you look at the XML output available from validator.w3.org and base any of this of that? |
| 15:26 | <hsivonen> | Lachy: I did |
| 15:26 | <hsivonen> | Lachy: it is non-streamable |
| 15:27 | <hsivonen> | or they are, rather |
| 15:27 | <hsivonen> | "I think there are two problems with the SOAP and Unicorn formats: they are unnecessarily complex and they don’t support streaming output." |
| 15:28 | <hsivonen> | though it is possible that I overvalue streaming |
| 15:34 | <hsivonen> | Lachy: also, the W3C validator format uses the entity-escaped HTML anti-pattern |
| 15:34 | <Lachy> | what does that mean? |
| 15:35 | <hsivonen> | Lachy: it's like RSS |
| 15:35 | <Lachy> | like the <![CDATA[ stuff with escaped HTML? |
| 15:35 | <hsivonen> | Lachy: yes |
| 15:37 | <Lachy> | oh I see, like in the <m:explanation> and <m:source> elements where it should just use XHTML. |
| 15:37 | <Lachy> | http://validator.w3.org/docs/api.html |
| 15:38 | <zcorpan> | hsivonen: are clients expected to reject responses where the root element is "messages" in some other namespace? |
| 15:38 | <hsivonen> | zcorpan: yes |
| 15:40 | <hsivonen> | and now it is time to spec a JSON format... |
| 15:40 | <Lachy> | I don't understand the <parse-tree> section |
| 15:41 | <hsivonen> | Lachy: if I didn't salt the namespaces and someone loaded it in a browser, it would be a script injection bug to the validator.nu domain |
| 15:42 | <zcorpan> | how can you serialize the parse tree to xml if you parsed html that is not serializable as xml? |
| 15:42 | <Lachy> | no, I understood the reason for the namespace change, but not how you serialise a parse tree |
| 15:42 | <Lachy> | wouldn't that just be the entire document? |
| 15:42 | <hsivonen> | zcorpan: I intend to cheat |
| 15:43 | <hsivonen> | zcorpan: in the case of white space |
| 15:43 | <zcorpan> | hsivonen: what about e.g. colons in local names? |
| 15:43 | <hsivonen> | with non-NCName names, I guess I'm going to drop the parse tree |
| 15:43 | <zcorpan> | ok |
| 15:43 | <hsivonen> | Lachy: yes |
| 15:44 | <zcorpan> | perhaps the parse tree could be serialized as http://simon.html5.org/specs/sdf instead? :) |
| 15:44 | <hsivonen> | I expect the parse tree part to be the least useful and the most controversial |
| 15:44 | <Lachy> | how do you intend to represent the parse tree of HTML documents, especially for the cases where HTML and XML differ |
| 15:44 | <zcorpan> | though i need to support character escaping for that format |
| 15:45 | <hsivonen> | Lachy: I intend not to |
| 15:45 | <hsivonen> | zcorpan: yeah, perhaps I should use SDF instead |
| 15:46 | <Lachy> | hsivonen, yeah, sdf would make sense |
| 15:47 | <gsnedders> | http://simplepie.org/support/viewtopic.php?id=1167 —that's so typical of PHP bugs. |
| 15:47 | <zcorpan> | "\u000C" for a form feed? |
| 15:48 | <zcorpan> | perhaps i should look at how css escapes characters and do the same |
| 15:48 | <hsivonen> | zcorpan: if you want to represent arbitratry DOMs, you should probably represent UTF-16 code point sequences |
| 15:48 | <hsivonen> | not character sequences |
| 15:49 | <hsivonen> | \uXXXX |
| 15:49 | <hsivonen> | is fine |
| 15:49 | <zcorpan> | ok |
| 15:49 | <hsivonen> | with surrogates as two escapes |
| 15:49 | <zcorpan> | makes sense |
| 15:50 | <Lachy> | yeah, \uXXXX would be compatible with JavaScript escape sequences too |
| 15:50 | <Lachy> | which might be useful |
| 15:55 | <zcorpan> | perhaps the whole spec should talk about utf-16 code points instead of characters |
| 15:55 | <hsivonen> | yeah |
| 15:56 | <hsivonen> | code units, actually |
| 15:56 | <Lachy> | zcorpan, in SDF, do all attribute nodes have to occur before any other child nodes of an element? |
| 15:56 | <zcorpan> | Lachy: no |
| 15:56 | <zcorpan> | U+0022 QUOTATION MARK -> 0x0022 UTF-16 code unit ? |
| 15:57 | <hsivonen> | yeah |
| 15:57 | <hsivonen> | but saying QUOTATION MARK is still useful |
| 15:57 | <Lachy> | so this would be conforming: |
| 15:57 | <Lachy> | e "div" |
| 15:57 | <Lachy> | t "foo" |
| 15:57 | <Lachy> | a "class" "bar" |
| 15:57 | <zcorpan> | yes |
| 15:58 | <hsivonen> | zcorpan: any particular reason for not have exactly one way of doing it? (so that results would be byte-wise comparable for equality) |
| 15:58 | <zcorpan> | i haven't thought much about serializing to sdf yet |
| 15:59 | <Lachy> | I would have expected that to be non-conforming and for each node to represented in document order |
| 16:00 | <zcorpan> | yeah, although attributes are unordered :) and if you implement it using dom methods it doesn't really matter if an attribute appears after other nodes |
| 16:02 | <zcorpan> | if you write sdf by hand you might not want to care about what order you place the attributes (though it makes sense to put them before child nodes of the element) |
| 16:02 | <zcorpan> | but when serializing and comparing you'd probably sort the attributes |
| 16:07 | <Lachy> | hmm. sdf is a lot nicer than the format onsgmls outputs, which emits attributes before the element |
| 16:09 | <zcorpan> | is there a good character that would be encoded as a surrogate pair in utf-16 that i can use in the examples? |
| 16:10 | <Lachy> | zcorpan, U+1047E |
| 16:11 | <zcorpan> | thanks |
| 16:11 | <Lachy> | that's the one that appears in Hixie's email sig |
| 16:11 | <zcorpan> | noticed (by googling for it) :) |
| 16:31 | <zcorpan> | ok, specced escaping |
| 16:34 | zcorpan | hopes he got the example right |
| 16:36 | <Lachy> | zcorpan, can CR and LF occur anywhere within a string? |
| 16:37 | <Lachy> | so if a text node spans multiple lines, it's just represnted like this:.. |
| 16:37 | <zcorpan> | they can |
| 16:37 | <Lachy> | t "this string contains |
| 16:37 | <Lachy> | a new line" |
| 16:38 | <zcorpan> | yes |
| 16:39 | <Lachy> | you should probably change "It is not intended for data exchange, but rather intended to be used for test suites.", especially if Henri ends up using it for the validator's parse tree output |
| 16:40 | <Lachy> | maybe say that it's primarily intended for test suites |
| 16:43 | <hsivonen> | communicating DOMs on XML without using XML doesn't quite feel right |
| 16:44 | hsivonen | likes the images of the JSON spec |
| 16:45 | <zcorpan> | Lachy: changed |
| 17:00 | <zcorpan> | how is "0x000A" generally pronounced? oh-ecks-... or zero-ecks-... ? |
| 17:01 | <Lachy> | zcorpan, I pronounce it zero-ecks... |
| 17:01 | <zcorpan> | ok |
| 17:01 | <Lachy> | why? |
| 17:01 | <zcorpan> | "A 0x..." vs "An 0x..." |
| 17:02 | <gsnedders> | zcorpan: I pronounce it oh-ecks-… |
| 17:02 | <zcorpan> | oh well :) |
| 17:03 | <gsnedders> | zcorpan: though normally only in my head. I have little occasion to actually talk about such things :) |
| 17:04 | <zcorpan> | yeah, but it might be annoying to read text that assumes the other pronounciation (by having "a" or "an" in front of it) |
| 17:05 | <zcorpan> | "a FAQ" vs "an FAQ" is the same |
| 17:05 | <Lachy> | whenever I read something like that, I just change it mentally to read it how I like it |
| 17:06 | <zcorpan> | yeah, that's what is annoying (not very but still) |
| 17:06 | <Lachy> | "an FAQ" is correct because I pronounce it F-A-Q rather than "Fack" |
| 17:07 | zcorpan | pronounces it as "fack" |
| 17:09 | <gsnedders> | F-A-Q |
| 17:11 | <Lachy> | zcorpan, SDF should probably say something about character encodings |
| 17:12 | <Lachy> | should a consumer assume SDF files are encoded as UTF-8 or is that left explicitly undefined? |
| 17:13 | <zcorpan> | dunno |
| 17:15 | <zcorpan> | haven't thought much about it, but so far i've just used bomless utf-8 |
| 17:15 | <Lachy> | maybe say encoded as UTF-8 or UTF-16 (with an appropriate BOM) or otherwise specified by a higher level protocol or container format |
| 17:19 | <hsivonen> | Lachy: why bother with UTF-16 if you have a chance of mandating UTF-8? |
| 17:19 | <Lachy> | dunno |
| 17:20 | <zcorpan> | thinking about it, perhaps the spec should talk about characters instead of code units for the things that are integral to the syntax itself, and that it's just the \uXXXX escaping that should talk about utf-16 code units |
| 17:21 | <hsivonen> | zcorpan: the language itself should probably be defined in terms of characters but the string values it encodes should be strings of UTF-16 code units |
| 17:21 | <zcorpan> | yeah |
| 17:25 | <hsivonen> | are there any guides on JSON design patterns? |
| 17:25 | <hsivonen> | I tried to google, but I only found references to the use of JSON itself being an Ajax design pattern |
| 17:27 | <Lachy> | you should be able to just do a 1:1 mapping between elements in the XML format and objects in JSON, with attributes represented as properties. |
| 17:28 | <hsivonen> | Lachy: wouldn't such a JSON format suck compared to one designed as JSON? |
| 17:28 | <Lachy> | maybe |
| 17:28 | <hsivonen> | Lachy: how do you do mixed content in JSON? |
| 17:29 | <Lachy> | what kind of mixed content? |
| 17:29 | <hsivonen> | the kind that mixes text and inline markup |
| 17:29 | <hsivonen> | as far as I can tell, JSON doesn't do mixed content |
| 17:30 | <hsivonen> | which means no document-style XML |
| 17:30 | <hsivonen> | only "data"-style XML |
| 17:30 | zcorpan | isn't really sure how to spec the escaping part |
| 17:30 | <hober> | "foo ", {"bar": [{}. "baz"]} " quux" is something like "foo <bar>baz</bar> quux" |
| 17:30 | <Lachy> | yeah, you'd have to represent xml as strings |
| 17:31 | <hsivonen> | zcorpan: I suggest escaping everything except code units that correspond to printable ASCII |
| 17:31 | <hober> | (assuming you fix the missing commas, typos, etc. in my example) |
| 17:32 | <hsivonen> | zcorpan: assuming you want something that is hard to break and unambiguous for byte comparisons |
| 17:32 | <zcorpan> | ok |
| 17:34 | <hsivonen> | zcorpan: though that would make the format suck big time as a human-readable tree dump for real pages |
| 17:35 | <hsivonen> | as opposed to ASCII-dominated test cases |
| 17:35 | <zcorpan> | indeed |
| 17:37 | <zcorpan> | though mixing utf-16 escapes and utf-8 characters seems error prone |
| 17:38 | <zcorpan> | also, if all non-ascii is escaped it doesn't matter much what encoding is used for the format itself |
| 17:39 | <Lachy> | hsivonen, something like this could work http://tinyurl.com/ytldcz |
| 17:39 | <hsivonen> | Lachy: yes, except I'm going to leave elaboration out |
| 17:40 | <Lachy> | ok |
| 17:42 | <hsivonen> | http://wiki.whatwg.org/wiki/Validator.nu_JSON_Output |
| 17:43 | <Lachy> | what's extract-offset? |
| 17:44 | <hsivonen> | a pointer to the character of interest inside extract |
| 17:44 | <hsivonen> | since <m> and mixed content is not JSON-like |
| 17:46 | <Lachy> | ok, so it's value would have to give start and end points or start and length |
| 17:49 | <hsivonen> | no, only offset to the character of interest |
| 17:50 | <hsivonen> | <m> conceptually only designates one character, but it will have to expand not to cut a sequence of a base character plus combining characters |
| 17:50 | <hsivonen> | an offset in JSON does not have that problem |
| 17:54 | <Lachy> | ok, I thought <m> would mark the whole section. e.g. for an invalid attribute, it would mark the whole thing: <p <m>align="right"</m>> ... |
| 17:55 | <hsivonen> | Lachy: no, it is just for indicating which part of the extract corresponds to line & col |
| 17:55 | <Lachy> | ah |
| 17:55 | <hsivonen> | attribute errors will point to the last > of the start tag |
| 17:57 | <Lachy> | oh, that's unfortunate |
| 17:57 | <hsivonen> | I need to figure out how to do this without letting a huge string of combining diacritics to be used as a DoS attack |
| 17:57 | <Lachy> | that means for something making use of the API which wants to highlight the attribute itself would need to search the string itself |
| 17:58 | <hsivonen> | Lachy: SAX only allows source locations on a per-event basis |
| 17:58 | <Lachy> | oh right |
| 17:58 | <hsivonen> | Lachy: and the whole start tag is the event |
| 17:59 | <Lachy> | I suppose it wouldn't be too hard to use a regex to find the attribute in most cases |
| 18:00 | <hsivonen> | Lachy: well, in the case of elements, the element may be implied so you wouldn't find anything matching in the source |
| 18:00 | <hsivonen> | at least SAX does some source location |
| 18:00 | <hsivonen> | the DOM does none |
| 18:01 | <hsivonen> | which is one of the reasons why I wrote my own tree model |
| 18:02 | <Lachy> | I updated my example based on the current description in the wiki http://html5.lachy.id.au/clipboard |
| 18:03 | <hsivonen> | Lachy: is it OK to copy that into the wiki under the MIT license? |
| 18:05 | <Lachy> | as long as you comply with my copyright licence <http://lachy.id.au/about/copyright>;, then sure |
| 18:05 | <zcorpan> | ok, rewrote the escaping part |
| 18:08 | <zcorpan> | hmm |
| 18:08 | <hsivonen> | Lachy: unfortunately, I'm not at liberty to release my changes to the Public Domain |
| 18:08 | <Lachy> | no worries |
| 18:09 | <Lachy> | everything after the first paragraph is basically a non-binding suggestion |
| 18:09 | <Lachy> | maybe I should update it to say "or a free licence" |
| 18:10 | <hsivonen> | Lachy: I realized it was non-binding |
| 18:11 | <hsivonen> | speccing this stuff is so much more tedious than just implementing it and telling people to view source and guess |
| 18:14 | <zcorpan> | there, i think i got it right now |
| 18:15 | <Lachy> | I updated it |
| 18:19 | <Lachy> | hsivonen, I think you should change "extract-offset" to "offset" |
| 18:19 | <Lachy> | and consider shortening "non-document-error" |
| 18:22 | Philip` | updates http://canvex.lazyilluminati.com/misc/ref/ref.html so all the examples get automatically conformance-checkered when the HTML is generated |
| 18:24 | <hsivonen> | Lachy: any suggestions for "non-document-error"? it won't waste too many bytes as there will be unlikely to be more than two per result |
| 18:24 | <hsivonen> | Philip`: with what software? |
| 18:27 | <Philip`> | hsivonen: It's using the validator.nu API (with text output) |
| 18:27 | <Philip`> | (via httplib) |
| 18:29 | <hsivonen> | Philip`: ok |
| 18:29 | <Lachy> | my example had syntax errors, I fixed them and updated the wiki |
| 18:29 | <hsivonen> | Lachy: thanks |
| 18:32 | <Lachy> | hsivonen, probably best to avoid "-" characters in property names because it allows scripts to use this notation: result.messages[0].line, which can't be done with result.result.parse-tree |
| 18:32 | <Lachy> | authors would have to use result.result["parse-tree"] instead |
| 18:33 | <hsivonen> | Lachy: ok. is any of the names a reserved word by any chance? e.g. "class" is a usual suspect |
| 18:34 | <hsivonen> | Lachy: should I use camelCase instead? |
| 18:34 | <Lachy> | yeah, that would work |
| 18:36 | <hsivonen> | class will be reserved in the future |
| 18:36 | <hsivonen> | http://developer.mozilla.org/en/docs/Core_JavaScript_1.5_Reference:Reserved_Words |
| 18:37 | <hsivonen> | should I have type and subtype instead of class and type? |
| 18:38 | <Lachy> | yeah |
| 18:41 | <virtuelv> | (mildly OT: What is the status of selector support in Firefox 3? does http://virtuelvis.com/download/2007/09/backgammon/ render correctly?) |
| 18:48 | <Philip`> | virtuelv: Doesn't work in a build from a few days ago |
| 18:49 | <Philip`> | http://www.css3.info/selectors-test/test.html indicates that everything from nth-child to nth-last-of-type is unsupported |
| 18:52 | <hsivonen> | What technical change to HTML5 is called for in the message about braille printing charts? |
| 19:13 | <Lachy> | wow, IE only has 63.86% market share now according to http://www.itproductivity.org/browser.htm |
| 19:27 | <Philip`> | Lachy: And apparently Netscape has 10% - that sounds a bit odd |
| 19:28 | <Lachy> | yeah, it's nowhere near the vales wikipedia lists from several other sources |
| 19:28 | <zcorpan> | krijnh: would be nice to pick up on /me lines also :) |
| 19:30 | <Lachy> | krijnh, it would be nice to not use grey text on a green background also. it's so hard to read |
| 19:30 | <zcorpan> | /^.{8}\* ([^ ]+)/ |
| 19:30 | <Philip`> | "The survey is accurate to +/- 1.0%" is peculiar - that number of decimal places doesn't really seem justified |
| 19:31 | <kingryan> | any idea of the survey's methodology? |
| 19:31 | <Philip`> | "a majority of the NS users are on version 4.x"! |
| 19:33 | <Lachy> | see what slashdot readers have to say about it http://slashdot.org/article.pl?sid=07/09/11/1222256 |
| 19:34 | <zcorpan> | the change is also percent units, not percents |
| 19:34 | <Philip`> | http://www.e-janco.com/Samples/BrowserSample.pdf lists the sites they got data from |
| 19:35 | <Philip`> | (I like how their you-must-register-to-download-it form is circumvented by disabling JavaScript) |
| 19:36 | <Lachy> | hmm. 9 web sites isn't really a reasonable sample size |
| 19:36 | <Philip`> | By the way, how can they say 63.86% +/- 1.0%? |
| 19:36 | <zcorpan> | e.g. opera had 0.78% in 2006 and 1.87% in 2007 according to that table, that's a change of +240% |
| 19:38 | <zcorpan> | Philip`: i would guess they rounded the numbers arbitrarily and picked 1.0% out from the blue :) |
| 19:40 | <markp> | re: http://krijnhoetmer.nl/irc-logs/whatwg/20070906#l-554 |
| 19:40 | <markp> | html5lib does NOT have a validator |
| 19:40 | <markp> | it has a prototype of an idea of a concept of a half-baked scheme of a validator |
| 19:40 | <markp> | i hope that clarifies things |
| 19:40 | <Philip`_> | How come Netscape usage increased from 6.3% to 10.2% in the past two years, when a majority of Netscape users are (apparently) Netscape 4.x? Does that mean everybody was using NS4 two years ago, or more people have started using NS4 now than two years ago? |
| 19:41 | <markp> | and a disturbing amount of it was written on a red-eye flight from CA to NC |
| 19:41 | <Hixie> | netscape usage isn't anywhere near 6% or 10% |
| 19:41 | <kingryan> | markp: I didn't mean to overrepresent it as something useful :) |
| 19:42 | <markp> | :) |
| 19:42 | <zcorpan> | some bots might identify themselves as nn4 |
| 19:42 | <zcorpan> | though i wouldn't expect that to change the number much |
| 19:43 | <Philip`_> | The numbers don't seem entirely convincing |
| 19:44 | <Philip`_> | and I'd prefer to not be convinced that people still actually use Netscape 4 :-) |
| 20:19 | zcorpan | is unsure how to implement the escaping with javascript |
| 20:20 | <Dashiva> | What kind of escaping? |
| 20:21 | <zcorpan> | in sdf |
| 20:21 | <zcorpan> | http://simon.html5.org/specs/sdf |
| 20:21 | <zcorpan> | see the last example at the bottom |
| 20:25 | <Dashiva> | Newlines are serialized as actual newlines? |
| 20:26 | <zcorpan> | LFs, yes |
| 20:27 | <Dashiva> | Isn't that likely to cause trouble outside browsers? |
| 20:27 | <zcorpan> | why? |
| 20:28 | <Dashiva> | Could be misinterpreted as a line separator? |
| 20:29 | <Lachy> | zcorpan, see http://lachy.id.au/dev/mozilla/sidebar/Unicode/character-tools.html |
| 20:30 | <Lachy> | that has the code to deal with parsing \uXXXX |
| 20:30 | <zcorpan> | Dashiva: you mean you would parse line for line instead of character by character? |
| 20:30 | <zcorpan> | Lachy: thanks |
| 20:31 | <Dashiva> | I just know I've used text editors that do weird stuff when faced with both CR and LF mixed |
| 20:31 | <Lachy> | String.fromCharCode(codepoint) where code point is XXXX |
| 20:32 | <zcorpan> | Lachy: but will that work correctly with surrogate pairs? |
| 20:32 | <Lachy> | you have to do each one separately, JS can't deal with astral characters |
| 20:32 | <zcorpan> | Dashiva: but this format will only have LFs |
| 20:33 | <Dashiva> | Oh! I missed that |
| 20:33 | <zcorpan> | Lachy: oh. well then that makes it simpler |
| 20:33 | <Lachy> | zcorpan, you should probably make that clearer about LFs, even though that's actually specified in HTML5 |
| 20:33 | <Dashiva> | What's the reason for not escaping them inside strings, though? |
| 20:34 | <Dashiva> | It makes the code more complex, after all |
| 20:34 | <zcorpan> | Lachy: html5? |
| 20:34 | <Lachy> | doesn't HTML5 specify how to serialise a DOM |
| 20:35 | <zcorpan> | umm. to html and xml using innerHTML |
| 20:35 | <Lachy> | well, actually, it defines that CR and CRLF wil be replaced with LF in the input stream, so CRs won't appear in the output |
| 20:35 | <Lachy> | unless they're added later |
| 20:35 | <zcorpan> | Lachy: right, but the dom can have CRs |
| 20:36 | <zcorpan> | Lachy: CRs have to be escaped in sdf |
| 20:36 | <Lachy> | why? |
| 20:36 | <zcorpan> | to avoid ending up with a mix of CRs and LFs |
| 20:37 | <zcorpan> | some text editors will be smart and make sure the file has consistent line endings |
| 20:38 | <zcorpan> | Dashiva: i'm not sure what is more complex by having LFs in strings |
| 20:39 | <Dashiva> | You need an additional check if ( c == 0x000A ) { emit LF } |
| 20:39 | <Lachy> | zcorpan, this works http://html5.lachy.id.au/output?data=%3Cscript%3E%0D%0Avar+c1+%3D+parseInt%28%22D801%22%2C+16%29%3B%0D%0Avar+c2+%3D+parseInt%28%22DC7E%22%2C+16%29%3B%0D%0Avar+chars+%3D+String.fromCharCode%28c1%29+%2B+String.fromCharCode%28c2%29%3B%0D%0A%0D%0Adocument.write%28chars%29%3B%0D%0A%3C%2Fscript%3E&type=text%2Fhtml%3B+charset%3DUTF-8 |
| 20:39 | <Philip`> | LFs in strings would mean you couldn't use a text editor to indent a chunk of SDF |
| 20:40 | <zcorpan> | Philip`: why not? |
| 20:40 | <Dashiva> | And from the syntax, it seems like SDF strings are intended to be JS-parseable, which the LFs break |
| 20:40 | <Lachy> | zcorpan, because you'd introduce extra spaces into the string |
| 20:40 | <Philip`> | Because the editor would add indentation at the start of each line, and if the start of a line is the inside of a string, then you'd be modifying the inside of the string unintentionally |
| 20:41 | <zcorpan> | good points |
| 20:41 | zcorpan | takes out the LF special case |
| 20:41 | <Lachy> | zcorpan, maybe allow \n and \r to occur within strings |
| 20:42 | <zcorpan> | maybe. if i do then i could also allow the other shorthands |
| 20:42 | <Lachy> | like \t? |
| 20:43 | <zcorpan> | yeah, and \f |
| 20:43 | <Lachy> | oh right |
| 20:43 | <Lachy> | yeah, those are compatible with various libraries and programming languages, so that shouldn't add too much complexity |
| 20:44 | <Philip`> | Can you just define it to be a JSON string? |
| 20:44 | <Dashiva> | That's pretty much what it is |
| 20:44 | <zcorpan> | well then :) |
| 20:44 | <Dashiva> | http://pastebot.nd.edu/3535 |
| 20:45 | <zcorpan> | i guess i could simplify my parser after these changes |
| 20:47 | <Dashiva> | One concern about the current form is that it uses a lot of bytes for non-printables inside the ascii range, but that's hardly a big deal |
| 20:50 | zcorpan | puts sdf in version control |
| 20:52 | <hsivonen> | hmm. I wonder if a reply from me is expected to John Foliot's public-html follow-up to my message |
| 20:58 | <zcorpan> | ok, changed string to JSON string |
| 21:01 | <Hixie> | any italians able to translate this for us? :-) http://scaccoalweb.vnunet.it/2007/09/html-5-ritorno-.html |
| 21:01 | <zcorpan> | google translate seems to work :) |
| 21:01 | <Lachy> | Hixie http://translate.google.com/translate?u=http%3A%2F%2Fscaccoalweb.vnunet.it%2F2007%2F09%2Fhtml-5-ritorno-.html |
| 21:02 | <Hixie> | wow, a translation service that works? *skeptical* |
| 21:03 | zcorpan | points to http://hsivonen.iki.fi/test/translate.html |
| 21:06 | <zcorpan> | i use that bookmarklet a lot :) |
| 21:06 | <Hixie> | :-) |
| 21:06 | <Hixie> | it doesn't work on my portal |
| 21:06 | <Hixie> | i guess encodeURIComponent() doesn't work on IDN uris |
| 21:43 | <Hixie> | www.satogo.com has a pretty cool AT tool |
| 21:43 | <Hixie> | seems on par with JAWS in many ways |
| 21:44 | <Hixie> | (there's a free trial if you want to try it) |
| 23:02 | <Hixie> | i'm amused by the arguments that say that x% of pages omit alt attributes, since they seem to forget the x% of pages that do include alt attributes but do so in fundamentally useless or bad ways |
| 23:15 | <Philip`> | It matters that the x% omitting alt because of lazy/stupid/careless/etc authors is non-zero and non-negligible, because then the 0.00y% that omits alt because their authors conscientiously followed the HTML5 spec (and that has indistinguishable syntax from the first case) will have to be processed with the "lazy/stupid/careless author forgot about alt" semantics, not the "critical part of content" semantics, in order to work in the majority of cases |
| 23:39 | zcorpan | finds it somewhat ironic that http://blog.whatwg.org/result-format seems to have comments disabled |
| 23:40 | <zcorpan> | hsivonen: was that intentional? |
| 23:50 | <hsivonen> | zcorpan: no. the admin interface tells me they are enabled |
| 23:50 | <hsivonen> | Lachy: see above. fallout from the upgrade? |
| 23:50 | <hsivonen> | nn |
| 23:51 | <zcorpan> | nn hsivonen |