| 00:01 | <Tripknotix> | whats, whatwg stand for |
| 00:02 | <Hixie> | Tripknotix: Web Hypertext Application Technology Working Group |
| 00:03 | <Tripknotix> | so whys this the place to talk about html5 ? |
| 00:03 | <Hixie> | WHATWG is one of the groups working on HTML5 |
| 00:04 | <Tripknotix> | oh |
| 00:04 | <Tripknotix> | interesting |
| 00:04 | <Hixie> | see also www.whatwg.org and http://wiki.whatwg.org/wiki/FAQ |
| 00:29 | <Hixie> | hsivonen: yt? |
| 05:15 | <MikeSmith> | hsivonen_: please ping me when you get back on -- if/when you have time to chat a little. I want to ask some boneheaded questions about how to actually get the v.nu error-handling backend to report warnings for exceptions that have warningness set |
| 05:19 | <MikeSmith> | hsivonen_: in the mean time I'll read more through the code to try to give myself some clue |
| 05:19 | <MikeSmith> | and your thesis |
| 07:22 | <hsivonen_> | Hixie: I'm here now |
| 07:22 | <hsivonen_> | looks like MikeSmith just left :-( |
| 07:25 | <hsivonen_> | oh. a new ED of XHTML2 |
| 07:43 | <Hixie> | hsivonen_: i don't recall why i was pinging you, sorry. |
| 07:45 | <hsivonen_> | Hixie: ok |
| 07:58 | <Philip`> | "<zcorpan> so when's WHATWG LC? <Hixie> probably monday" - it's late :'-( |
| 07:58 | <Philip`> | in most timezones |
| 07:58 | <Hixie> | yeah, gonna be tomorrow night my timezone i think |
| 08:12 | <a-ja> | anyone know of a microdata extractor/tester service that works with recent spec changes? |
| 08:13 | a-ja | only found one for circa may timeframe |
| 08:15 | <Hixie> | i'm not aware of any, but it wouldn't be hard to make one |
| 08:16 | <a-ja> | was just playing with trying to make some calendars / vcards "multi-lingual"....uF and microdata |
| 08:17 | <Hixie> | cool |
| 08:18 | <a-ja> | it's not exactly pretty :) |
| 08:23 | <jgraham> | a-ja: I am in the process of updating my python implementation to the latest spec. It will probably take a few more evenings since I'm not trying very hard |
| 08:24 | <a-ja> | oh, cool....i'll check it again soon then |
| 08:28 | <Philip`> | I'm not in the process of updating my JS implementation |
| 08:28 | <Philip`> | Someone else could feel free to fix it for me :-) |
| 08:28 | <a-ja> | fwiw....seemed to me that included format adr & geo should get added to html5. and perhaps let vevent location have addr/geo instead of just text |
| 08:29 | <a-ja> | erm....adr/geo/vcard |
| 08:30 | <a-ja> | Philip`: URL? |
| 08:34 | <Philip`> | a-ja: http://philip.html5.org/demos/microdata/demo.html |
| 08:35 | <a-ja> | tks......seen it before and have it bookmarked already |
| 09:04 | <hsivonen_> | Is a control named isindex magic in IE even if the submission encoding type is not the default? |
| 09:08 | <hsivonen> | never mind |
| 09:12 | <hsivonen> | eww. Are browsers really supposed to submit the localized string for "Submit" when the submit button has no specified value? |
| 09:15 | <Hixie> | they already do, no? |
| 09:16 | <zcorpan_> | doesn't the spec just talk about the button's label? |
| 09:16 | <Hixie> | hm, maybe the spec doesn't require it? |
| 09:16 | <Hixie> | i dunno |
| 09:16 | <Hixie> | file bugs if the spec is wrong |
| 09:16 | <zcorpan_> | "If the element has a value attribute, the button's label must be the value of that attribute; otherwise, it must be an implementation-defined string that means "Submit" or some such." |
| 09:16 | <zcorpan_> | what does "it" refer to? |
| 09:18 | <zcorpan_> | i guess it can only refer to "the button's label" |
| 09:18 | zcorpan_ | looks at the form submission algorithm |
| 09:23 | <zcorpan_> | "The value IDL attribute is in mode default." |
| 09:24 | <zcorpan_> | "On getting, if the element has a value attribute, it must return that attribute's value; otherwise, it must return the empty string." |
| 09:26 | <zcorpan_> | "Each input element has a value, which is exposed by the value IDL attribute." |
| 09:27 | <zcorpan_> | "Otherwise, append an entry in the form data set with name as the name, the value of the field element as the value, and type as the type." |
| 09:29 | <zcorpan_> | is that right? |
| 09:31 | <hsivonen> | Hixie: it seems that Gecko already submits the localized string for "Submit" |
| 09:32 | <hsivonen> | it seems like a bad idea to make the submission dependent on localization, but I guess that's the way the legacy is |
| 09:34 | <Hixie> | does it affect the DOM also? |
| 09:34 | <Hixie> | oh, crap, the IETF is having a meeting again, so I can't update the I-D |
| 09:34 | <Hixie> | man, the IETF is really annoying |
| 09:35 | <zcorpan_> | opera even sets the content attribute value="Submit" |
| 09:35 | <zcorpan_> | but that's a bug |
| 09:39 | <zcorpan_> | Hixie: don't we want to have the same requirement for <title> and the others? |
| 09:40 | Philip` | sees that websocket has long surpassed http://tools.ietf.org/html/draft-ietf-dnsext-mdns-47 in number of revisions |
| 09:42 | <Hixie> | zcorpan_: hm? |
| 09:43 | <Hixie> | Philip`: it's still far behind html5! |
| 09:43 | <zcorpan_> | Hixie: the restriction you added to <style> |
| 09:43 | <Hixie> | i don't really want to have it on <style>, let alone anything else |
| 09:43 | gsnedders|work | wonders what happens after 99 |
| 09:43 | <Philip`> | gsnedders|work: 100 |
| 09:44 | <Philip`> | gsnedders|work: I thought you knew how to count |
| 09:44 | gsnedders|work | wonders if any I-D has ever hit that |
| 09:44 | <Philip`> | gsnedders|work: websocket seems to be the highest that I can see, and dnsext-mdns is second |
| 09:44 | <zcorpan_> | Hixie: ok. i guess it's mostly just a problem for script and style anyway |
| 09:44 | <gsnedders|work> | Philip`: But they're deleted after six months, so that's not very interested |
| 09:44 | <gsnedders|work> | *interesting |
| 09:45 | <Philip`> | Hixie: Sadly HTML5 still has far to go before catching up with KDE in number of revisions |
| 09:46 | <Hixie> | There's only one of me :-P |
| 09:46 | <Philip`> | gsnedders|work: Deleted from tools.ietf.org too? |
| 09:46 | <gsnedders|work> | Philip`: Ah, no, they aren't. |
| 09:46 | <gsnedders|work> | Hixie: Clone yourself. |
| 09:47 | Philip` | was looking at stuff like http://www.google.co.uk/search?q=site%3Atools.ietf.org+inurl%3Ahtml%2Fdraft+inurl%3A30 |
| 09:51 | <hsivonen> | hrm. VideoPress supports Theora but their front page tries to serve Flash to Minefield |
| 09:55 | <zcorpan_> | hsivonen: because Minefield isn't Firefox 3.5? |
| 09:55 | <hsivonen> | zcorpan_: dunno |
| 10:00 | <Hixie> | Philip`: for 8047, any reason I couldn't just invoke the WebIDL algorithm directly? |
| 10:04 | <Philip`> | Hixie: That would be a type error, because the WebIDL algorithm expects an ECMAScript value whereas HTML5 has an unlimited-precision decimal number |
| 10:04 | <Hixie> | yeah, but other than that |
| 10:05 | <DrEvil> | can web workers 'watch' page transitions? |
| 10:05 | <Hixie> | DrEvil: shared workers can listen to what their pages tell them, but they have no view to the outside world other than that |
| 10:05 | <Philip`> | and I think the type error has practical consequences, because it affects how rounding is computed |
| 10:05 | <Hixie> | k |
| 10:06 | <DrEvil> | Hixie: so a web worker cannot watch to see the window location changing? |
| 10:06 | <Hixie> | correct |
| 10:06 | <Hixie> | a |
| 10:06 | <Hixie> | a dedicated worker will be shut down when the pages moves on |
| 10:06 | <Hixie> | a shared worker will live on if other pages are talking to it only |
| 10:07 | <Hixie> | Philip`: can you think of any reason why the parsing algorithm shouldn't just always convert to a float? |
| 10:07 | <Philip`> | Maybe it'd be much easier to replace the whole float-parsing thing with ECMAScript ToNumber, and have it return a proper floating-point number rather than an impossible hypothetical unlimited-precision number |
| 10:07 | <DrEvil> | hmm so there is not a way for which to view informations on page changes with browsers? |
| 10:08 | <Hixie> | DrEvil: how do you mean? |
| 10:08 | <DrEvil> | say for analysing requests and responses from server |
| 10:09 | <Hixie> | DrEvil: what are you trying to do? |
| 10:09 | jgraham | thought that most ES algorithms were written in terms of abstract, unlimited precision numbers and converted to concrete representations at the last possible moment |
| 10:09 | <webben> | DrEvil: Isn't that what your server access logs do? |
| 10:09 | <DrEvil> | on the client I mean |
| 10:09 | <DrEvil> | with javascript |
| 10:09 | <jgraham> | (but I don't recall exactly what you are talking about and I don't fedel incline to check unless someone wants me to) |
| 10:09 | <Philip`> | jgraham: I don't know about most algorithms, but it defines how to convert a string into a Number (i.e. a 64-bit floating point value), and WebIDL defines how to convert a Number to an IDL float (32-bit) |
| 10:09 | <webben> | DrEvil: Sure there is. Firebug and LiveHTTPHeaders do so, for instance. They just don't run out of a page. |
| 10:09 | <Philip`> | which seems to be what's needed here |
| 10:10 | <DrEvil> | webben: I am looking for something to run as code on demand not to be requiring addon |
| 10:10 | <jgraham> | Philip`: So you're saying that HTML5 should just work in terms of some concrete representation rather than an abstract number? |
| 10:11 | <webben> | DrEvil: There's no reason a browser can't incorporate those features if it wants. |
| 10:11 | <DrEvil> | as an example - I want to create a javascript application which can be included into page - it will read the details of requests/response for page and provide a status bar with details at the bottom of the pgae |
| 10:11 | <DrEvil> | but I cannot see this information with javascrit |
| 10:11 | <DrEvil> | I cannot do with workers? |
| 10:12 | <webben> | DrEvil: You can see it if your script is making the requests. |
| 10:12 | <DrEvil> | but if the browser location is changing? |
| 10:12 | <DrEvil> | I want to host the script and have authors include this one script in their pages |
| 10:12 | <Philip`> | jgraham: The view ES/WebIDL takes (if I'm not mistaken) is that floating-point numbers are an abstract concept of a finite set of real numbers, and the byte-level representation is irrelevant |
| 10:13 | <Philip`> | jgraham: so it's still non-concrete |
| 10:13 | <DrEvil> | but at the moment I cannot see how it is I can get this information with javascript |
| 10:13 | <webben> | DrEvil: It sounds to me like allowing any old script in a page to capture requests/responses even after the page was closed would contravene the browser security model left right and center, and is it's easy for developers to use tools like Firebug, I don't really see what real-world problem you are solving. |
| 10:13 | <webben> | *as it's |
| 10:14 | <DrEvil> | I don't think that is much security concern |
| 10:15 | <DrEvil> | maybe for request you would remove authorization |
| 10:15 | <webben> | Really? What would be the limitation on someone including a script in a page and using it to track all my subsequent movements? |
| 10:15 | <DrEvil> | but for response it is not risk |
| 10:15 | <Philip`> | I think I'm saying that instead of HTML5 defining how to parse a string into an unlimited-precision number, and then defining how to convert it to a member of the finite floating-point set of numbers for IDL, it should instead parse the string into a member of the finite set directly in the same way that ES does |
| 10:15 | <Philip`> | (and then the conversion to ES Number is trivial) |
| 10:15 | <DrEvil> | what http headers are you thinking of in response that present risk? |
| 10:15 | <Philip`> | (and WebIDL handles the conversion to IDL float) |
| 10:15 | <webben> | DrEvil: All of them. It's /my/ traffic. Not to be sniffed by random websites I visit. |
| 10:15 | <Philip`> | (unless I'm getting stuff mixed up) |
| 10:16 | <DrEvil> | it's not sniffed random the author has to chose to include script |
| 10:16 | <DrEvil> | so this is correct - it is possible for author to design now custom solution with on click events for links |
| 10:16 | <webben> | DrEvil: Yes. But you want the sniffing to continue once one has left the page. |
| 10:17 | <DrEvil> | no for response information available each page |
| 10:17 | <DrEvil> | individually |
| 10:17 | <jgraham> | Philip`: I see. There is some level of concreteness in Ecmascript through the to (U)Int algorithms which are needed for things like bitwise operators applied to Numbers |
| 10:17 | <webben> | DrEvil: You asked what happens if .location changes - that's leaving the page. |
| 10:17 | <DrEvil> | this is interesting because some sites they use URI for sensitive information - this is available already to javascript |
| 10:17 | <jgraham> | Philip`: But you plan sounds like a rather sensible approach |
| 10:18 | <webben> | DrEvil: Not to scripts loaded on previously visited pages. |
| 10:18 | <hsivonen> | hrm. XiphQT doesn't with QuickTime Player X |
| 10:18 | <DrEvil> | I do not need to scripts on previous pages I was thinking more to find a solution |
| 10:18 | <DrEvil> | if it is possible some other way that my javascript can include on the page and get that inforamtion each time that is better |
| 10:19 | <DrEvil> | so I can access the response information - more than just location |
| 10:19 | <webben> | DrEvil: If you "do not need to scripts on previous pages", what do you mean by tracking traffic across .location changes? |
| 10:19 | <Philip`> | (Oh, I might be a little mixed up - if the IDL uses float, then HTML5 probably needs the value to be part of the set of 32-bit floats. Or HTML5 could just use double.) |
| 10:19 | <gsnedders|work> | hsivonen: AFAIK it should work if you have QT7 installed |
| 10:19 | <DrEvil> | I say wrong - I mean to say that I wish to view response information on each page |
| 10:19 | <hsivonen> | gsnedders|work: yes, I noticed that |
| 10:20 | <webben> | DrEvil: What if a page requests a resource (e.g. an iframe src) on another domain? |
| 10:20 | <gsnedders|work> | (QT X has no native support extensions, and just uses QTKit for unknown codecs if it is installed) |
| 10:20 | <webben> | DrEvil: Sniffing that would contravene same-origin policies. |
| 10:20 | <gsnedders|work> | s/ if it is installed// |
| 10:20 | <webben> | DrEvil: Your use-case is handled so much better by tools like Firebug. |
| 10:21 | <DrEvil> | webben: I am not understanding this is relevant I am not saying sniffing |
| 10:21 | <hsivonen> | gsnedders|work: do QT X and QT 7 have totally different back ends? |
| 10:21 | <gsnedders|work> | hsivonen: As I understand it, yes. |
| 10:21 | <DrEvil> | I meant to say that when a page loads I wish to see the information after the fact |
| 10:21 | <DrEvil> | from the response |
| 10:21 | <webben> | DrEvil: s/sniffing/viewing/ |
| 10:21 | <gsnedders|work> | hsivonen: Both can be used via QTKit, and QTKit just chooses the best for what codec you want |
| 10:21 | <DrEvil> | so I can show maybe ETag headers.. ? |
| 10:21 | <DrEvil> | or something like this |
| 10:21 | <hsivonen> | gsnedders|work: I see. That wasn't exactly clear from marketing material or from the Snow Leopard installer |
| 10:22 | <DrEvil> | I have some application which it presents standard information in the server responses |
| 10:22 | <DrEvil> | in the headers |
| 10:22 | <DrEvil> | if I am use XHR I can get this |
| 10:22 | <gsnedders|work> | (and for anything apart from the default built-in codecs in QT X, that means using QT7) |
| 10:22 | <hsivonen> | gsnedders|work: and QT Player X deliberately configures QTKit never to use the QuickTime 7 back end? that's odd |
| 10:22 | <gsnedders|work> | hsivonen: No |
| 10:22 | <webben> | DrEvil: Yes I understand what you're saying - but it's totally undesirable from a security/privacy standpoint AFAICT. |
| 10:22 | <DrEvil> | but for new page javascript cannot see - all javascript sees is the location |
| 10:23 | <DrEvil> | webben: what information in response you are talking about security ? |
| 10:23 | <hsivonen> | why doesn't QT Player X choose the QT7 back end for ogg, then? |
| 10:23 | <gsnedders|work> | hsivonen: It should |
| 10:23 | <hsivonen> | hmm. |
| 10:23 | <webben> | DrEvil: Whereas in-built browser tools or add-ons like Firebug can do more and do it better and without damaging security/privacy |
| 10:23 | <DrEvil> | webben: there must be trust given to code on demand |
| 10:23 | <hsivonen> | does the new back end have extension APIs at all or just new extension APIs? |
| 10:23 | <DrEvil> | has to be there is sensitive information in pages which use javascript |
| 10:24 | <gsnedders|work> | hsivonen: It has none AFAIK |
| 10:24 | <webben> | DrEvil: On the contrary, such code should be distrusted as much as possible. |
| 10:24 | <DrEvil> | this is just more of information in response that is in headers |
| 10:24 | <DrEvil> | not more or less sensitive |
| 10:24 | <DrEvil> | just more information |
| 10:24 | <hsivonen> | gsnedders|work: well that's an interesting change of policy |
| 10:24 | <DrEvil> | if it is html page I can still use javascript extract values from page elements |
| 10:24 | <hsivonen> | Snow Leopard installer didn't make it exactly clear what the "older media formats" that require QuickTime Player 7 are |
| 10:24 | <DrEvil> | so what of the headers - they are just information to me |
| 10:24 | <DrEvil> | I want to understand what it is in the headers you are worrying about |
| 10:24 | <hsivonen> | QT7 back end is installed by default even though the player isn't, right? |
| 10:24 | <DrEvil> | so I can think about this |
| 10:25 | <webben> | DrEvil: Yes. However, if there's an iframe on the page and its from another domain you cannot introspect its DOM. |
| 10:25 | <hsivonen> | since the QT7 Player option was just a couple of MB |
| 10:25 | <webben> | DrEvil: It would be similarly inappropriate to allow scripts to introspect its headers. |
| 10:25 | <DrEvil> | no - this would not be the case either for javascript is only capable of seeing request information from its current window |
| 10:25 | <DrEvil> | same as location etc. |
| 10:26 | <webben> | DrEvil: So you're not asking for the ability to view HTTP information for resources included in the page from other domains? |
| 10:26 | <DrEvil> | no |
| 10:26 | <DrEvil> | just for current page |
| 10:26 | <DrEvil> | window.XHResponse.getHeader('Etag') |
| 10:26 | <webben> | DrEvil: When you say the current page... do you mean the HTML page as opposed to the resources it includes? |
| 10:26 | <DrEvil> | or something like this |
| 10:27 | <DrEvil> | I mean for the URI really |
| 10:27 | <webben> | I see. |
| 10:27 | <hsivonen> | I guess I should have read the John Siracusa treatment of SL |
| 10:27 | <DrEvil> | for the response but html/javascript is main use case |
| 10:27 | <DrEvil> | I assume this is good place asking? |
| 10:27 | <DrEvil> | I will stop if that is wrong |
| 10:27 | <webben> | It's not a bad place. |
| 10:28 | <DrEvil> | but this makes sense yes? it is the same as window.location |
| 10:28 | <DrEvil> | but for the other metadata in the response also |
| 10:28 | <DrEvil> | I cannot see the way to do this now |
| 10:28 | <Hixie> | nn |
| 10:29 | <webben> | DrEvil: Can you think of a use-case other than creating a poor man's Firebug/wireshark? |
| 10:29 | <DrEvil> | for some standard of server responses and metadata framework |
| 10:29 | <DrEvil> | for standard information on a resource |
| 10:30 | <DrEvil> | create standard widget which inspect response on page load |
| 10:30 | <webben> | for what? |
| 10:30 | <DrEvil> | provide metadata widget |
| 10:30 | <DrEvil> | from information in headers |
| 10:30 | <webben> | why would a user want a "metadata widget from information in headers"? |
| 10:30 | <DrEvil> | because there are standards which are using this |
| 10:31 | <DrEvil> | such as new OCCI standard using this |
| 10:31 | <DrEvil> | and they make use http headers so they page load but need to render the metadata to a page for users |
| 10:31 | <webben> | DrEvil: What metadata that a user would want to see is being send as an HTTP response header? |
| 10:31 | <webben> | *sent |
| 10:32 | <DrEvil> | they have standards such as resource ID, link relations, resource category |
| 10:32 | <DrEvil> | this is something that can be in html |
| 10:32 | <webben> | Indeed. |
| 10:32 | <DrEvil> | but mostly for display authorative is headers |
| 10:32 | <DrEvil> | so for system pruposes and standard |
| 10:32 | <DrEvil> | easier if standard library which is included in page for this |
| 10:32 | <webben> | Doesn't sound like it needs changes to the DOM ... just put the information in the body of the response. |
| 10:33 | <webben> | That way, it will work without scripting. |
| 10:33 | <webben> | So the end-user benefits. |
| 10:33 | <DrEvil> | yes it is possible to do this but to go with standard it will need analyse the headers |
| 10:34 | <DrEvil> | so user can use browser viewing raw data as machine |
| 10:34 | <DrEvil> | also html is useful but to view true headers is necessary use case |
| 10:35 | <DrEvil> | there is to say there is not to be this in browser ever a solution? |
| 10:35 | <DrEvil> | :( |
| 10:35 | <webben> | DrEvil: I'm not sure I understand what you're saying |
| 10:36 | <annevk2> | probably not a generic solution for a long time |
| 10:36 | <DrEvil> | this is information which is there for the browser to present to javascript |
| 10:36 | <annevk2> | for reading response headers |
| 10:36 | <DrEvil> | yes? |
| 10:36 | <webben> | DrEvil: Could you link to the standard which says you must use HTTP headers not the body to present the information? |
| 10:36 | <DrEvil> | yes ok 2 mins |
| 10:38 | <DrEvil> | http://occi.googlecode.com/hg/docs/occi-core.html |
| 10:38 | <DrEvil> | metadata standardised with headers |
| 10:38 | <DrEvil> | this means all metadata can be provided regardless of format |
| 10:39 | <DrEvil> | so can still have useful response for jpeg/pdf/html/OVF/etc. |
| 10:39 | <DrEvil> | better standardise this way gives more flexibility for providers |
| 10:40 | <DrEvil> | annevk2: for a long time? |
| 10:40 | <DrEvil> | a couple of weeks? |
| 10:40 | <DrEvil> | :)) |
| 10:40 | <DrEvil> | I would be interested to know how long this is you are thinking |
| 10:40 | <DrEvil> | that is bad news if it is very very long time, there are many people look now to use headers for this purpose |
| 10:41 | <annevk2> | many different nicks which I suspect are the same person certainly are |
| 10:41 | <annevk2> | other than that, not so much |
| 10:42 | <DrEvil> | ? |
| 10:42 | <DrEvil> | occi spec is example |
| 10:43 | <DrEvil> | there is other case like couchdb use etag and things like this |
| 10:43 | <DrEvil> | amazon web service also |
| 10:43 | <DrEvil> | annevk2: you not say how much time you were meaning |
| 10:44 | <annevk2> | a decade or so |
| 10:44 | <DrEvil> | oh wow - why do you think this? |
| 10:44 | <DrEvil> | it is not easy just ammend where window.location to include also window.response something |
| 10:45 | <annevk2> | no |
| 10:45 | <DrEvil> | :( what are the problems here |
| 10:46 | <webben> | DrEvil: That OCCI spec specifically mentions in-band link relationships in HTML, I don't see a requirement to keep them in HTTP only. |
| 10:46 | <DrEvil> | webben: that is fine but does not to say it is more useufl to be using the authorative information which is headers |
| 10:46 | <DrEvil> | it is leaky |
| 10:47 | <DrEvil> | please can I ask for some more information why it is this is not happening for decade |
| 10:47 | <webben> | DrEvil: Sorry "does not to say it is more useufl to be using the authorative information which is headers" ... where does it say that's the headers are more authorative? |
| 10:47 | <DrEvil> | well this is the spec is defined against this method |
| 10:48 | <DrEvil> | they are required so it is natural this is cae |
| 10:48 | <DrEvil> | case |
| 10:48 | <annevk2> | DrEvil, there's hardly any demand so I sort of doubt it will happen a decade from now to be honest |
| 10:48 | <annevk2> | DrEvil, there's more important things to work on and worry about |
| 10:48 | <webben> | DrEvil: Where does the spec say you must supply metadata out-of-band rather than in-bad? |
| 10:48 | <webben> | *in-band |
| 10:48 | <DrEvil> | it is for making tool consistent |
| 10:48 | <DrEvil> | this information must be consistent format for headers |
| 10:49 | <webben> | DrEvil: for example the spec says "The root (“/”) should expose collections in-band and/or out-of-band in order for clients to discover resources." note the "and/or" |
| 10:49 | <DrEvil> | that is data not metadata |
| 10:49 | <DrEvil> | no wonder |
| 10:49 | <DrEvil> | annevk2: how it is not important to do this I don't think |
| 10:50 | <DrEvil> | many seeing more use in this area would be easier making browsers more easy to be involved |
| 10:50 | <DrEvil> | it is not so hard to make this object available for scripts, no? |
| 10:50 | <DrEvil> | what is the difficulty for this? |
| 10:51 | <annevk2> | added complexity |
| 10:51 | <DrEvil> | it is complex to do this? really? |
| 10:51 | <annevk2> | yes, each feature adds complexity |
| 10:51 | <webben> | DrEvil: It seems like the approach you are suggesting is actually user-hostile since it encourages publishers to make the display of information the user is interested in dependent on scripts. |
| 10:51 | <DrEvil> | yes that is obvious statement if I not to be rude |
| 10:51 | <annevk2> | and could potentially be some kind of attack vector, etc. |
| 10:51 | <annevk2> | yes it is |
| 10:52 | <annevk2> | so we need to evaluate which features to add based on their merit |
| 10:52 | <annevk2> | to only increase complexity so much |
| 10:52 | <DrEvil> | hmm that is sad it is maybe moving development idea faster if possible |
| 10:53 | <DrEvil> | thinking of news things to do |
| 10:53 | <DrEvil> | but cannot because browsers too slow to react |
| 10:53 | <DrEvil> | no? |
| 10:53 | <annevk2> | I've a hard time following your ramblings |
| 10:53 | <DrEvil> | better browser focus low level see as open platform |
| 10:53 | <DrEvil> | allow build on top |
| 10:53 | <DrEvil> | bottom up |
| 10:54 | <DrEvil> | not making complicated solution top down that is slow to react in fast pace world |
| 10:54 | <annevk2> | I suggest you start here: http://wiki.whatwg.org/wiki/FAQ#Is_there_a_process_for_adding_new_features_to_a_specification.3F |
| 10:54 | <DrEvil> | that is bad for me for everyone |
| 10:54 | <DrEvil> | yes I am just talking with you |
| 10:54 | <DrEvil> | I can read processes but processes seem they are the problem no? |
| 10:55 | <annevk2> | no idea, but your ramblings won't get you anywhere |
| 10:55 | <annevk2> | that piece of process might |
| 10:55 | <DrEvil> | I see that browsers they have a bad time because there are many features - but I think approach is old and bad |
| 10:56 | <DrEvil> | need more open and start at the basics |
| 10:56 | <DrEvil> | provide a platform for build complexity |
| 10:56 | <annevk2> | sure sure, brave new world thinking is always fun |
| 10:56 | <DrEvil> | I think the world it is too rapid for your beuracracy |
| 10:56 | <DrEvil> | you are slowing progression I think |
| 10:56 | <annevk2> | but that ain't gonna fly if we want today's web to work as well |
| 10:57 | <annevk2> | so route around me |
| 10:57 | <DrEvil> | I think disruption sometimes it is necessary |
| 10:57 | <DrEvil> | maybe we fire you all and make new browsers |
| 10:57 | <DrEvil> | :D |
| 10:57 | <annevk2> | go for it |
| 10:57 | <DrEvil> | it is a shame that is so difficult - maybe if we have operating systems more for the web |
| 10:57 | <DrEvil> | you will hopefully become less important |
| 10:57 | <DrEvil> | and the world will be better more advancing place faster |
| 10:58 | <DrEvil> | not waiting for you to decide all the time |
| 10:58 | <DrEvil> | why it is not just to make the very essentials easy accessible and programmable? |
| 10:58 | <DrEvil> | you say there is no demand but that may be more reflection of supply |
| 10:58 | <DrEvil> | and that is you.. no? |
| 10:59 | <annevk2> | what are you still wasting your time here for? go plot and scheme how to destroy the establishment! |
| 10:59 | <Dashiva> | Or just make a browser |
| 11:00 | <DrEvil> | I tihnk google are best direction |
| 11:00 | <DrEvil> | I not do a lot better than they are |
| 11:00 | Philip` | wonders why annevk2 is being unusually hostile |
| 11:00 | <DrEvil> | he is homo |
| 11:00 | <DrEvil> | lady name |
| 11:01 | <DrEvil> | i joke this is not me I am friendly just like to talk about this to understand |
| 11:01 | <Philip`> | That's not a productive way to continue the conversation either :-p |
| 11:02 | <DrEvil> | I think interesting to see why not to implement the core functionality for web |
| 11:02 | <DrEvil> | like this and then build on top |
| 11:02 | <annevk2> | Philip`, getting tired of unstructured ramblings mostly |
| 11:02 | <Dashiva> | It seems to end up that way whenever HTTP is invoked |
| 11:03 | <DrEvil> | I see |
| 11:03 | <DrEvil> | HTTP it is the web mostly! |
| 11:03 | <annevk2> | DrEvil, the reason we have process is so that not every feature that is being proposed has to be done |
| 11:03 | <ttepasse> | Given that Last Call is dooming it's probably to late to repropose <line>, isn't it? |
| 11:03 | <Dashiva> | That misconception is probably a big part of it |
| 11:03 | <annevk2> | DrEvil, the process is there so that feature proposals get more structured and can be more easily evaluated against other proposals |
| 11:03 | <DrEvil> | Dashiva: that is not mis-thinking that is true |
| 11:03 | <annevk2> | DrEvil, it makes it clear what problems are being solved, what the use cases are, etc. |
| 11:04 | <Dashiva> | HTTP is just a transfer protocol, it's largely irrelevant |
| 11:04 | <DrEvil> | annevk2: that is nice politician talking |
| 11:04 | <annevk2> | DrEvil, now if you think that is not needed I'm afraid you won't get much further here |
| 11:04 | <DrEvil> | Dashiva: HTTP is irrelevant?! that is crazy man :) |
| 11:05 | <DrEvil> | annevk2: I think that is poor justification and protective approach for beauracracts |
| 11:05 | <DrEvil> | to be important and decide things |
| 11:05 | <Dashiva> | Sure, it's the one we happen to use |
| 11:05 | <Dashiva> | But there's nothing special about HTTP |
| 11:05 | <DrEvil> | I wonder why1 |
| 11:05 | <gavin> | "http is irrelevant" and "http is irrelevant to whether I want peanut butter on my sandwich" are two very different statements |
| 11:05 | <DrEvil> | really? |
| 11:05 | <gavin> | I think dashiva's statement was a variant of the latter :) |
| 11:06 | <DrEvil> | if http is irrelevant |
| 11:06 | <DrEvil> | why it is you add PUT/DELETE? |
| 11:07 | <DrEvil> | I am saying something you all dead now? |
| 11:07 | <Dashiva> | That's the thing in a nutshell, isn't it |
| 11:07 | <Dashiva> | The web is adding PUT/DELETE |
| 11:07 | <DrEvil> | no http has had this for since 1.1 |
| 11:07 | <Dashiva> | Yes, and nobody cared for a long time |
| 11:07 | <DrEvil> | this is something html has not had |
| 11:08 | <DrEvil> | Dashiva: I think people have cared for while but more since XHR it came to fix a little bit the shitness of HTML |
| 11:08 | <DrEvil> | but there is still many problems |
| 11:09 | <DrEvil> | html it is very having lots of shitness |
| 11:09 | <Dashiva> | Roy would be proud of you |
| 11:09 | <DrEvil> | you are angry people with nothing to say |
| 11:09 | <DrEvil> | you attack |
| 11:09 | <DrEvil> | that is all |
| 11:09 | <DrEvil> | do you listen? |
| 11:09 | <DrEvil> | no because you are the most important |
| 11:10 | <DrEvil> | I think browsers and html have such big responsibility but you do not have ability to fulfill this |
| 11:10 | <Dashiva> | Now that's just asking to be taken out of context :) |
| 11:10 | <DrEvil> | maybe because of process, people, both |
| 11:10 | <DrEvil> | but you are failing the world |
| 11:11 | <DrEvil> | there is missed opportunity because of your bad processes |
| 11:11 | <DrEvil> | fact. |
| 11:11 | <Dashiva> | Axiomatic proof |
| 11:11 | <DrEvil> | Dork comment |
| 11:12 | <DrEvil> | this is just too bad |
| 11:13 | <DrEvil> | you people do not deserve to be in control if you are not willing to acknowledge the scope of your responsibilities |
| 11:13 | <Dashiva> | Nobody is in control |
| 11:13 | <Dashiva> | You're free to go out and make a new web |
| 11:14 | <DrEvil> | you're an idiot if you think that statement has merit |
| 11:14 | <DrEvil> | I don't think you do. |
| 11:15 | <Dashiva> | Oh, but it does |
| 11:15 | <Dashiva> | Becausae HTTP is most important, right? |
| 11:15 | <Dashiva> | You can just add some minor stuff on top of it and you're done |
| 11:15 | <DrEvil> | no HTTP is just a foundation |
| 11:15 | <DrEvil> | that is unavoidable fact |
| 11:15 | <DrEvil> | the same as TCP is important |
| 11:15 | <DrEvil> | unavaoidable |
| 11:15 | <DrEvil> | I do not see why it is you contest this |
| 11:16 | <DrEvil> | you have HTTP notions in your spec |
| 11:16 | <DrEvil> | (apart from your pathetic conception of resources) |
| 11:16 | <DrEvil> | oh look.. |
| 11:16 | <DrEvil> | make anothe rstupid comment to avoid the point |
| 11:16 | <DrEvil> | pathetic. |
| 11:17 | <DrEvil> | like it or not, HTTP is a brilliant piece of engineering and is the reason the web works |
| 11:17 | <Dashiva> | Do you have a point? |
| 11:17 | <DrEvil> | HTTP/URI's lead to hypermedia formats |
| 11:17 | <DrEvil> | html happens to be the one that took off |
| 11:18 | <DrEvil> | it's shit. |
| 11:18 | <DrEvil> | and there's virtually nothing in it that's of massive complexity or importance |
| 11:18 | <DrEvil> | all the clever stuff is in the protocols |
| 11:18 | <DrEvil> | boo hoo |
| 11:18 | <Dashiva> | The importance is in the _content_ |
| 11:19 | <DrEvil> | you can get sand in your vadge over that if you want |
| 11:19 | <DrEvil> | I don't care |
| 11:19 | <DrEvil> | no actually it's |
| 11:19 | <DrEvil> | *information* |
| 11:19 | <DrEvil> | and |
| 11:19 | <Dashiva> | No, it's the content. Really. |
| 11:19 | <DrEvil> | *relationships* |
| 11:19 | <DrEvil> | No. Really. Socrates. It's not. |
| 11:19 | <Dashiva> | You may have some kind of distorted HTTP view of reality |
| 11:19 | <DrEvil> | you can live in your robotic dork world |
| 11:19 | <Dashiva> | But the people use the web for content |
| 11:19 | <DrEvil> | where that is your reality |
| 11:19 | <DrEvil> | but it's bullshit, and you're in idiot. |
| 11:20 | <DrEvil> | people use the web |
| 11:20 | <DrEvil> | they dont use html |
| 11:20 | <DrEvil> | the web is related information |
| 11:20 | <DrEvil> | it's not "lots of amazing wonderful html document" |
| 11:20 | <DrEvil> | nobody gives a shit. |
| 11:20 | <DrEvil> | XML/JSON/RIAs |
| 11:20 | <DrEvil> | Ajax |
| 11:20 | <beowulf> | is DrEvil a bot? |
| 11:21 | <hsivonen> | beowulf: DrEvil is mookid |
| 11:21 | <DrEvil> | I did a really good job of pretending to be foreign though |
| 11:21 | <DrEvil> | give me cretdit |
| 11:21 | <Dashiva> | No, not really |
| 11:21 | <zcorpan_> | Hixie: where's the data about <meta content-language>? |
| 11:21 | <DrEvil> | I'm not actually mookid btw |
| 11:21 | <Dashiva> | You have content-type:troll/mookid written all over you |
| 11:21 | <beowulf> | DrEvil: foreign? |
| 11:22 | <DrEvil> | yes it is correct I am make you think that |
| 11:22 | <beowulf> | DrEvil: foreign to where? |
| 11:22 | <beowulf> | or whom? |
| 11:22 | <DrEvil> | yeah whatever foreign to where I am actually located |
| 11:23 | <DrEvil> | smarty pants. |
| 11:23 | <beowulf> | i still think it's a bot |
| 11:23 | <DrEvil> | if it is the singularity just happened |
| 11:23 | <DrEvil> | because I am on FIRE right now |
| 11:23 | <ment> | hi, anybody fluently speaking html5? |
| 11:23 | <DrEvil> | yep <bag of shit /> |
| 11:23 | <ment> | i think i don't understand how <LI> open tag algorithm works, (9.2.5.10 The "in body" insertion mode). |
| 11:24 | <Dashiva> | What part of it is unclear? |
| 11:24 | <ment> | becouse it seems the algorithm does nothing |
| 11:24 | <jgraham> | ment: What are you expecting it to do? |
| 11:25 | <ment> | it skips formatting and phrasing elements to produce "node" variable and then it does nothing with it |
| 11:25 | <Dashiva> | It uses it to close previous <li> elements |
| 11:25 | <Dashiva> | Step 3 |
| 11:27 | <DrEvil> | well I hope you're going to apologise for how you've treated me this morning |
| 11:27 | <DrEvil> | I feel very upset |
| 11:27 | <DrEvil> | how dare you accuse me of caring about the application protocol web developers work with |
| 11:28 | <DrEvil> | and then daring to relate that to html |
| 11:28 | <ment> | Dashiva: 3) close LI if on top 4) if (node->tag not in Formatting or Phrasing) goto 6; 5) node = previous node on stack; goto 4; 6) this is the last step 7) insert_li_on_top() |
| 11:28 | <Dashiva> | No, goto 3 |
| 11:28 | <DrEvil> | goodbye cruel world! |
| 11:28 | <Dashiva> | Where node is reused |
| 11:29 | <ment> | Dashiva: 5. Otherwise, set node to the previous entry in the stack of open elements and return to the step labeled loop. |
| 11:29 | <annevk2> | ment, the node is inserted in step 6 |
| 11:29 | <Dashiva> | ment: And the labeled step is #3 |
| 11:29 | <annevk2> | ment, the node variable is just a temporary thing to check various conditions |
| 11:29 | <ment> | Dashiva: no, as #3 |
| 11:29 | <ment> | hrmz, maybe i have just the wrong version |
| 11:30 | <Dashiva> | Which page are you using? |
| 11:30 | annevk2 | is looking at http://html5.org/spec |
| 11:30 | <Dashiva> | http://www.whatwg.org/specs/web-apps/current-work/multipage/tokenization.html#parsing-main-inbody |
| 11:30 | <ment> | http://www.w3.org/TR/html5/syntax.html#parsing-main-inbody |
| 11:30 | <Dashiva> | Ah, that one is outdated |
| 11:30 | <Dashiva> | Very much so |
| 11:30 | <annevk2> | but not for this algo it seems |
| 11:31 | <zcorpan_> | TR/ should be called outdated/ |
| 11:31 | <Dashiva> | ment: In that version, your issue is indeed present |
| 11:32 | <Dashiva> | But it seems to have been fixed since then |
| 11:32 | <ment> | are there any changelogs for the whatwg specs to see what is different? |
| 11:32 | <gsnedders|work> | ment: http://html5.org/tools/web-apps-tracker |
| 11:32 | <Dashiva> | Yes, but there's a ton to read :) |
| 11:32 | <annevk2> | Dashiva, what is different? |
| 11:33 | <Dashiva> | annevk: The loop label has moved from step 4 to step 3 |
| 11:33 | <ment> | gsnedders|work: thanks |
| 11:33 | <annevk2> | aaah |
| 11:34 | <ment> | however, any _big_ changes in html5 parser? (for example something like, we abandoned adoption agency algorithm, we are using something completely different and three times as complicated instead) |
| 11:34 | <DrEvil> | couldn't you just stop talking crap in here and go implement my window.response object ? |
| 11:34 | <DrEvil> | that would be far more productive |
| 11:35 | <DrEvil> | please, I forgot to say please |
| 11:35 | <gsnedders|work> | ment: Depends how "big" is big :) |
| 11:35 | <Dashiva> | ment: There's been some back-and-forth about escaped text in <script> and related elements |
| 11:35 | <gsnedders|work> | ment: I guess all the new script stuff, the removal of insertion modes, is big |
| 11:36 | <Dashiva> | Table taint is gone, I think? |
| 11:36 | <gsnedders|work> | s/insertion modes/content model flags/ |
| 11:37 | <gsnedders|work> | Yeahj |
| 11:37 | <ment> | aw, i like content model flags :/ |
| 11:37 | <gsnedders|work> | resetting the insertion mode has changed |
| 11:37 | <Dashiva> | I think the changes have been in the direction of less complexity |
| 11:37 | <Dashiva> | Not more |
| 11:38 | <gsnedders|work> | Yeah, it is also more similar to how you wanted to have it for perf reasons before |
| 11:39 | <gsnedders|work> | ment: you implementing the parser? |
| 11:39 | <ment> | well if you want less complexity, the tokenizer could be shortened by few pages by replacing the unrolled state automaton with regexps and lookback buffer :) |
| 11:39 | <ment> | gsnedders|work: i've almost implemented tokenizer and parser |
| 11:40 | <gsnedders|work> | ment: in what language? |
| 11:40 | <ment> | gsnedders|work: C |
| 11:40 | <hsivonen> | ment: the [R]CDATA changes might count as 'big' |
| 11:40 | <gsnedders|work> | ment: That's more complex if you want to implement it without regexp :) |
| 11:40 | <jgraham> | Indeed more complexity != longer description |
| 11:41 | jgraham | wonders why his copy of spotify is behaving so badly today |
| 11:41 | <gsnedders|work> | Mine is behaving fine, FWIW |
| 11:41 | <DrEvil> | http://clouddevelopertips.blogspot.com/2009/10/what-language-does-cloud-speak-now-and.html |
| 11:41 | <gsnedders|work> | (And I guess that's slightly more useful than some random person saying that) |
| 11:42 | <jgraham> | Yeah I guess it's not network problems |
| 11:42 | <jgraham> | then |
| 11:42 | <gsnedders|work> | Or maybe I'm using the entire bandwidth of the office :P |
| 11:42 | <ment> | gsnedders|work: but this description is more complicated to implement, because i have to put the regexps together from state machine (not an pleasant thing to do) and then unroll it again into my representation |
| 11:43 | <Dashiva> | Maybe there can be some informative un-unrolled state examples made when the spec is finalized |
| 11:43 | <jgraham> | ment: If you are making an implementation that is significantly different from the spec it is always more complex to be sure you are right |
| 11:43 | <gsnedders|work> | ment: FWIW, all the implementations thus far wouldn't have gained by having it in regex |
| 11:43 | <Dashiva> | But I believe the unrolled state is better for review |
| 11:44 | <gsnedders|work> | ment: Why do you need to go from spec -> regex -> your repr? Why the middle-step? |
| 11:44 | <ment> | gsnedders|work: http://ibawizard.net/~thement/com.txt |
| 11:45 | jgraham | wonders if that is supposed to be an answer |
| 11:45 | <ment> | gsnedders|work: then i can after <!-- just fill buffer and every time i see '>' i would run this regexp |
| 11:45 | gsnedders|work | wonders if doing it in a regexp way is very performant |
| 11:46 | <ment> | gsnedders|work: of course unrolled into five-line c code |
| 11:46 | <gsnedders|work> | ment: Most of the impls do similar quick grabbing from the input stream, and regex wouldn't have helped |
| 11:47 | <gsnedders|work> | (AFAIK the longest look-behind any have is one character_ |
| 11:47 | <gsnedders|work> | (And the validator/Gecko one has no look-behind or look-ahead) |
| 11:48 | <jgraham> | If anyone sees rubys, can they remind me to talk to him about html5lib-commits? |
| 11:49 | <ment> | gsnedders|work: i need to store the comment in dom tree anyway. but if i wanted DFA, i could just unroll my own from those regexs and it would be much less error prone |
| 11:49 | <gsnedders|work> | DFA? |
| 11:49 | <ment> | deterministic automaton |
| 11:50 | <gsnedders|work> | Ah. I think the basic state is that using regex in places would benefit some and hurt others |
| 11:51 | <hsivonen> | ment: do you mean writing regexps and turning those into automata would be less error prone than hand-implementing the specced state machine? |
| 11:51 | jgraham | can't see why regexps should ever be faster assuming a sufficiently low-level implementation |
| 11:51 | <hsivonen> | jgraham: indeed |
| 11:52 | <hsivonen> | besides, most "regexp" implementations out there are worse than DFA perf-wise |
| 11:52 | <jgraham> | hsivonen: I think the paper that claimed that really only applied to a special case |
| 11:53 | <hsivonen> | jgraham: I mean that the regexps you get from Perl, Java, Python, JS, etc. aren't DFA |
| 11:53 | <jgraham> | At least I recall a forum discussion with a (perl?) person suggesting that the paper focused on a type of regexp which is uncommon and is very bad in real implementations |
| 11:53 | <jgraham> | but ignored the fact that real (non-DFA) implementations are faster for commoner cases |
| 11:54 | <jgraham> | (maybe not faster in terms of algorithmic complexity but actually faster to run) |
| 11:54 | <jgraham> | (but I don't recall the details so I might be wrong) |
| 11:54 | hsivonen | doubts they are faster but they do enable things like capturing parentheses |
| 11:55 | gsnedders|work | thinks you should try slower languages like PHP |
| 11:55 | <gsnedders|work> | but there again, PCRE is fairly easy to segfault, which is suboptimal |
| 11:55 | jgraham | might be mistaken about the whole thing |
| 11:55 | <ment> | http://swtch.com/~rsc/regexp/regexp1.html |
| 11:57 | <ment> | hsivonen: html5 tokenizer is in non-machine readable format, so i have either step-by-step implement it or decompile it into high-level representation (states and regexes) and then compile it into tokenizer |
| 11:58 | <ment> | hsivonen: and the latter is much easier given that in the former i have to explicitly check if some state hasn't "now dance monkey dance" action (and i'm only human) |
| 12:00 | <hsivonen> | having implemented it by hand, I doubt that you could get the buffering just right efficiently by using a generic compilation tool |
| 12:00 | <hsivonen> | I use Java as the abstraction from which I generate C++ |
| 12:01 | <ment> | hsivonen: specification in regexes -> NFA -> DFA, then some state minimization and i've got the _exact_ same tokenizer |
| 12:01 | <ment> | hsivonen: and i have a guarantee my tokenizer doesn't end in endless loop due to missing state in html5 specs |
| 12:02 | <jcranmer> | and you rely on the correctness of three other internal generation schemes |
| 12:04 | <hsivonen> | ment: but the state machine isn't enough. it also needs to store e.g. element names as it traverses over them |
| 12:05 | <jcranmer> | you can do semantic action processing on state transitions |
| 12:05 | <jcranmer> | although that's not amenable to minimization |
| 12:07 | <jgraham> | hsivonen: http://www.perlmonks.org/?node_id=597364 |
| 12:08 | <ment> | jcranmer: not to underestimate you, but those three algorithms has been proven to produce correct result and it's implementation takes few KB instead of half-meg of specs |
| 12:09 | <jcranmer> | I'm reminded of that Knuth quote |
| 12:09 | <jcranmer> | Beware of bugs in the above code; I have only proved it correct, not tried it. -- Donald E. Knuth |
| 12:11 | <hsivonen> | jgraham: well, the Unicode concern is a bit on the side of the less common case. For every DFA operating on Unicode characters, there's an equivalent DFA (potentially with more states) operating on UTF-8 |
| 12:11 | <hsivonen> | jgraham: and the most common regexps out there probably match on ASCII code points anyway |
| 12:12 | <hsivonen> | I can appreciate the utility of backreferences, etc., though |
| 12:12 | gsnedders|work | doesn't see why you can't just compile [^a] to $char != 'a' instead of $char == 'b' || $char == 'c' etc. |
| 12:12 | <jcranmer> | well, backrefs easily make languages context-sensitive |
| 12:13 | <jcranmer> | (.*)\1 is neither regular nor context-free |
| 12:13 | <hsivonen> | jcranmer: sure. Perl "regular expressions" aren't regular |
| 12:13 | <jgraham> | hsivonen: His claim that the real world performance of perl-like implementations will likely be better seems at least plausible |
| 12:14 | <hsivonen> | jgraham: yes, it is plausible. and probably also depends on when expression compilation happens and how often |
| 12:15 | <Philip`> | ment: http://canvex.lazyilluminati.com/svn/tokeniser/tokeniser_spec.ml <-- there's a machine-readable version of the tokeniser :-) |
| 12:15 | <jgraham> | Of course. In any case the general idea that it is dangerous to extrapolate from a single pathological case to general ideas about goodness seems sound |
| 12:15 | <jgraham> | Philip`: s/machine-readable/out-of-date/ |
| 12:16 | <gsnedders|work> | Philip`: r2548? wow. |
| 12:16 | <ment> | Philip`: cute :) |
| 12:18 | <Philip`> | ment: and it can be converted to graphs like http://canvex.lazyilluminati.com/misc/tokeniser_states.png and code like http://philip.html5.org/tools/parser/tokeniser_auto.js etc |
| 12:18 | <Philip`> | It is indeed horrendously out of date :-( |
| 12:20 | <hsivonen> | jgraham: well, at least with DFAs the failure mode is running out of heap during pattern compilation instead of running out of stack space during matching. :-) |
| 12:20 | <Philip`> | So with DFAs your program will fail to execute at all, whereas with NFAs it'll work fine with normal inputs |
| 12:21 | <ment> | that's another thing, the <!doctype>. why not just grab <!([^>]+)> and then parse it in sub-grammar instead of pluging it directly in the main tokenizer? |
| 12:21 | <hsivonen> | ment: to touch the memory fewer times |
| 12:22 | <hsivonen> | ment: anyway, the spec allows you to implement any algorithmic transformation you see fit as long as it doesn't change the black-box behavior |
| 12:22 | <Philip`> | Is it possible to statically determine whether a Perl-style regexp will run in non-exponential time/space on any input? |
| 12:24 | <jgraham> | Philip`: Doesn't perl allow you to run arbitary functions as part of the process? I would guess in those cases the answer is no... |
| 12:24 | <ment> | hsivonen: you don't event know how ridiculous that argument sounds in HTML5 context |
| 12:25 | Philip` | would like to be able to detect bugs like the one that made the html5lib sanitiser exponentially slow |
| 12:25 | <Philip`> | jgraham: That's true, so I was thinking more of other cases :-) |
| 12:25 | <gsnedders|work> | jgraham: I guess if you ignore the e flag |
| 12:26 | <hsivonen> | ment: what's ridiculous? |
| 12:26 | hsivonen | has implemented some things differently from the spec concept but with the same behavior |
| 12:26 | <Philip`> | gsnedders|work: In Perl regexps? That's only relevant for substitutions, not pattern matching |
| 12:26 | <ment> | hsivonen: "to touch the memory" wtf? |
| 12:27 | <ment> | hsivonen: you are running javascript-sqllite-svg virtual machine and now you are going to count cycles? |
| 12:27 | <hsivonen> | ment: in infrastructure of the beast, sure |
| 12:28 | <ment> | hsivonen: moreover, the while(c != '>') add_char(); parse_doctype(); would be still faster given that the doctype string is reasonably small (few KBs tops) |
| 12:29 | <hsivonen> | ment: you are permitted to implement it that way |
| 12:30 | <Philip`> | hsivonen: Your earlier statements seem to be contradictory - if you can implement any black-box-equivalent behaviour, the spec could easily say to grab the whole string and then parse it as a later step, and it wouldn't affect the memory use patterns of implementations |
| 12:31 | <hsivonen> | Philip`: yeah, the real reason why the doctype stuff is in the same state machine is more likely that the idea that > needs to break out even within quotes came relative late in the speccing cycle |
| 12:32 | <Philip`> | ment: Why would that be faster? (I'd have thought that parse_doctype would have to do the same amount of work that an exploded tokeniser would do, and you're also doing an extra loop over the characters) |
| 12:34 | <ment> | Philip`: you don't have to jump back-and-forth in the whole state machine and it breaks down to while (c == ...) grab(); while (c != ...) skip(); so jump prediction would work as expected (given that it fits in L1 so no "touching memory" occurs) |
| 12:35 | <ment> | Philip`: but cycle counting is still ridiculous in this context |
| 12:35 | <hsivonen> | ment: if all the state transitions are compiled down to gotos, surely code locality-based caching should do the right thing when you are jumping among the doctype-related states? |
| 12:36 | hsivonen | intends to make that stuff use gotos instead of switch eventually |
| 12:37 | gsnedders|work | could see what effect using goto instead of switch in PHP 5.3 has… I doubt much |
| 12:37 | <hsivonen> | gsnedders|work: PHP has goto? |
| 12:37 | <gsnedders|work> | hsivonen: Since 5.3 |
| 12:37 | <hsivonen> | yay |
| 12:38 | hsivonen | wishes Java had goto |
| 12:38 | <gsnedders|work> | And the manual page for goto has http://xkcd.com/292/ on it |
| 12:38 | <ment> | hsivonen: the doctype parser states are jump-forward only (given that in one state you can have small loop on != or ==), so yes |
| 12:39 | <gsnedders|work> | (The original PHP tokenizer used methods, but the function call overhead made that stupidly slow0 |
| 12:39 | <ment> | hsivonen: replacing switch with goto? the compiler reorders your code anyway. have you even tried to profile it? |
| 12:39 | <gsnedders|work> | ment: You still have to go through all the options in the switch each time |
| 12:40 | <hsivonen> | ment: I haven't profiled switch vs. goto |
| 12:40 | <hsivonen> | ment: I don't see why goto would be slower than switch, though. dunno if it's going to be a win. |
| 12:40 | <ment> | gsnedders|work: have you ever heard of things like jump tables? or binary trees? |
| 12:41 | <gsnedders|work> | ment: Yes, but I would still expect goto to be quicker |
| 12:41 | <gsnedders|work> | (Though perhaps the difference is almost irrelevantly small) |
| 12:45 | <ment> | gsnedders|work: for(;;) switch (state) { case STATE_A: ...; state = STATE_B; break; case STATE_B: ...; } would have the break compiled as goto anyway |
| 12:46 | <hsivonen> | ment: goto to STATE_B? which compiler? |
| 12:47 | <ment> | hsivonen: gcc |
| 12:47 | <hsivonen> | ment: cool. |
| 12:47 | gsnedders|work | doesn't remember seeing that before |
| 12:47 | <Philip`> | I guess you have to be careful not to do anything (like read the next character) outside the switch for that to work |
| 12:47 | hsivonen | should disassemble gcc output some time |
| 12:47 | <Philip`> | but that doesn't seem too hard |
| 12:50 | <ment> | you could define a macro and use it instead of break: #define NEXT(st) state = st; c = getc(); break |
| 12:53 | <ment> | or better #define NEXT(st) c=getc(); state = st; break |
| 12:53 | hsivonen | has trouble seeing how hiding break; inside a macro is any more righteous than goto. |
| 12:54 | <jcranmer> | it isn't, IMHO |
| 12:59 | <zcorpan_> | does audio/ogg; codecs=1 mean Ogg container with PCM audio? |
| 13:01 | <ment> | btw, what does "fragment case" exactly mean? |
| 13:01 | <hsivonen> | ment: innerHTML |
| 13:01 | <ment> | ah, this thing. thanks |
| 13:02 | <gsnedders|work> | (It also applies to some things like HTML fragments in Atom) |
| 13:03 | <hsivonen> | and createContextualFragment |
| 13:03 | <gsnedders|work> | And numerous other beautiful parts of the web arch |
| 13:04 | hsivonen | wonders if there's any sane way to report line numbers in document.written content |
| 13:05 | <hsivonen> | hmm. is it Oct 30th yet? http://www.w3.org/2005/Incubator/app-backplane/XGR-app-backplane-20091030/ |
| 13:06 | <Dashiva> | Is it LC yet? |
| 13:06 | <annevk2> | hsivonen, might be the pubdate |
| 13:06 | <annevk2> | sometimes editor's drafts are a bit ahead of time |
| 13:06 | <hsivonen> | annevk2: ah. right. it's not under TR |
| 13:07 | hsivonen | wonders what "X" in "XG" means except that X is always cool |
| 13:08 | <zcorpan_> | extensible group? |
| 13:08 | <ment> | btw is there any list of html5 elements together with their flags (like IS_SCOPING, IS_FORMATTING, etc.)? |
| 13:09 | <Dashiva> | The X means 'incubator' |
| 13:09 | <hsivonen> | ment: http://hg.mozilla.org/mozilla-central/file/57e42052ea96/parser/html/nsHtml5ElementName.cpp |
| 13:10 | <gsnedders|work> | ment: Look in the stack of opoen elmenets] |
| 13:10 | <gsnedders|work> | (just spelt correctly) |
| 13:11 | <Dashiva> | Maybe IG was taken by interest group, so they decided to pick another letter |
| 13:11 | <ment> | gsnedders|work: phrasing category seems to be defined negatively |
| 13:12 | <gsnedders|work> | ment: Well, it's impossible to define it otherwise |
| 13:13 | <gsnedders|work> | ment: As it has to cover an infinite number of elements |
| 13:13 | <ment> | gsnedders|work: what do you mean? i thought that the set of html5 elements (not in foreign content) is finite |
| 13:14 | <gsnedders|work> | ment: <acompletelyfakeelement> is a phrasing element for the parser |
| 13:14 | <gsnedders|work> | ment: There's an infinite number of HTML elements that are non-conforming |
| 13:15 | <ment> | gsnedders|work: oh, so in html5 i could create new elements on the fly? |
| 13:15 | <gsnedders|work> | ment: No, it's non-conforming, but the parser defines how to parse any byte-stream |
| 13:16 | <ment> | gsnedders|work: i see your point. but that's quite a good idea! |
| 13:16 | <gsnedders|work> | ment: Everyone making up their own elements a good idea? |
| 13:17 | <gsnedders|work> | ment: Shouldn't they go to standards organizations to specify them? |
| 13:18 | <Philip`> | gsnedders|work: Not if they put colons in the names |
| 13:19 | <ment> | gsnedders|work: if they don't have any special meaning, why not? |
| 13:19 | <gsnedders|work> | Philip`: Oh yeah, right |
| 13:19 | <gsnedders|work> | ment: What do you gain by having your own elements? |
| 13:19 | <Philip`> | ment: We might want to give them special meaning in the future |
| 13:21 | <zcorpan_> | Philip`: we can solve that with namespaces! |
| 13:21 | <zcorpan_> | or versioning |
| 13:21 | <ment> | gsnedders|work: for example, i could build a new languages on html5 |
| 13:21 | <jgraham> | ment: What's wrong with class? |
| 13:21 | <gsnedders|work> | ment: Why? |
| 13:21 | <zcorpan_> | or both |
| 13:21 | <jgraham> | zcorpan_: Versioned namespaces? |
| 13:22 | <ment> | Philip`: for example, reserve all element names with name shorter than 6 characters |
| 13:22 | <hsivonen> | it's interesting how namespace proponents ignore the situation we had with the XHTML2 WG and the HTML WG using the same namespace |
| 13:22 | <Philip`> | jgraham: Namespaced versions |
| 13:22 | <hsivonen> | and that issue had a social solution--not a technical solution |
| 13:22 | <gsnedders|work> | Philip`: of namespaces? |
| 13:22 | <TabAtkins> | ment: unknown elements have *no* semantics. They're <div>s to any user agent, which means they're horrible for accessibility. |
| 13:23 | <hsivonen> | TabAtkins: it's safer to talk about <span>s to avoid giving any wrong ideas about behavior |
| 13:23 | <ment> | jgraham: i don't know. i just plugged my own tags and attributes in the parser. |
| 13:23 | <TabAtkins> | As well, as noted, it pollutes the language with private extensions that may not be well-thought out, and may prevent us from using appropriate names in the future. |
| 13:23 | <jgraham> | Philip`: Nah, we should have syntax like using:xmlns="http://some/long/uri/version/2"; so that we can change the namespace processing of documents in the future |
| 13:23 | <ment> | TabAtkins: because ajax web site is so much more accessible for blind user |
| 13:23 | <TabAtkins> | hsivonen: True, and actually they act more like a <span> anyway. |
| 13:23 | <gsnedders|work> | ment: Unless they're standardized, they bring no advantage to anyone over data-* attributes for example |
| 13:23 | <gsnedders|work> | ment: It is entirely possible for them to be accessible |
| 13:23 | <jgraham> | ment: (I should note that there is a certain amount of sarcasm going around) |
| 13:24 | <gsnedders|work> | jgraham: Or maybe a URI as the namespace in curly brackets with an ISO 8601:2004 date following it? |
| 13:24 | <TabAtkins> | Also, we really shouldn't be going out of our way to make sites less accessible. Rather the opposite, if possible. |
| 13:25 | <DrEvil> | what about i-hate-blind-people.com ? |
| 13:25 | <jgraham> | ment: I still don't understand why you want private extensions beyond what is already possible |
| 13:25 | <jgraham> | I see why browser vendoes might but as noted on the mailing list, inventing elements doesn't work in that case (even with namespaces) |
| 13:26 | <ment> | jgraham: i just think it's nice thing to have parser and tokenizer capable of working with non-standardized tags just like any other <div> (for example for future extensions) |
| 13:26 | <ment> | jgraham: or any other purpose |
| 13:26 | <jgraham> | ment: They can. It's just that you shouldn't create extensions using element names except as part of the HTML standard |
| 13:27 | <TabAtkins> | Indeed, the parser definitely needs to be *able* to handle unknown elements, exactly for extending the language - we don't want a repeat of FF2 or IE6/7's unknown tag handling. |
| 13:27 | <gsnedders|work> | But IE6 is awesome. |
| 13:27 | <TabAtkins> | But extending the language element-wise is *not* something we want to enable for everyone, precisely because it's a *bad* extensibility mechanism. |
| 13:27 | <TabAtkins> | gsnedders|work: You're right, I should apologize. |
| 13:28 | <TabAtkins> | IE6 leaves me in awe, I just won't say of what. |
| 13:28 | <gsnedders|work> | The smell of it reminds of when I went to antique stores as a kid. |
| 13:29 | <TabAtkins> | Heh, I like your quit message. |
| 13:29 | gsnedders|work | wonders what it is |
| 13:29 | <gsnedders|work> | Oh, that |
| 13:41 | <Philip`> | Floating-point is irritating |
| 13:41 | <Philip`> | I think what's currently in the spec is still slightly impossible to implement correctly |
| 13:43 | <Philip`> | because of numbers like 1.000000000000000111022302462516 |
| 13:43 | <Philip`> | which could be indefinitely long |
| 13:48 | <Philip`> | Oh, hmm, Firefox doesn't implement what ECMAScript seemingly requires |
| 13:48 | <Philip`> | (for parsing floats in JS) |
| 13:56 | <Philip`> | http://software.hixie.ch/utilities/js/live-dom-viewer/saved/293 |
| 13:57 | ment | desperately needs modern text browser |
| 13:57 | <Philip`> | Firefox, Safari and Opera on Windows give one output |
| 13:57 | <Philip`> | IE on Windows gives a second output |
| 13:57 | <Philip`> | Opera on Linux gives a third |
| 13:59 | <Philip`> | (Firefox on Linux gives the first output) |
| 13:59 | <Philip`> | I suppose it's alright to expect the first output, then |
| 14:00 | <Philip`> | which I think is what the HTML5 text expects |
| 14:00 | <Philip`> | so that's okay |
| 14:01 | <zcorpan_> | on mac my browsers give 2.220446049250313e-16 for the first four and 0 for the last two |
| 14:06 | <Philip`> | zcorpan_: That matches the first outputs I got |
| 14:06 | <Philip`> | so it's just IE and Opera-on-Linux that differ |
| 14:09 | <Philip`> | I think only IE matches what ECMAScript requires |
| 14:09 | <Philip`> | (It returns one 2.2etc and then all 0s) |
| 14:11 | <Philip`> | Hmm, I think Opera matches what ES requires too |
| 14:11 | <Philip`> | on Linux |
| 14:11 | <Philip`> | (though it's a different output to IE) |
| 14:16 | <DrEvil> | if I used a data URI to embed a jpeg or pdf file in my page |
| 14:16 | <DrEvil> | can a user still get the normal save as/download dialogue? |
| 14:18 | <Philip`> | DrEvil: It should act the same as any normal JPEG or PDF that doesn't use funny HTTP headers |
| 14:19 | <Philip`> | (i.e. you can't use content-disposition to force a dialog box) |
| 14:21 | <zcorpan_> | but you can use data:application/octet-stream;base64,... which will probably bring up the save as dialog |
| 14:21 | <DrEvil> | oooh snazzy |
| 14:21 | <DrEvil> | so we can actually have any XHR'd data presented for download to users |
| 14:21 | <DrEvil> | that would immense. |
| 14:21 | <DrEvil> | be^ |
| 14:24 | <Philip`> | Oh, I think I was interpreting ES as being stricter than it is |
| 14:24 | <Philip`> | and actually all browsers are acceptable given its requirements |
| 14:25 | <zcorpan_> | Philip`: are there other outputs that are acceptable too? |
| 14:25 | <Philip`> | zcorpan_: ES says to truncate after 20 significant digits, and then optionally add 1 to the 20th digit |
| 14:26 | <Philip`> | and it doesn't say you have to choose the same option all the time |
| 14:26 | <Philip`> | so it's acceptable to examine an unlimited number of digits before deciding whether to move the 20th digit up or not |
| 14:26 | <Philip`> | (which is what most implementations do) |
| 14:30 | <Philip`> | (IE ignores digits after the 20th, Opera on Linux ignores digits after the 21st, if I'm not mistaken) |
| 14:33 | <zcorpan_> | has anyone written tests for the new script parsing? |
| 14:37 | <gsnedders|work> | I'd quite like them for html5lib soon, FWIW |
| 14:38 | <TabAtkins> | Seriously, why are we talking about @longdesc? The discussion was over a long time ago - of the tiny fraction of pages that use it, almost all of them use it incorrectly, and screenreaders tend to ignore the attribute because of this. |
| 14:39 | <Philip`> | TabAtkins: We're talking about it because someone submitted a change proposal, and if nobody submits a counter-proposal then it will get accepted by default |
| 14:40 | <TabAtkins> | Philip`: Ok. Then let's just make the counterproposal and stop talking about it. It's just spam. |
| 14:40 | <Philip`> | TabAtkins: Indeed |
| 14:42 | <Philip`> | Seems like the counter-proposal needs to be a summary of all the arguments against longdesc from the past many years |
| 14:42 | <Philip`> | though I suppose it should be called "rationale" rather than "arguments" |
| 14:44 | <zcorpan_> | gsnedders|work: have you started writing tests? |
| 14:44 | <gsnedders|work> | zcorpan_: No |
| 14:44 | <gsnedders|work> | zcorpan_: If nobody else has by the time I finish implementing it, I will |
| 14:44 | <zcorpan_> | can we run html5lib tests in opera these days? |
| 14:45 | <gsnedders|work> | zcorpan_: We already do |
| 14:45 | <zcorpan_> | ok so where should i get cracking? |
| 14:45 | gsnedders|work | can't remember where he has them though |
| 14:46 | <gsnedders|work> | Somewhere on gsnedders.html5.org |
| 14:46 | <TabAtkins> | Well, since I picked up the project to document why features exist, and haven't done anything about it yet, I suppose I could do so. Documenting why features *don't* exist is about as useful in some cases. |
| 14:48 | <gsnedders|work> | I guess we need to do something about contentModelFlags.test and escapeFlag.test now |
| 14:48 | <gsnedders|work> | And probably need to add some way to start the tokenizer in a specific state |
| 14:48 | <gsnedders|work> | (As otherwise the tokenizer tests could never test the script data states) |
| 14:56 | zcorpan_ | clones html5lib |
| 15:35 | <zcorpan_> | hmm, it's a bad idea to drink coffee, sneeze and have your macbook on the lap at the same time |
| 15:38 | <jgraham> | zcorpan_: Assiming your computer isn't currently giving you an electric shock, what does the test you just changed now expect? Two adjacent text nodes? |
| 15:39 | <zcorpan_> | jgraham: yes |
| 15:39 | <jgraham> | Hmm, I thought that didn't happen |
| 15:39 | <zcorpan_> | happens when there's a token in between |
| 15:40 | <zcorpan_> | it was changed in the spec at some point |
| 15:40 | jgraham | wonders if html5lib can deal with that |
| 15:42 | <Philip`> | It should be changed so it can :-) |
| 15:42 | <Philip`> | and also so it doesn't have O(n^2) behaviour in any text-concatenation cases, hopefully |
| 15:42 | <ment> | btw is there any html5lib for C? |
| 15:43 | <jgraham> | ment: Nothing that actually works |
| 15:43 | <jgraham> | I keep starting writing bits of it and then get bored quite quickly |
| 15:43 | <gsnedders|work> | jgraham: You only append a char to a text node if the text node was the last node added to the document |
| 15:43 | <Philip`> | There's hubbub, I think |
| 15:44 | <jgraham> | Oh yes, there is a C implementation |
| 15:44 | <Philip`> | http://www.netsurf-browser.org/projects/hubbub/ |
| 15:44 | <jgraham> | But not a html5lib (tm) |
| 15:44 | <jgraham> | Note: not a real trademark |
| 15:44 | <Philip`> | jgraham: What makes an HTML5 implementation a html5lib (tm)? |
| 15:44 | <Philip`> | s//parser/ |
| 15:44 | <jgraham> | Philip`: It is in the html5lib repository |
| 15:45 | <jgraham> | (as far as I'm concerned the Pyhton implementation is the only true html5lib; the ruby one was but it died) |
| 15:45 | <Philip`> | In these days of decentralised VCSs, what is "the ... repository"? |
| 15:46 | <jgraham> | Philip`: On code.google.com |
| 15:46 | <ment> | i think i'll name my library 5libhtml or lib5html |
| 15:46 | <gsnedders|work> | jgraham: So PHP is a just an illusion? |
| 15:46 | <ment> | jgraham: yes, it is boring, because the standard isn't machine readable :) |
| 15:46 | <gsnedders|work> | Probably the best way, though |
| 15:47 | <Philip`> | ment: My machine can easily the spec and display its words to me |
| 15:47 | <Philip`> | *read the spec |
| 15:47 | gsnedders|work | notes it wasn't reading the spec that was what bored jgraham |
| 15:49 | <Philip`> | If nothing changed since I last checked, it should be possible for html5lib-in-Python to avoid ever concatenating strings multiple times (which can be very slow) |
| 15:49 | <Philip`> | but as far as I'm aware it doesn't avoid that |
| 15:49 | <Philip`> | so someone should fix that :-) |
| 15:50 | <gsnedders|work> | My work on the tokenizer does that a fair bit |
| 15:50 | <gsnedders|work> | But mainly for temporaryBuffer |
| 15:50 | <gsnedders|work> | Though I guess we could have that as a list |
| 15:50 | <Philip`> | (The idea was that you'd keep a list of text nodes to be inserted, instead of inserting immediately, and when you insert any other node you first concatenate and flush that list and then carry on) |
| 15:51 | <Philip`> | (I hope that's what the spec allows) |
| 15:51 | <ment> | text nodes ought to be coalesced together? ie. "text</ignoredendtag>text" => "texttext" ? |
| 15:53 | <Philip`> | ment: http://whatwg.org/html5#insert-a-character |
| 15:54 | <Philip`> | i.e. yes |
| 15:54 | <Philip`> | in your particular example |
| 15:59 | <gsnedders|work> | zcorpan_: May I, however, point out that every TC you change delays the release of html5lib 0.12 :P |
| 16:01 | <zcorpan_> | gsnedders|work: i was just changing that tc to make sure that i could commit properly |
| 16:01 | <Philip`> | gsnedders|work: You need a release branch |
| 16:01 | <zcorpan_> | gsnedders|work: my objective is to create new tests for script parsing |
| 16:02 | <Philip`> | on which you can ignore non-obvious bugs such as spec mismatches |
| 16:07 | <jgraham> | Philip`, gsnedders|work I think we should do that at WHATWG LC and aim to make 0.12 as correct per the LC draft as possible |
| 16:07 | <jgraham> | Then we should call it 1.0 or something |
| 16:08 | <jgraham> | gsnedders|work: Can't hear a word you're saying |
| 16:08 | <zcorpan_> | is there a policy i should adhere to before pushing a new set of tests? |
| 16:08 | <jgraham> | zcorpan_: Like what? |
| 16:08 | <gsnedders|work> | jgraham: Ah, is this this "selective hearing" thing? :P |
| 16:08 | <zcorpan_> | dunno, like trying to run the tests or something |
| 16:08 | <jgraham> | gsnedders|work: No more good headphones |
| 16:08 | <gsnedders|work> | jgraham: So we should release 0.12, and then call it 1.0? |
| 16:08 | <jgraham> | + extra commas |
| 16:08 | <gsnedders|work> | jgraham: But you could hear I was speaking |
| 16:09 | <gsnedders|work> | jgraham: Your headphones blatantly aren't good enough |
| 16:09 | <jgraham> | zcorpan_: You should check that you didn't cause any regressions. But if you only changed tests you can't really |
| 16:10 | <jgraham> | I guess in an ideal world you would ping html5lib-discuss about new tests that the current code fails |
| 16:10 | <jgraham> | But that never happens |
| 16:11 | zcorpan_ | has now pushed some untested work in progress tests for script |
| 16:12 | <zcorpan_> | feel free to fix any bugs :) |
| 16:12 | <gsnedders|work> | zcorpan_: Can't I just nag you? |
| 16:13 | <gsnedders|work> | zcorpan_: What's the diff between line 109 and 121? |
| 16:13 | <gsnedders|work> | zcorpan_: and 192 and 204 |
| 16:13 | <zcorpan_> | a trailing space if i'm guessing correctly which lines you're talking about |
| 16:16 | <gsnedders|work> | Ah |
| 16:50 | <gsnedders|work> | jgraham: Spotify working for you? |
| 16:50 | <jgraham> | gsnedders|work: No |
| 16:50 | <Rik`> | gsnedders|work: it just restarted here and not working since |
| 16:51 | gsnedders|work | has it offline, but still has the rest of the track he's on in buffer |
| 16:51 | <jgraham> | This is what happens when I decide to pay for it :( |
| 16:51 | <gsnedders|work> | (which is rather nice, being a 17 min long track) |
| 16:51 | <gsnedders|work> | Oh, and now it's just come back |
| 17:03 | <Philip`> | Someone ought to stop me from filing bugs that would make HTML5 harder to read |
| 17:03 | <Philip`> | (like http://www.w3.org/Bugs/Public/show_bug.cgi?id=8087) |
| 17:37 | <ment> | in which state starts fragment parser? |
| 17:37 | <gsnedders|work> | ment: That depends |
| 17:37 | <gsnedders|work> | ment: see http://www.whatwg.org/specs/web-apps/current-work/#html-fragment-parsing-algorithm |
| 17:38 | <ment> | i see, thanks |
| 19:47 | <Hixie> | wtf is johab |
| 19:48 | <Philip`> | http://unicode.org/Public/MAPPINGS/OBSOLETE/EASTASIA/KSC/JOHAB.TXT |
| 19:48 | Philip` | shrugs |
| 19:48 | <Philip`> | All I know is that it appears to exist |
| 19:49 | <Hixie> | man, jshin has been at this a while |
| 19:51 | <Hixie> | it appears to be KS C-5601-1992 |
| 19:52 | <Hixie> | hah, Wikipedia is explicitly missing 1361 |
| 19:53 | <Hixie> | apparently KS C 5601 is now KS X 1001 |
| 19:53 | <Hixie> | oh no wait |
| 19:53 | <Hixie> | that's the character set |
| 19:53 | <Hixie> | not the encoding |
| 19:54 | <Hixie> | but the encoding is listed in KS X 1001 annex 3 |
| 19:55 | <Philip`> | http://unicode.org/mail-arch/unicode-ml/Archives-Old/UML011/0205.html - "Johab should NOT be considered as included in KS C 5601-1992" |
| 19:56 | Philip` | knows nothing about this, except that iconv supports an encoding under the name "johab", and is quite happy not knowing any more :-) |
| 19:57 | <Hixie> | i was quite happy not knowing that much! |
| 19:57 | <Hixie> | now i am less happy :-P |
| 20:03 | <Hixie> | Philip`: your no-c-end is wrong |
| 20:03 | <Hixie> | Philip`: because it doesn't allow a string ending in "-" |
| 20:04 | gsnedders | wonders whether he should bother going to uni |
| 20:04 | <Hixie> | i suppose we could add [ %x2d [ %x2d ] ] at the end of it |
| 20:04 | <Hixie> | gsnedders: i would encourage it, personally. e.g. it is orders of magnitude harder to get hired by google if you don't have at _least_ an undergrad degree |
| 20:06 | <gsnedders> | Hixie: I may end up doing something that wouldn't help me get hired by Google anyway :P |
| 20:06 | <TabAtkins> | gsnedders: I'd recommend going for the same reason. It's not as necessary in the States as I understand it to be in the UK, but still, very helpful. |
| 20:06 | <Hixie> | gsnedders: how so? |
| 20:07 | <gsnedders> | Hixie: How beneficial is English Language and Linguistics to getting hired by Google? |
| 20:07 | <Hixie> | gsnedders: significantly more than Physics |
| 20:07 | <Hixie> | gsnedders: we have a whole linguistics team |
| 20:07 | <gsnedders> | Hixie: And you were trying to convince me to do physics before :P |
| 20:07 | <Hixie> | gsnedders: but the subject matter doesn't matter half as much as the fact that you got a degree |
| 20:07 | <jcranmer> | not surprising |
| 20:08 | <jcranmer> | since, after all, Google is first and foremost a search engine |
| 20:08 | <Hixie> | i was trying to convince you to do something broadly useful, not something specific to getting hired by google :-) |
| 20:09 | <jcranmer> | hmm, I keep forgetting that we're supposed to leave our sense of logic by the door |
| 20:09 | <gsnedders> | Applying for uni would in general be easier if: a) I had decent exam results from my last year of school; b) I had a clue about what subject to apply for; c) Where to apply; d) Whether I really want to go to uni. |
| 20:10 | <jcranmer> | I suspect I could have gotten a job without uni, but it tends to be harder to do so |
| 20:10 | <jcranmer> | treat uni as 3 or 4 years of laying back |
| 20:11 | <inimino> | and getting into debt ;-) |
| 20:11 | <jcranmer> | not in my case |
| 20:12 | <jcranmer> | 3 years + parents who have fiscal sanity |
| 20:12 | <hsivonen> | so according to engadget, android 2.0 adds html5 support |
| 20:12 | <hsivonen> | does it mean canvas and h.264 <video> or something else? |
| 20:13 | <otherarun> | gsnedders, or you could write a great work of literature, and forgo all this "job" tomfoolery. |
| 20:14 | <Philip`> | Hixie: Oh, good point |
| 20:14 | <gsnedders> | Gah. decisions that have some vague effect on your future! |
| 20:14 | <otherarun> | *sigh these days there's no specificity in the term "HTML5 support." |
| 20:14 | <otherarun> | It's become a marketing catchphrase. |
| 20:15 | <Hixie> | Philip`: it gets way more complicated for the script case |
| 20:16 | <Philip`> | Hixie: The script case seems alright since it's not invalid ABNF |
| 20:16 | <Philip`> | or at least relatively alright |
| 20:16 | <erlehmann> | inimino, i live in a state without horrendous study debts. berlin ;) |
| 20:16 | <Philip`> | (though it's kind of cheating to use ABNF and then stick big lumps of prose in it) |
| 20:16 | <roc> | otherarun: this actually works to our advantage in the end, I think |
| 20:17 | <gsnedders> | In Scotland you can still expect to spend around £30k in total over the course of uni, primarily just on living |
| 20:17 | <inimino> | erlehmann: yes, a lot of it depends on where you live (and on your grades/test scores) |
| 20:17 | <Hixie> | otherarun: did you get my e-mail re thursday? |
| 20:18 | <jcranmer> | I estimate my total university cost at around $100K |
| 20:18 | <gsnedders> | That just seems completely ridiculous. |
| 20:18 | <erlehmann> | inimino, in most german states, studying is essentially free of charge |
| 20:19 | <otherarun> | roc, yeah, the W3C developer thing that brendan, fantasai and I are talking at has attracted lots of folks intrigued by HTML5. maybe you're right. |
| 20:19 | <otherarun> | Hixie, yep got it. responding now-ish.... |
| 20:19 | <jcranmer> | hey, it's an out-of-state school |
| 20:19 | <Hixie> | otherarun: k |
| 20:20 | <gsnedders> | jcranmer: Going to anywhere in the UK would cost me around 30k GBP, going to somewhere in Sweden would cost me around the same… |
| 20:20 | Philip` | had something like £12K of student loans and spent a bit less than that over three years, if he remembers correctly |
| 20:20 | <erlehmann> | gsnedders: come to germany, we have hackers here :3 |
| 20:21 | <gsnedders> | Philip`: So £9k on tutition fees, and £1k/year on living :P |
| 20:21 | <gsnedders> | erlehmann: That means learning German :P |
| 20:22 | <erlehmann> | gsnedders it isnt so hart. listen and repeat: JAWOHL HAUPTKOMMISSAR SCHWEINEHUND |
| 20:22 | <erlehmann> | ZU BEFEHL FEUER SAUERKRAUT |
| 20:22 | <Philip`> | gsnedders: Tuition fees were only added in my last year |
| 20:22 | <gsnedders> | Philip`: Silly English. |
| 20:23 | <gsnedders> | erlehmann: Hey, I'm going to Berlin for Christmas. I'm learning, dammit! |
| 20:23 | <Philip`> | The only German I learnt and remembered is from Wolfenstein 3D |
| 20:23 | <erlehmann> | gsnedders, you are attending 26c3 ? are you with the kaminsky crowd ? |
| 20:23 | <Dashiva> | What about Indiana Jones? |
| 20:24 | <gsnedders> | erlehmann: wah? |
| 20:25 | <gsnedders> | erlehmann: (judging by the fact I've never heard of either, no) |
| 20:25 | <erlehmann> | gsnedders: 26C3 is the 26th chaos communication congress, 3 to 4 days between christmas and sylvester. several thousand attendees, also international (hey, i unexpectedly ran into bre pettis there) |
| 20:25 | <erlehmann> | go there, it is awesome |
| 20:25 | <gsnedders> | erlehmann: Looking it up, I could only go to the first day |
| 20:26 | <gsnedders> | (I leave for my parent's on the 28th) |
| 20:27 | <erlehmann> | then go there for a day |
| 20:27 | <erlehmann> | its the most awesome event to attend in that timeframe |
| 20:27 | <gsnedders> | erlehmann: I'm unlikely to have the time |
| 20:27 | <gsnedders> | Also, I should probably finally get around to having supper |
| 20:28 | <erlehmann> | sad thing. |
| 20:29 | gsnedders | turns oven on, and realizes he has to wait |
| 20:30 | gsnedders | was blatantly spoiled growing up with his parents with an Aga and not having to wait |
| 20:30 | <Hixie> | mmm agas |
| 20:30 | <Hixie> | i wish they had those here |
| 20:31 | <erlehmann> | what is an „aga“? |
| 20:31 | <gsnedders> | erlehmann: http://en.wikipedia.org/wiki/AGA_cooker |
| 20:31 | <Philip`> | Is it like a microwave? |
| 20:31 | <Philip`> | You don't have to wait for those |
| 20:33 | <Hixie> | you'll wait forever if you wait for them to cook the food properly |
| 20:33 | <erlehmann> | Philip`, it seems to be an expensive, energy-hungry cooking artifact from 80 years ago. |
| 20:34 | gsnedders | guesses the one at his parent's house is around 30 years old |
| 20:35 | <Hixie> | it's usually used as a replacement or supplement for the house hot water or radiator heating, so it turns out not to be more wasteful than most systems |
| 20:35 | <Hixie> | though it's certainly not what one would call "Efficient" |
| 20:36 | <gsnedders> | When you're dealing with an Edwardian house where heat seeps out through the walls and windows like there's no tomorrow, keeping one room at a reasonable temperature is a challenge in and of itself |
| 20:37 | <gsnedders> | (And keeping a room at a sane temp. basically means having a gas fire on continuously 24/7 over the winter) |
| 20:38 | <gsnedders> | (As even if you turn it off overnight, you then spend the first 10 hours of the day getting the room up to any temp. where you can bear to be in it) |
| 20:38 | <Hixie> | or, one could use something known as "insulation" |
| 20:38 | <Hixie> | but don't feel bad, californians don't know what that is either |
| 20:38 | <gsnedders> | Hixie: You can't really do much about the outside walls of the house |
| 20:38 | <gsnedders> | Hixie: (at least without significant structural changes, which with a listed building…) |
| 20:38 | <Hixie> | this isn't my area of expertise, but i'd be surprised if there was really nothing one could do |
| 20:39 | <TabAtkins> | You just tear out the drywall and stuff insulation in there, then put it back. (Though your houses may not be built to work like that...) |
| 20:39 | <gsnedders> | TabAtkins: Edwardian houses don't have Drywall in general :) |
| 20:39 | <Hixie> | hey so good news everyone! |
| 20:39 | <Hixie> | i'm at 0,1,0 again |
| 20:39 | <TabAtkins> | Woo! |
| 20:40 | <TabAtkins> | gsnedders: Then your houses are built wrong. ^_^ |
| 20:40 | <gsnedders> | TabAtkins: No, not really. You have brick walls with plaster directly applied to the, |
| 20:41 | <TabAtkins> | so the walls themselves are brick? |
| 20:41 | <gsnedders> | TabAtkins: Yes |
| 20:41 | <gsnedders> | A hundred years ago, if you wanted a wall in house, you used bricks. |
| 20:42 | <TabAtkins> | Interesting. I've only ever seen brick used as a facade. Still, if you don't mind losing a few inches off the room dimensions, you could improvise drywall into the room, allowing you to sandwich insulation between the wall layers. |
| 20:43 | <gsnedders> | TabAtkins: See: http://en.wikipedia.org/wiki/Listed_building |
| 20:43 | <Philip`> | Oops, I failed to notice that cp1361 == johab |
| 20:44 | <Philip`> | or forgot from when I last looked at it |
| 20:44 | <TabAtkins> | Ah, that sort of mod probably would count as being against code, then. |
| 20:44 | <mpilgrim> | gsnedders, jgraham: which html5 outliner should i point people to? |
| 20:44 | <gsnedders> | mpilgrim: Mine, his is out of date |
| 20:45 | <Philip`> | Why are approximately none of the rationale comments getting through to public-html-bugzilla? |
| 20:45 | <gsnedders> | TabAtkins: Esp. when it is listed because it has almost all original fittings |
| 20:45 | <mpilgrim> | ok, thanks |
| 20:45 | <gsnedders> | TabAtkins: And when there's a lot of fine-wood around the walls in places… |
| 20:51 | <gsnedders> | jgraham: Where can you buy tissues here? I've not seen anywhere selling them anywhere… |
| 20:55 | <ment> | what for are the attributes in closing tags? </b xxx=zz> ? |
| 20:55 | <Hixie> | nothing |
| 20:55 | <Philip`> | Hmph, it takes more effort to close bugs than to file them |
| 20:59 | <Hixie> | hah |
| 20:59 | <Philip`> | Also it's hard to find which bugs are mine |
| 20:59 | <Hixie> | even with the IP address? |
| 20:59 | <Philip`> | because I have dozens of different IP addresses |
| 21:00 | <Philip`> | I use at least three different internet connections each day, none of which have static IPs |
| 21:06 | <jgraham> | gsnedders: Apoteket for sure. also I think the supermarket |
| 21:07 | <Philip`> | Hmm, I think I got them all now |
| 21:07 | <Philip`> | and only had to reopen one |
| 21:07 | <Dashiva> | gsnedders: Normal food stores have them |
| 21:09 | <Hixie> | Philip`: good times |
| 21:10 | <ment> | script-data-double-escaped-dash-dash-state ... what the fuck |
| 21:11 | <gsnedders> | jgraham: hmm, I didn't see any when I looked (albeit very briefly before) |
| 21:12 | <Dashiva> | They're well hidden |
| 21:12 | <Dashiva> | Sometimes there's just a single type and a single brand |
| 21:14 | <jgraham> | gsnedders: BTW you are really making oo big a deal out of this whole university thing. You should just pick something |
| 21:14 | <jgraham> | It doesn't matter what so much |
| 21:14 | <jgraham> | or where |
| 21:14 | <jgraham> | (wihin reasonable limits in both cases) |
| 21:15 | <jgraham> | (i.e. it has to be something you expect to enjoy somewhere you can bear to live and aren't embarassedto graduate from) |
| 21:15 | <jgraham> | You can't make a prefect objective decision because all the necessary facts aren't there |
| 21:20 | <JonathanNeal> | is the source code for outliner available? |
| 21:22 | <Hixie> | Philip`: yt? |
| 21:23 | <Hixie> | Philip`: is there a version of your multipage script that takes an argument for which index file to process and an argument for where to post the results? |
| 21:23 | <Hixie> | Philip`: or alternatively, can you mail me the multipage version of http://www.whatwg.org/specs/web-apps/2009-10-27/ ? |
| 21:23 | <Philip`> | Hixie: There's not an online version that does that, only the offline Python script |
| 21:24 | Philip` | could do that in an hour or so |
| 21:24 | <Hixie> | cool, thanks |
| 21:25 | Philip` | should make sure the sections are sensibly split, too |
| 21:26 | <ment> | i know i will be probably banned for this statement, but the author (or team of authors) who created html5 tokenizer (the latest version) deserves to get an inoperable tumor at the base of his spine |
| 21:26 | <Hixie> | ment: :-( |
| 21:28 | <ment> | now i have to either retype 300KB worth of DFA state machine or draw several pages of diagrams to figure out all the hidden exceptions of latest tokenizer |
| 21:29 | <Hixie> | ment: anything in particular that can make it sier for you? i think the changes actually made the parser simpler, not more complex (though it has more states) |
| 21:29 | <Hixie> | (the changes were necessary to be compatible with legacy content) |
| 21:31 | <jgraham> | JonathanNeal: There is an ouliner as part of anolis which I think is the basis for gsnedders online one |
| 21:31 | <jgraham> | look at hg.gsnedders.org or so |
| 21:31 | <gsnedders> | s/org/com/ |
| 21:31 | <ment> | Hixie: well if i ever finish the diagrams, i will show you simpler way to describe automatons like this |
| 21:32 | <Hixie> | ment: that'd be great |
| 21:32 | jgraham | finds the current spec reaonably simple to understand |
| 21:32 | <jgraham> | *reasonably |
| 21:33 | <JonathanNeal> | This is great, thanks jgraham and gsnedders |
| 21:33 | <jgraham> | (certianly not so complex that I am wishing painful death on Hixie) |
| 21:34 | <gsnedders> | So you just do that for different reasons? |
| 21:35 | <Hixie> | ok, anyone have any reason i shouldn't hit "submit" on this blog post? |
| 21:35 | <gsnedders> | Because the button is labelled, "Publish" |
| 21:35 | <Hixie> | btw i only created a static copy of the html5 spec, not the complete, vocabs, and workers specs |
| 21:35 | <Hixie> | because that takes way the hell too long per spec |
| 21:35 | <Hixie> | and it's not clear to me what the benefit is |
| 21:35 | <Hixie> | gsnedders: good reason, any others? :-) |
| 21:36 | <jgraham> | Hixie: No |
| 21:36 | <Hixie> | ok i have scheduled it for 21:55, which i guess is in about 10 minutes |
| 21:36 | <Philip`> | Because there's no multipage version yet and so you'll crash people's browsers when they look at it? :-) |
| 21:38 | <Hixie> | the blog post just links to the current-work/ spec: http://blog.whatwg.org/html5-at-last-call |
| 21:38 | <Hixie> | actually it links to current-work/multipage/ |
| 21:41 | Philip` | tries multipaging it now |
| 21:43 | <Hixie> | 4 minutes to go! |
| 21:43 | <mpilgrim> | future-posts are evil |
| 21:44 | <mpilgrim> | i tweeted it and then realized that the link is 404 for anyone not logged into blog.whatwg.org |
| 21:44 | <gsnedders> | The past and the future are only illusions, there is only now. |
| 21:44 | <erlehmann> | quick, mpilgrim, submit it to slashdot, so subscribers can see posts IN THE FUTURE |
| 21:44 | <Hixie> | mpilgrim: hah |
| 21:44 | Hixie | advances the publication time! |
| 21:44 | <mpilgrim> | no no, i don't want to be responsible for html5 going to last call before it's ready! |
| 21:45 | <Hixie> | we're live baby! |
| 21:45 | <erlehmann> | Hixie, you need an additional status code, akin to 410 gone. 4xx - Not There Yet for stuff reserved for future use ;) |
| 21:46 | <Hixie> | erlehmann: heh |
| 21:46 | <Hixie> | erlehmann: i'm sure that exists in some HTTP extension somewhere |
| 21:47 | <Dashiva> | erlehmann: Just set up a chain of "see other" redirects that goes on until the resource comes into existence |
| 21:48 | <mpilgrim> | damn, someone replied to my tweet saying the link was 404 |
| 21:48 | <erlehmann> | Dashiva, browsers will stop if there are dozens of redirect |
| 21:48 | <erlehmann> | s |
| 21:48 | <mpilgrim> | damn the real-time web! |
| 21:49 | <mpilgrim> | seriously though, this is a historic moment |
| 21:49 | <jgraham> | Hixie: Might have been good o mention that the HTMLWG don't consider the spec in LC yet |
| 21:49 | <mpilgrim> | i'm just glad i was here to help fuck it up |
| 21:49 | <Dashiva> | erlehmann: Okay, make it only one redirect, but it hangs (or writes junk headers) until the location exists |
| 21:49 | <gsnedders> | Hmm, free shipping on deviantart today |
| 21:50 | <Philip`> | Hixie: http://philip.html5.org/misc/multipage-web-apps-2009-10-27.tar.gz might be it |
| 21:50 | <gsnedders> | jgraham: Also, what's the whole deal with recieving large packages in the post here? |
| 21:51 | <Hixie> | jgraham: i figured if i qualified it as "at the WHATWG" each time, that'd be clear enough. I don't want to sound like I'm saying "neener neener" or something like that. |
| 21:51 | <Hixie> | our relations are strained enough as it is |
| 21:51 | <Philip`> | Dashiva: Why redirect? Just let the server hold onto the HTTP session forever, until the response is ready |
| 21:52 | <Philip`> | Dashiva: (sending an "X-Ping: .........." header with a new byte every few seconds to stop it timing out) |
| 21:52 | <Dashiva> | I imagined a dedicated server handling these long connections |
| 21:52 | <TabAtkins> | gsnedders: Man, wish I had some money to order a few prints, then. |
| 21:52 | <Dashiva> | So the real content doesn't have to worry |
| 21:52 | <jgraham> | gsnedders: They end up at the postofice usually |
| 21:52 | <jgraham> | Well at the place on Nygatan |
| 21:53 | <gsnedders> | jgraham: The one parallel to Vasa vagen? |
| 21:53 | <jgraham> | Hixie: I think the current text could be construed as misleading |
| 21:53 | <gsnedders> | TabAtkins: I wish I could find what I thought I had saved as a favourite before :( |
| 21:53 | <jgraham> | gsnedders: Er not sure. Maybe I am wrong abou the name of the stree |
| 21:54 | <TabAtkins> | gsnedders: Yeah, that's often the problem. The site's too big! o_O |
| 21:54 | <gsnedders> | jgraham: Well, where? |
| 21:54 | <jgraham> | gsnedders: Near Hemkop |
| 21:54 | <jgraham> | (the one near St Lars) |
| 21:55 | <gsnedders> | Where there? |
| 21:55 | <erlehmann> | i hope moot implements some audio / video (don't they have a flash board?) |
| 21:55 | <jgraham> | gsnedders: In the spel + tobac (sp?) |
| 21:55 | <jgraham> | you get a note through the door |
| 21:56 | <jgraham> | (I guess it might end up in a different place for you) |
| 21:56 | <erlehmann> | i predict YTMND will be the most obnoxious HTML5 page in the next decade |
| 21:56 | <gsnedders> | From where I know where places are, I don't think we live far enough apart for there to really be any difference |
| 21:56 | <Hixie> | jgraham: ok, updated |
| 21:56 | <Hixie> | jgraham: (added a note at the bottom) |
| 21:57 | <mpilgrim> | just put a note that says "it's all over but the screaming" |
| 21:57 | <gsnedders> | TabAtkins: And now I keep finding things that aren't available as prints. gah. |
| 21:57 | <othermaciej> | Hixie: I have a question about bug hygiene |
| 21:57 | <Hixie> | othermaciej: yes sir |
| 21:57 | <othermaciej> | Hixie: you resolved a couple of my bugs in a way that completely addresses the original problem, but introduces what I think is a new, totally different problem |
| 21:58 | <othermaciej> | Hixie: should I reopen the bugs or file new ones? |
| 21:58 | <TabAtkins> | gsnedders: I just don't understand creators that don't make everything print-availabl. It's silly. |
| 21:58 | <Hixie> | othermaciej: either is fine by me, i'd recommend asking the chairs for their advice but that's you :-P |
| 21:58 | gsnedders | would really like http://pyromaniac.deviantart.com/art/Saving-all-my-words-24031392 as a print |
| 21:58 | <othermaciej> | Hixie: yeah I wanted to see if you have a preference |
| 21:58 | gsnedders | knows he'll be mocked by jgraham for that |
| 21:58 | <Hixie> | othermaciej: i think generally new bugs is better, but in practice it makes little difference |
| 21:59 | <othermaciej> | to be specific the problem is that HTMLCollection subclasses now violate the Liskov Substitution Principle |
| 21:59 | <othermaciej> | I'll file a new bug |
| 22:00 | <Dashiva> | That sounds serious |
| 22:00 | <mpilgrim> | ok, i looked that up on wikipedia and now i'm even more confused |
| 22:00 | <Philip`> | gsnedders: Surely it'd be better to get something less depressing |
| 22:00 | <Hixie> | othermaciej: i don't see how changing the type would change that |
| 22:00 | <mpilgrim> | it's probably my fault for dropping out of CS and pursuing a philosophy degree |
| 22:01 | <mpilgrim> | but could someone explain the liskov substitution principle without using any greek letters? |
| 22:01 | <Hixie> | othermaciej: regardless of what the interface says, the parent interface is always going to return only Element objects, whereas the subclasses can return more |
| 22:01 | <othermaciej> | Hixie: LSP is about the contract, not the behavior |
| 22:01 | gsnedders | did the same, it's just he doesn't have a degree at all |
| 22:01 | <othermaciej> | Hixie: you can't promise in the base class to return only Element if subclasses might return more general types |
| 22:01 | <Hixie> | othermaciej: that seems like a distinction without a difference, as people say |
| 22:02 | <Hixie> | othermaciej: we do promise (in prose) to only return Element objects |
| 22:02 | <Hixie> | regardless of what the interface says |
| 22:02 | <Philip`> | mpilgrim: I think the idea is that whenever you write code based on the requirements/guarantees of class(/interface) A, then you can safely pass it an object of any subclass of A and it won't violate the contract |
| 22:02 | <ment> | html5 specs is written by philosophers? :)) |
| 22:02 | <Hixie> | othermaciej: and regardless of what the interface says, you can always screw around with the prototype and have any object return anything for anything |
| 22:02 | <othermaciej> | Hixie: the way it's a practical problem is that you can't generate interfaces for any statically typed language from the WebIDL as currently written (including Java and Objective-C) |
| 22:02 | <gsnedders> | Philip`: That really isn't very dark |
| 22:02 | <Hixie> | ah, that's a more interesting problem |
| 22:03 | <Philip`> | (e.g. a subclass might override a method to return a subtype of the original type, but can't return a supertype) |
| 22:03 | <gsnedders> | Philip`: Maybe my perception of such things is a bit off, though |
| 22:03 | <Hixie> | othermaciej: file that as the bug :-) |
| 22:03 | <othermaciej> | Hixie: doing so |
| 22:03 | <mpilgrim> | Philip`: thanks, that makes sense |
| 22:04 | <Philip`> | gsnedders: It's all blacks and greys and crying. You should get something like http://veganya.deviantart.com/art/Pony-41998085 which has nice colours and happy animals |
| 22:05 | <gsnedders> | Philip`: Not available as a print, sorry |
| 22:05 | <mpilgrim> | for the benefit of people reading this discussion later, i recommend http://c2.com/cgi/wiki?LiskovSubstitutionPrinciple instead of the wikipedia article |
| 22:05 | <jgraham> | gsnedders: Also the artist seems to see himself as a short man with an enormous red penis hanging from his forhead |
| 22:06 | <othermaciej> | Hixie: http://www.w3.org/Bugs/Public/show_bug.cgi?id=8092 |
| 22:06 | <Hixie> | othermaciej: thanks |
| 22:07 | <ment> | speaking of deviantart, does anybody know why it has <keeper for="Smoothie.thumbs"></keeper><smoothie q="special:dd" qx="1"...> tags in body or what they are good for? |
| 22:07 | Philip` | notes that he doesn't remember hearing of the LSP in his CS degree at all |
| 22:08 | <othermaciej> | Philip`: I had Barbara Liskov for a professor |
| 22:08 | <othermaciej> | (though she didn't bring up the principle by that name in class) |
| 22:08 | <Philip`> | I didn't :-( |
| 22:09 | <gsnedders> | Why is it that the majority of "People" prints on deviantART are either half-naked or completely-naked? |
| 22:09 | jgraham | thinks that constrining subtypes to always have the same behaviour as supertypes sounds rather limiting |
| 22:09 | <othermaciej> | the rule isn't that they have to have "the same" behavior |
| 22:10 | <othermaciej> | just that if a caller follows the contract, a subtype will respect the contract |
| 22:10 | <othermaciej> | HTMLVideoElement doesn't have the same behavior as HTMLElement, but all HTMLElement methods do what is guaranteed by contract |
| 22:10 | <jgraham> | Hmm, the c2.com wiki says "What is wanted here is something like the following substitution property: If for each object o1 of type S there is an object o2 of type T such that for all programs P defined in terms of T, the behavior of P is unchanged when o1 is substituted for o2 then S is a subtype of T." |
| 22:10 | <Dashiva> | Apparently you have to be logged in to buy the naked prints :P |
| 22:11 | <gsnedders> | Thankfully, I'm logged in :P |
| 22:13 | <Philip`> | jgraham: I don't believe that's the same as the definition I usually associated with the term |
| 22:13 | <jgraham> | othermaciej: But you could check if an element had a play method, for example and that would be a program that didn't run unchanged when a HTMLVideoElement was passed in instead of a HTMLElement |
| 22:13 | <othermaciej> | the idea is if you have a only reference to the base type, any code you write that works with that interface and follows its side of the contract should remain correct, if you actually pass it an object that's an instance of a subtype |
| 22:13 | <Philip`> | *usually see associated |
| 22:13 | <othermaciej> | that's true - but lacking a "play" method is not part of HTMLElement's contract |
| 22:14 | <jgraham> | So how do you tell which parts are contract and which parts are not? |
| 22:14 | <Philip`> | The static type information is part of the contract |
| 22:14 | <Philip`> | i.e. the types in the IDL |
| 22:14 | <Dashiva> | The interfaces only promise existence, not non-existence |
| 22:15 | <Philip`> | and then I guess anything else is typically in comments |
| 22:15 | <Philip`> | (or, rather, typically not stated at all, but if was anywhere it'd probably be in comments) |
| 22:15 | <jgraham> | It isn't obvious to me why existence is more important than non existance |
| 22:16 | <jgraham> | (or why types are more important than other things) |
| 22:16 | gsnedders | wishes he could find what he was looking for |
| 22:17 | <gsnedders> | "But I still haven't found what I'm looking for" |
| 22:17 | <Dashiva> | Try looking in the last place |
| 22:17 | <othermaciej> | I don't think there's a question of important or not - just what is or isn't promised |
| 22:17 | <othermaciej> | Web IDL is a contract - it guarantees that if you have an instance of a certain interface, the methods and properties have the stated type signatures |
| 22:17 | <othermaciej> | it doesn't promise that an object lacks other interfaces |
| 22:18 | <othermaciej> | it does promise that an object also implements the inherited interfaces |
| 22:18 | <Philip`> | Existence is important because typically you'll write code that acts on objects, and you know the interface of the objects, and you want to be sure that if you call some method with arguments that follow certain requirements (like types) it'll return something that follows certain requirements (like types), else you've got no hope of writing working code |
| 22:18 | <jgraham> | gsnedders: I wouldn't worry. U2 didn't manage that and they have still been quite successful |
| 22:18 | <jgraham> | Oh you said that |
| 22:18 | <jgraham> | don't mind me |
| 22:18 | <Philip`> | and you'll rarely write code that assumes you'll get an exception if you call a certain method that wasn't defined |
| 22:19 | <Philip`> | (especially in statically-typed languages where you'd get a compile error) |
| 22:19 | <Dashiva> | Unless it's a method throwing NotImplementedException |
| 22:19 | <jgraham> | It seems like a reasonable strategy in non-static languages though |
| 22:20 | <Philip`> | Non-static languages are weird and crazy and you've got no hope of writing working code |
| 22:21 | <Philip`> | At least with Web IDL you can work under the impresion that you've got a proper class-based language |
| 22:22 | <jgraham> | (also LSP seems like it should want to say something about duck typing since that is about interfaces if not subtypes. But in practice it is quite common for similar types to not fully replace each other, and the world doesn't collapse) |
| 22:22 | <othermaciej> | I really shouldn't have cited LSP for my issue anyway |
| 22:23 | <othermaciej> | since in this case the subtyping fails on a syntactic / type signature level, not just a semantic level |
| 22:23 | <Philip`> | Hixie: Did you see the multipage thing I mentioned here earlier? |
| 22:23 | jgraham | is not meaning to suggest that LSP is not useful or anything, just to try and understand it better |
| 22:23 | <othermaciej> | Liskov's original statement was something like "Let q(x) be a property provable about objects x of type T. Then q(y) should be true for objects y of type S where S is a subtype of T." |
| 22:23 | Philip` | is wondering since the URL still redirects to current-work/multipage |
| 22:24 | <othermaciej> | so then it depends on what "provable property" means |
| 22:24 | <othermaciej> | which depends on how T is defined |
| 22:24 | <othermaciej> | if T is defined in terms of some sort of contract, then everything that can be inferred from that contract is a "provable property" |
| 22:24 | <Hixie> | Philip`: i had not, but got it now, thanks. downloading... |
| 22:25 | <othermaciej> | so going back to the HTMLElement interface, it's a provable property that any HTML element has a click() method with no arguments and void return type |
| 22:26 | <othermaciej> | it's not a provable property that an HTMLElement *lacks* a play() method |
| 22:27 | <othermaciej> | but anyway a contravariant return type is a violation of more basic notions of subtyping than LSP |
| 22:55 | <GPHemsley> | Hixie: I don't think you ever replied to the <cite> proposal. |
| 22:56 | <roc> | it's Liskov-WING substitution, please |
| 22:57 | <roc> | poor Jeannette |
| 22:57 | <doublec> | she has to suffer like the 'et al' in the design patterns authors :) |
| 22:58 | <Philip`> | Al coauthored an awful lot of books |
| 22:58 | <Philip`> | roc: That makes it sound less like a science-fiction name |
| 23:01 | <roc> | actually, since Ms Wing is a preternaturally youthful martial-arts badass, she adds to the sci-fi cachet |
| 23:04 | <doublec> | it's a wonder it wasn't named the 'wing substitution principle' then |
| 23:05 | <roc> | come into the office, I'll tell you some stories |
| 23:07 | <doublec> | hehe, hopefully soon |
| 23:07 | <doublec> | my landlord is giving me free furniture |
| 23:13 | <ment> | ok, i have a question: parser is in script data state, on the input is '<!-- </script> --> </script>' |
| 23:13 | <ment> | s/parser/tokenizer/ |
| 23:14 | <ment> | after which </script> will the tokenizer switch to non-script state? |
| 23:15 | <ment> | (for example, data state) |
| 23:15 | <gsnedders> | ment: the second |
| 23:15 | <ment> | nope, the first (if i haven't missed something) |
| 23:17 | <gsnedders> | hmm, that appears to be right, and that surely has to be a bug |
| 23:18 | <ment> | i can draw you a diagram how it looks like now |
| 23:22 | <gsnedders> | Oh, no, it is right, I think |
| 23:22 | <ment> | http://ibawizard.net/~thement/sds.png |
| 23:25 | <ment> | what does it behave like that? it makes no sense |
| 23:27 | <gsnedders> | ment: Legacy content. |
| 23:27 | <gsnedders> | (and not wanting to reparse) |
| 23:39 | <aboodman> | let's wait one round |
| 23:44 | <Hixie> | GPHemsley: which one? there were hundreds over the years :-) |
| 23:49 | <Lachy> | Hixie, are you aware that the new complete copy of the spec here hangs in Minefield? http://www.whatwg.org/specs/web-apps/current-work/complete.html |
| 23:50 | <Lachy> | which browser can actually handle loading it? |
| 23:50 | <Hixie> | works fine in safari nightlies and chrome nightlies last i tried |
| 23:51 | Hixie | tries firefox |
| 23:51 | <Hixie> | works fine for me on a firefox nightly with html5.enable |
| 23:52 | <Hixie> | and in opera 10.something |
| 23:52 | <Hixie> | oh, wait, firefox is hanging now |
| 23:52 | <Hixie> | is one of the scripts doing bad things in firefox? |
| 23:52 | <othermaciej> | seems to load ok in Safari |
| 23:53 | <Lachy> | yeah, it's working in safari and opera 10 |
| 23:53 | <Hixie> | bringing up the error console in safari is a bad idea because it tries to show the 5MB DOM |
| 23:53 | <othermaciej> | (but I'm also on a really fast network so likely not as much incremental layout was needed) |
| 23:53 | <othermaciej> | man we so need a datagrid for the Web Inspector |
| 23:53 | <Hixie> | scrolling in opera is painful |
| 23:54 | <JoePeck> | othermaciej: mention it in #webkit-inspector |
| 23:55 | <Lachy> | scrolling in opera isn't painful for me |
| 23:55 | <Lachy> | but I have Smart Scroll installed on my system, which is probably helping |
| 23:58 | <ment> | may i ask you again what's the motivation behind having attributes in closing tags? |
| 23:58 | <ment> | like </a href="xxx"/> |
