| 03:50 | <Hixie> | heh, this e-mail talking about discovery was itself subject to discovery, that's funny: http://edge-op.org/iowa/www.iowaconsumercase.org/011607/0000/PX00622.pdf |
| 04:25 | <jwalden> | hah |
| 06:50 | <Hixie> | 67% complete in categorising the table feedback, yay |
| 06:56 | <Hixie> | 85% |
| 07:12 | <Hixie> | woot, sorted all teh feedback for tables |
| 07:12 | <Hixie> | now i guess i have to read it! |
| 07:13 | <Hixie> | well this is weird |
| 07:15 | <Hixie> | for all the talk about headers/id, there's very little work on it other than what ben and james have done |
| 07:18 | <Hixie> | so i guess what i'm going to have to do is implement this table cell/header association algorithm |
| 07:18 | <Hixie> | skipping step 1 |
| 07:18 | <Hixie> | and see if it gives better or worse results than just obeying header/id |
| 07:19 | <Hixie> | to determine if header/id is used correctly or not |
| 07:19 | <Hixie> | if header/id is used correctly and the algorithm gets the same result, then we might as well use the algorithm |
| 07:20 | <Hixie> | if header/id is used correctly but the algorithm gets the wrong results, then we probably need header/id |
| 07:20 | <Hixie> | if header/id is used incorrectly the vast majority of the time, then it doesn't much matter, we should probably drop it anyway |
| 11:59 | <hsivonen> | aargh. lynx from MacPorts and lynx from fink break lines differently |
| 11:59 | hsivonen | is diffing the lynx text dump of the spec |
| 12:34 | <annevk> | hsivonen, web-apps-tracker is no good? |
| 12:35 | <annevk> | http://blog.kilobox.mobi/2008/03/02/whatwg-are-you-creating-html/ |
| 12:41 | <zcorpan_> | annevk: i guess it's harder to read with all the markup |
| 12:48 | <hsivonen> | annevk: not enough context |
| 12:49 | hsivonen | diffs with -U 200 |
| 12:50 | <hsivonen> | annevk: also, there have been too many interdependent changes in the tokenizer, so I find I can't do spec rev by spec rev, bug by bug patching |
| 12:50 | <hsivonen> | instead, I'm sweeping over the diff of the entire chapter in one go |
| 12:50 | <annevk> | I suppose we could make context an option |
| 12:53 | <hsivonen> | the main problem is that the tracker doesn't say what the heading before a given change was |
| 12:53 | <hsivonen> | so it isn't obvious which tokenizer state or insertion mode is affected |
| 12:55 | <annevk> | philip asked about that too |
| 12:55 | <annevk> | that would be a bit tricky though |
| 12:55 | <annevk> | i suppose you could make a diff with a lot of context |
| 12:55 | <annevk> | then traverse back from the changes until you hit <hx>, and drop everything before that or so |
| 12:56 | <hsivonen> | I think a large context alone is good enough |
| 13:13 | <annevk> | hsivonen, ok, there's now a context feature |
| 13:14 | <annevk> | http://html5.org/tools/web-apps-tracker?from=1356&to=1357&context=70 |
| 13:21 | <hsivonen> | annevk: cool. thank you |
| 13:46 | <Dashiva> | annevk: Is that blog viewable without a mobile device? |
| 13:48 | <Philip`> | Dashiva: It seems to be perfectly viewable on my non-mobile computer |
| 13:49 | <Dashiva> | I guess it's just my connection that sucks then |
| 13:49 | <Dashiva> | I need a RART |
| 13:49 | <Philip`> | A regional asset recovery team? |
| 13:50 | <Dashiva> | A router attitude readjustment tool |
| 13:52 | hsivonen | notes that requiring spaces between a quoted attribute value and the next attribute name will make valid HTML4 invalid. |
| 13:54 | <zcorpan_> | that's true |
| 13:54 | Philip` | would estimate that to be about zero HTML4 documents in practice |
| 13:56 | <zcorpan_> | Philip`: how about any HTML documents (not just counting valid HTML4)? |
| 13:56 | <Philip`> | I still estimate about zero |
| 13:56 | <Philip`> | based on no data, except I don't remember ever seeing people intentionally omit spaces between attributes |
| 13:57 | <Philip`> | (so my view doesn't really count) |
| 13:57 | <zcorpan_> | i don't either, but i have seen it in some text/html-XHTML documents (so it was not intentional) |
| 13:57 | <hsivonen> | Philip`: actually, when the W3C Validator fixed this for XHTML, there were (angry) people who found their previously 'valid' pages were in fact invalid |
| 13:58 | <zcorpan_> | would be interesting to see if pages that lack spaces between attributes do so because of mismatching quotes or not |
| 13:59 | <zcorpan_> | (since that's what the new requirement is supposed to help with) |
| 13:59 | <Philip`> | Oh, maybe it's more common that I expected |
| 14:00 | <Philip`> | I see /<[a-z]+ [a-z]+="([^"]+)"[a-z]/ on about ... |
| 14:00 | <zcorpan_> | also, perhaps the other new requirements are sufficient for catching mismatched quotes |
| 14:00 | <hsivonen> | Hixie: any particular reason why < didn't become invalid in unquoted attribute values along =, ' and ? |
| 14:00 | Philip` | waits for it to continue finding matches |
| 14:01 | <Philip`> | ... lots of pages |
| 14:02 | <Philip`> | About 5K (out of 130K), plus lots of margin of error |
| 14:03 | <Philip`> | though it looks like a lot of them are from the same site |
| 14:03 | <Philip`> | (which says <d |
| 14:03 | <Philip`> | iv id="gb_city/cottonwood-az_5"class="act_gbar" style="left: 25px; height: 18px" ></div>) |
| 14:03 | <Philip`> | (...without the line wrapping) |
| 14:05 | <zcorpan_> | i guess the whitespace between attributes requirement has a high noise to signal ratio |
| 14:05 | <Philip`> | <font face="arial color="silver">X - Y - </font> |
| 14:06 | <zcorpan_> | quote in attribute name would catch that one |
| 14:06 | <Philip`> | is the only error I've found after seeing about a dozen that are simply skipping the space |
| 14:07 | <Philip`> | <img src="http://www.w3.org/Icons/valid-xhtml10"alt="Valid XHTML 1.0!" height="31" width="88" /> - hmm |
| 14:07 | <zcorpan_> | heh |
| 14:08 | <hsivonen> | lol |
| 14:09 | Philip` | goes away for a while |
| 14:10 | <zcorpan_> | i can't come up with a case where requiring whitespace between attributes helps to catch mistakes that quotes in attribute names don't |
| 14:12 | <zcorpan_> | unless *two* quotes were omitted, e.g. <input value="foo disabled="disabled> |
| 14:12 | <zcorpan_> | but that's very unlikely |
| 14:12 | <hsivonen> | I am not opposed to making it an error, fwiw. (at least not yet) |
| 14:13 | <zcorpan_> | if it doesn't catch any harmful mistakes, isn't it just adding noise to the list of validation messages? |
| 14:14 | <hsivonen> | that's a possibility |
| 14:15 | <hsivonen> | though this way the tokenizer spec comes closer to be hackable into an XML5 tokenizer that reports XML 1.0 4th ed. errors non-fatally |
| 14:17 | <hsivonen> | every time I patch Ælfred2 I come closer to biting the bullet and turning the Validator.nu HTML parser into a non-Draconian XML parser as well. |
| 14:17 | <hsivonen> | if it weren't for DTDs (internal subset in particular), I'd be hacking on that right now |
| 14:18 | <zcorpan_> | there are more html authors that validate their documents than there are non-drocanian xml tokenizer implementors |
| 14:18 | <hsivonen> | true |
| 14:27 | <gsnedders> | zcorpan_: even among feed readers? |
| 14:39 | <hsivonen> | the recipe at http://www.w3.org/QA/2008/03/html-charset.html contradics the HTML 5 spec |
| 14:48 | <hasather_> | annevk: the tracker uses -p for the diff command which makes it output "function receiver(e) {" and other garbage on each "@@"-line |
| 14:53 | <zcorpan_> | gsnedders: i would think so, yes |
| 14:59 | <hsivonen> | for some reason, the "HTML" page on wikipedia draws a lot of vandalism |
| 14:59 | <hsivonen> | much more than e.g. XHTML |
| 15:00 | <gsnedders> | hsivonen: but it says it isn't obsolete! it must be wrong! |
| 15:01 | <hsivonen> | aside: http://en.wikipedia.org/wiki/MathML#Example |
| 15:02 | <gsnedders> | brilliant example of why I hate MathML. It takes _so_ much code to represent anything. |
| 15:02 | <hsivonen> | Hixie: I think the MathML example is so bad the we should concede to converters being necessary and shouldn't try to make stuff implied in the text/html parser |
| 15:35 | <Philip`> | If I look for /<[a-z]+ [a-z]+="([^"]+)"[a-z][^= ]*/ in the subset of the 130K pages that have a URI whose MD5 starts with '8', then I find things like |
| 15:35 | <Philip`> | <table width="100% align="center"> |
| 15:35 | <Philip`> | <a href="<div align="center"> |
| 15:35 | <Philip`> | <a href="pages/leaders.html"Leadership""> |
| 15:35 | <Philip`> | <img src="http://cgi.knc.ne.jp/cgi/counter/counter.pl?USER=cci&URL=index.html"alt"=counter align="bottom" width="35" height="16" border="0"> |
| 15:35 | <Philip`> | <a title="Ver Video class="VERVIDEOLINK" 2005" onfocus="this.blur()" onclick="NewWindow(this.href,'vervideo','625','420','no','center');return false" href="http://raremusicvideos.org/videos/L7-the-word.html">; |
| 15:36 | <Philip`> | <a href="http://www.myrope.com/"index.asp"; target="_top"> |
| 15:36 | <Philip`> | <noscript><A HREF="http://www.world1000.com">; <img src="http://w108.hitbox.com/Hitbox?cd=1&bt;&hb;</A>; </noscript><script language="javascript1.2"> |
| 15:36 | <Philip`> | and several more that are simply forgetting a closing '"' |
| 15:38 | <Philip`> | Oops, I meant /<[a-z]+ [a-z]+="([^"]+)"[a-z][^= ]*"/ |
| 15:39 | <zcorpan_> | all of those are covered by banning " in attribute names, afaict |
| 15:40 | <Philip`> | It would probably be much more useful if I looked for all the cases that hit the relevant parse error in the tokeniser, but that would take more effort than I want to bother with now |
| 15:41 | <zcorpan_> | i think this is useful enough data for sending to the list |
| 15:42 | <Philip`> | Given what I was searching for, I don't think it'd be possible to find anything that wouldn't be interpreted as a quote in an attribute name |
| 15:44 | <zcorpan_> | good point :) |
| 16:02 | <zcorpan_> | hmm, getAttribute() returns the empty string instead of null in ie8 if the attribute is absent |
| 16:09 | <Philip`> | Is that consistent with other browsers? |
| 16:10 | Philip` | wonders why his spec-processing script creates output saying "<a�href="a">a<table><a�href="b">b</table>x" with something funny in place of the spaces, only in that particular instance and nowhere else in the document |
| 16:15 | <SadEagle> | Philip`: that's consistent with DOM, but not with the web. |
| 16:15 | <SadEagle> | Or is it other way around? |
| 16:16 | <SadEagle> | no, I got it right the first time around. |
| 16:17 | <SadEagle> | Philip`: the classic yahoo mail at least used to depend on getAttribute("missing") returning null. |
| 16:25 | <zcorpan_> | hmm, i think <input type=number value=""> should be valid html5 |
| 16:26 | <zcorpan_> | (haven't checked the spec but V.nu says it's invalid) |
| 16:29 | <annevk> | they implemented getAttribute() per spec? classic... |
| 16:30 | <zcorpan_> | i guess they went through the dom 2 core testsuite |
| 16:31 | <SadEagle> | did they start throwing WRONG_DOCUMENT_ERR too? |
| 16:32 | <zcorpan_> | dunno |
| 16:32 | <zcorpan_> | not very unexpectedly, they made changes to quirks mode and ie7 mode too |
| 16:33 | <annevk> | which makes their claims about "not breaking the Web" a lot less convincing... |
| 16:34 | <annevk> | it does further the cause of making Trident near impossible to imitate |
| 16:34 | <hsivonen> | rubys: hi |
| 16:35 | <rubys> | hsivonen: what do you mean by "above-dom"? |
| 16:35 | <hsivonen> | rubys: I mean code that reads the DOM |
| 16:35 | <Philip`> | It makes people's claims that IE would contain three independent frozen-in-time rendering engines incorrect |
| 16:35 | <hsivonen> | rubys: the SVG and MathML renderers |
| 16:35 | <hsivonen> | (and scripts that access the DOM, too) |
| 16:35 | <annevk> | Philip`, yes, that was me :) |
| 16:36 | <rubys> | cool. |
| 16:36 | <annevk> | guess I was ill informed |
| 16:38 | <hsivonen> | have I understood correctly that the IE8 CSS renderer shares the DOM impl. with the IE7 and IE5.5 modes? |
| 16:38 | <hsivonen> | but the DOM impl changes some of its behavior conditionally? |
| 16:40 | <annevk> | i think that's the case, yes |
| 16:40 | <annevk> | versioned APIs or something |
| 16:40 | <Philip`> | Would <style>a[title]{color:green}</style><a title>example</a> be a way of testing that, given what zcorpan_ said about IE8 dropping attributes? |
| 16:41 | <annevk> | probably or hasAttribute("title") maybe |
| 16:43 | <Philip`> | http://msdn2.microsoft.com/en-us/library/cc288472(VS.85).aspx - "The alt attribute is no longer displayed as the image tooltip when the browser is running in IE8 mode. Instead, the target of the longDesc attribute is used as the tooltip if present; otherwise, the title is displayed." |
| 16:43 | Philip` | hadn't seen that before |
| 16:44 | <billmason> | Given how much people mung up the content of the longdesc attribute, I have to wonder about the virtue of exposing it that way. |
| 16:45 | <hsivonen> | is there a screenshot of the longdesc thing happening? |
| 16:45 | <Philip`> | Given how little people use longdesc, it doesn't matter too much what you do with it |
| 16:45 | <billmason> | True. |
| 16:46 | Philip` | wonders what it'll do on Wikipedia images |
| 16:46 | <annevk> | hsivonen, is MathML just verbose or is it not possible to get Math compact in XML/HTML markup? |
| 16:48 | <gsnedders> | annevk: the latter |
| 16:48 | <hsivonen> | annevk: as far as I can tell, math requires lots of layout boxes, although CSS can have anonymous boxes, tweakable boxes practically have to be elements. that translates into very verbose markup |
| 16:48 | <Philip`> | Does <tex>e^{i\pi}=2</tex> count as "XML/HTML markup"? |
| 16:49 | <hsivonen> | Philip`: at least it would take anonymous box generation to a whole new level... |
| 16:49 | <hsivonen> | (I'm assuming here that we want a solution that integrates at least semi-sanely with a CSS formatter) |
| 16:50 | <Philip`> | At some level of abstraction <span>test test-test test­test test</span> does complex anonymous-box-like stuff to separate and wrap words |
| 16:50 | <rubys> | http://www.w3.org/TR/mathml-for-css/ |
| 16:50 | <Philip`> | (and I don't know what level of abstraction is the right one to look from) |
| 16:51 | <hsivonen> | rubys: not sufficient for Musings |
| 16:51 | <rubys> | understood, but he assumes that the spec is unchanging |
| 16:51 | <rubys> | if, however, you assume that the spec can be fixed, then it no longer is an HTML5 problem. |
| 16:52 | <annevk> | hsivonen, tweakable boxes could be ::foo |
| 16:53 | <hsivonen> | I'm assuming that in the near term fixing SVG and MathML as they are processed above the DOM should be out of scope for the WHATWG |
| 16:53 | <hsivonen> | I do think that the MathML 3 spec can still change, though |
| 16:54 | <hsivonen> | which reminds me that I should follow up on some of my comments on MathML 3 |
| 16:55 | <rubys> | opera's direction (last I heard) was to do mathml via css. Whether they do so, or don't do so, and whether their doing so is sufficient for Musings is not a problem that the HTML5 working group needs to solve. |
| 16:55 | <rubys> | Preventing MathML from being styled via CSS would be bad. |
| 16:56 | <hsivonen> | rubys: I agree on both counts |
| 16:56 | <rubys> | cool. I see the HTML5's WG's job to get the markup into the DOM. And that's as far as the job goes. |
| 16:56 | <hsivonen> | but as it stands, presentational MathML has useful bits that can't be emulated with a generic XML tree and CSS3 |
| 16:56 | <rubys> | then opera may have a problem with their approach. |
| 16:56 | <hsivonen> | so I think Gecko's MathML code and MathPlayer will still be relevant |
| 16:58 | <rubys> | if HTML5's HTML 5 deserialization produces the same result as HTML5's XHTML5 deserialization of equivalent documents (modulo syntax differences only), then the HTML5 working group's job is done. |
| 16:58 | <hsivonen> | rubys: I agree |
| 16:59 | <hsivonen> | which is why I think part of the solution is already constrained more than Hixie's email suggested |
| 17:02 | <billmason> | Philip`: I don't see IE8 doing anything with displaying longdesc in a tooltip, so I guess their documention is, as they say, 'preliminary'.... |
| 17:03 | <Philip`> | billmason: Okay, thanks |
| 17:03 | <Philip`> | billmason: Do you know if they still display alt in the tooltip? |
| 17:04 | <billmason> | They do not, from what I'm seeing in a quick test. |
| 17:06 | Philip` | sees that IE8 has added CSS printing support for widows and orphans, and thinks it is good that they are supporting broken families |
| 17:07 | <hsivonen> | didn't someone say that the IE7 engine was still used for printing? |
| 17:07 | <billmason> | I think their release notes say it, but I'm not looking at them right this moment. |
| 17:08 | billmason | finds the notes, and yes it's in the notes |
| 17:09 | <hsivonen> | hmm. not a total freeze of the IE7 CSS formatter then? |
| 17:11 | <Philip`> | It's still quite possible that they've only added to the IE8 CSS print formatter, but the current beta uses IE7 for printing |
| 20:15 | <annevk> | http://www.xs4all.nl/~egbg/counterscript.html |
| 21:18 | <Dashiva> | So what's the "right" way to get an arbitrary character sequence as a horizontal list delimiter. I'm stumped. |
| 21:20 | <Philip`> | Argh |
| 21:20 | <Philip`> | Remind me not to type "<meta http-equiv=x-ua-compatible" into the Live DOM Viewer, because it crashes IE before I get to the end |
| 21:21 | <Philip`> | Hmm, it just doesn't like not having a content attribute |
| 21:22 | <takkaria> | heh |
| 21:23 | <gsnedders> | hmm, I can't find anything in the XML PER that disallows characters not in Char |
| 21:25 | <Philip`> | http://philip.html5.org/tests/ie8/x-ua-compatible-crash.html |
| 21:26 | <Philip`> | It's great how the "Internet Explorer has stopped working" dialog box pops up in the background where I can't see it and won't notice it unless I happen to look at the taskbar |
| 21:26 | <gsnedders> | ah. here we go. |
| 21:26 | <gsnedders> | "Taken as a whole, it matches the production labeled document." |
| 21:29 | <gsnedders> | only the surrogate blocks, U+FFFE, and U+FFFF are disallowed. odd. |
| 21:29 | <Lachy> | annevk, othermaciej_ for selectors api, what are your thoughts about changing the interface to a single NodeSelector interface that is implemented by Documents, Elements and DocumentFragments? |
| 21:34 | <annevk> | wfm |
| 21:35 | <Philip`> | Hmm, I can't type a # (ctrl+alt+3) into IE8 |
| 21:40 | <othermaciej_> | Lachy: no problem with it |
| 21:57 | <Lachy> | ok, I'll make the change next time I edit the spec |
| 22:08 | <othermaciej> | wow, mathml is way uglier than I thought |
| 22:14 | <roc> | is MathML in Acid3? |
| 22:15 | <SadEagle> | don't think so. |
| 22:15 | <othermaciej> | no |
| 22:15 | <roc> | bummer, that's an oversight on our part :-) |
| 22:15 | <Lfe> | roc: you can read about what tests acid3 consists of here: http://www.webstandards.org/action/acid3/ |
| 22:16 | <othermaciej> | but I was just looking at the example that hsivonen sent a link to |
| 22:16 | <othermaciej> | where the quadratic formula takes 30 lines of XML to express |
| 22:17 | <SadEagle> | it's not meant to be human-readable... |
| 22:17 | <Philip`> | It only takes one line |
| 22:17 | <SadEagle> | seems like almost everyone converts to it from TeX or TeX-like syntax |
| 22:17 | <Philip`> | once you remove the whitespace |
| 22:17 | <gsnedders> | Philip`: :P |
| 22:17 | <Philip`> | since it's not like the whitespace makes it any more readable |
| 22:17 | <othermaciej> | and then the "semantic" 40-line version |
| 22:17 | <othermaciej> | it just makes me wonder what the point is |
| 22:18 | <roc> | I think the big lesson is that XML is rubbish for token-oriented languages like math and source code |
| 22:18 | <othermaciej> | humans can't read or write it, or probably even understand it enough to usefully apply CSS styling or scripting based on the MathML markup |
| 22:18 | <Philip`> | Most humans can't read or write HTML, but it's still useful for them |
| 22:19 | <roc> | othermaciej: you could say the same about SVG |
| 22:19 | <SadEagle> | othermaciej: I guess one could say that it should inherit fonts, etc., but then math fonts are special, anyway |
| 22:19 | <othermaciej> | the idea of CSS styling SVG is a little silly, but scripting it isn't totally insane |
| 22:19 | <roc> | and I think you'd be partially right, in both cases, but also partially wrong |
| 22:20 | <roc> | given decent inspector tools, it's not that hard to figure out how to script and style MathML or SVG |
| 22:21 | <othermaciej> | this is the link I'm talking about fwiw: http://en.wikipedia.org/wiki/MathML#Example |
| 22:22 | SadEagle | wonders about the quality of MathML rendering.. |
| 22:22 | gsnedders | can't render MathML well |
| 22:23 | <roc> | try ours |
| 22:23 | <othermaciej> | I guess what I'm saying is, I'm not sure that MathML is better than a <math> element which just contains magically parsed/rendered expressions would have been |
| 22:23 | <roc> | well |
| 22:23 | <SadEagle> | othermaciej: well, things like selection, etc, matter. |
| 22:23 | <othermaciej> | whereas for SVG there's no obvious better way to do it |
| 22:23 | <roc> | do you construct a math DOM or not? |
| 22:23 | <SadEagle> | of course, in most cased people just put up a PDF |
| 22:24 | <othermaciej> | gotta go for now |
| 22:24 | <roc> | if you don't, then you basically make scripting and styling impossible |
| 22:24 | <annevk> | I wonder if for text/html math could be done in such a way that you don't need to write most of the elements |
| 22:24 | <jgraham> | roc: MathML isn't in ACID3 because it doesn't have a widely implemented DOM API to test |
| 22:24 | <annevk> | they would be generated during parsing |
| 22:24 | <roc> | if you do, then using non-XML syntax could be confusing |
| 22:24 | <SadEagle> | roc: I am not sure how much of that is useful. Styling in particular. |
| 22:24 | <SadEagle> | colors, may be. |
| 22:24 | <Philip`> | annevk: Generated elements make styling and scripting harder |
| 22:25 | <jgraham> | gsnedders: Rendering MathML isn't a requirement for being a conforming MathML UA :) |
| 22:25 | <roc> | jgraham: what do you mean? you can use regular DOM APIs in our implementation |
| 22:25 | <gsnedders> | jgraham: I don't claim conformance though :) |
| 22:26 | <jgraham> | roc: What would you test that's MathML specific though? |
| 22:26 | <roc> | having an HTML <math> element that parses to MathML DOM could be very useful, but you'd need a way to put classes and IDs in there |
| 22:26 | <roc> | jgraham: stuff like MathML scriptlevel automatically changing font sizes based on formula structure |
| 22:27 | <jgraham> | roc: IIRC, that's not a normative requirement in the MathML spec |
| 22:27 | <roc> | sure is |
| 22:27 | <annevk> | Philip`, true, though probably not the common case |
| 22:28 | <SadEagle> | one would hope high-quality rendering would be an implementor's goal, though |
| 22:28 | Philip` | attempts to post bug reports to the IE8 newsgroup, which does not appear to be a place with an especially high signal-to-noise ratio |
| 22:29 | <roc> | the spec goes into some detail about how scriptlevel must interact with CSS font sizing |
| 22:29 | <roc> | I spent a few weeks implementing it last year |
| 22:29 | <roc> | I don't want to be told now that all that was non-normative :-) |
| 22:33 | <jgraham> | roc: http://www.w3.org/TR/MathML2/chapter7.html#interf.genproc (7.2.1) suggests that all the rendering rules are non-normative, but maybe I misunderstanding |
| 22:34 | jgraham | finds the MathML spec hard to understand in general |
| 22:36 | <roc> | I don't think the rules in 3.3.4.2.1 and 3.3.4.2.2 can be construed as non-normative |
| 22:36 | <roc> | http://www.w3.org/TR/MathML2/chapter3.html#id.3.3.4.2 |
| 22:56 | <jgraham> | roc: Practically I agree but pedantically "Since MathML renderers may be unable to make use of arbitrary font sizes with good results, they may wish to modify the mapping from scriptlevel to fontsize to produce better renderings in their judgment." implies that you need not change fontsize when scriptlevel changes. Although I guess there might be other testable assertions there |
| 22:57 | <jgraham> | Er, the next sentence from that quote would be useful "In particular, if fontsizes have to be rounded to available values, or limited to values within a range, the details of how this is done are up to the renderer." |
| 22:57 | <roc> | pedantically, CSS renderers are allowed to have "font-size:10px ! important" in the UA style sheet |
| 22:58 | <jgraham> | Fair enough :) |
| 22:59 | <jgraham> | Am I mistaken in thinking that MathML doesn't use RFC 2119 keywords? |
| 23:13 | <annevk> | i don't think so |
| 23:13 | <annevk> | it also doesn't appear to have any normative references? |