06:35
<zcorpan>
"Polyglot markup begins a comment with either "<!" or "<!--"."
06:36
<zcorpan>
<! polyglot is easy !>
10:48
<SteveF>
zcorpan: are we reaching some agreement re summary/details?
10:59
<zcorpan>
SteveF: maybe
11:04
<krit>
zcorpan: ping
11:04
<zcorpan>
krit: pong
11:04
<krit>
zcorpan: Do you think you need to do more edits on DOMPoint, Quad, Rect on http://dev.w3.org/fxtf/geometry/Overview.html?
11:05
<krit>
zcorpan: otherwise I finish the DOMMatrix stuff and we are closer to FPWD, what do you think?
11:08
<zcorpan>
krit: we might want to fix https://www.w3.org/Bugs/Public/show_bug.cgi?id=24219 although i guess it doesn't need to block fpwd. i also recall some markup bugs in the switch to bikeshed but same there
11:09
<krit>
zcorpan: didn't know that we have a bug tracker for the spec already :P
11:09
<zcorpan>
krit: the spec links to it :-)
11:09
<krit>
zcorpan: :O
11:10
<krit>
zcorpan: I will go over the issues before asking for FPWD
11:30
<zcorpan>
krit: there's little incentive to implement mathml if the geometry spec doesn't use it :-P
11:31
<krit>
zcorpan: :)
12:11
<zcorpan>
Hixie: in https://www.w3.org/Bugs/Public/show_bug.cgi?id=25140 stevef proposes <details><summary id=x> <label for=x>Foo</label> </summary> ... </details> for giving a label to the disclosure triangle to get a bigger click area
12:14
<SteveF>
Hixie: zcorpan: and also to provide an unambiguous accessible name for the triangle when summary includes controls with their own label text
13:58
<MikeSmith>
zcorpan: pushed some rel-handling updates to http://validator.w3.org/nu/
13:59
<MikeSmith>
saner error messages for rel="icon shortcut" etc
14:02
<zcorpan>
MikeSmith: does it just never mention anything about url/curie? just say that it's not registered?
14:03
<MikeSmith>
zcorpan: yeah never mentions
14:03
<MikeSmith>
only says "keyword not registered"
14:03
<zcorpan>
clever
14:05
<zcorpan>
MikeSmith: the curie support means that a url will basically always be treated as a curie (except if the scheme doesn't match Name) right?
14:05
<MikeSmith>
hmm yeah
14:06
<MikeSmith>
I guess it does
14:06
<MikeSmith>
hadn't thought about that
14:06
<MikeSmith>
but dunno anyway to get around it given the rdfa requirements
14:07
<zcorpan>
i think that was pointed out as a flaw way back when curie was a thing, but the response was safe_curie which didn't actually fix the problem
14:07
<zcorpan>
the whole thing is just a fuckup
14:07
<jgraham>
People still care about RDFa?
14:07
<jgraham>
Who knew
14:07
<MikeSmith>
zcorpan: yeah safe_curie which is now "deprecated"
14:08
<MikeSmith>
jgraham: I don't care but I have to make the validator care
14:08
<zcorpan>
MikeSmith: now that'll fix the problem :-)
14:08
<scor>
jgraham: http://schema.org/LocalBusiness (see examples at the bottom)
14:08
<jgraham>
MikeSmith: One would hope you only have to make the validator care if people care. But I suppose it is also possible that people think that other people care (but those other people don't actually)
14:09
<jgraham>
zcorpan: I guess you haven't had any time to look at the unstable tests?
14:10
<jgraham>
(not a criticism of course, just a question)
14:10
<zcorpan>
jgraham: no, sorry
14:10
<zcorpan>
i was busy eating chocolate
14:10
<jgraham>
OK, I guess I will have another look at some
14:10
<jgraham>
heh
14:11
<jgraham>
I am bothered by the ones that always seem to be unstable in automation and always seem to work fine on my machine
14:11
<MikeSmith>
jgraham: I only care because otherwise my fan-mail stream gets tainted with "why you no rdfa" screeching
14:13
<MikeSmith>
zcorpan: so I guess I don't need to have the code even do any valid-url check at all in this case, since it's already first checking if the tokens are valid curiess
14:13
<zcorpan>
MikeSmith: i think so
14:17
<MikeSmith>
zcorpan: ok sad but I guess that'll make the code slightly happier since it hates dealing with rdfa so much I have to always manhandle all the rdfa handling into it
14:18
<MikeSmith>
I think this guy is reading my mind https://twitter.com/mqsiuser/status/452085394813636609
14:19
<MikeSmith>
gsnedders might also find that more acutely relevant
14:22
<jgraham>
That's an amazing non-sequitr in that post
14:22
<jgraham>
I *didn't* study computer science and still got the same outcome
14:25
<Ms2ger>
...wat
14:26
<jgraham>
?
14:27
<zcorpan>
so that means everyone will have to work with w3c's bullshit standards for decades. since everyone will either have studied computer science, or not
14:27
<Ms2ger>
The tweet
14:28
<zcorpan>
but that's silly and therefore jgraham is a fool
14:28
<jgraham>
I think that it means that the existence of Bullshit W3C Standards, and your choice of whether to work with them or not, are mostly independent of your choice of degree
14:28
<MikeSmith>
it's the double hurt of him both wasting his time studying computer science and then having to try to apply that science to teach his computer how to handle RDFa
14:29
<MikeSmith>
his computer keeps saying But this is not science you promised me science
14:29
<jgraham>
You would have thought that computers, out of all entities, would know that computer science is not science
14:30
<MikeSmith>
and clearly he's talking about RDFa there even tough maybe he's not even read the RDFa "family of specifications" yet
14:30
<jgraham>
It's like a highly specialised branch of Mathematics that thinks "Science" is neat branding
14:30
<MikeSmith>
heh
14:30
<MikeSmith>
there's even "Web Science" now
14:31
<zcorpan>
is there HTML5 Science?
14:32
<MikeSmith>
zcorpan: there was but the lab exploded
14:33
<zcorpan>
they found <plaintext> and couldn't get out of it?
14:40
<MikeSmith>
heh
14:45
<MikeSmith>
zcorpan: on the plus side now I guess I really ought to write a bunch of validator tests for rel and you will have the fun of reviewing them
14:45
<MikeSmith>
so every cloud has a silver lining
14:45
<MikeSmith>
even the rdfa cloud
16:25
<gsnedders>
MikeSmith: Do you know what RelaxNG cannot express in html? Is it just the restrictions on valid attribute values? Can it express all the various content model constraints (if not, why not)?
16:25
<gsnedders>
(Anyone else feel free to answer, ofc)
16:30
<darobin>
gsnedders: one thing it can't do is wildcard on names
16:30
<darobin>
so data-* for instance
16:30
<MikeSmith>
bingo
16:33
<gsnedders>
Anything else?
16:33
<MikeSmith>
exclusions
16:34
<MikeSmith>
like old SGML exclusions
16:34
<gsnedders>
What exclusions?
16:34
<MikeSmith>
e.g., footnote can contain all flow content except footnote
16:34
<MikeSmith>
etc
16:35
<gsnedders>
Well you can enumerate all of them no?
16:35
<gsnedders>
So surely that poses no problem?
16:35
<MikeSmith>
it's a practical problem
16:36
<MikeSmith>
anyway lemme get you a URL
16:37
<gsnedders>
Ergh, train running late, cut back to Edinburgh. Yay for n extra change.
16:37
<gsnedders>
*an
16:37
<MikeSmith>
https://github.com/validator/syntax/blob/master/non-schema/java/src/org/whattf/checker/schematronequiv/Assertions.java
16:38
<MikeSmith>
ah required ancestors
16:38
<MikeSmith>
that's one
16:38
<gsnedders>
Ah, right. I guess again technicaly possible, but that really does get ugly.
16:39
<MikeSmith>
yeah
16:43
gsnedders
wonders about trying to programmatically generating a schema, and wonders if that'll be any quicker thn just doing the whole thing manually in Python
16:43
<MikeSmith>
I'd vote for doing it all manually in python
16:44
<MikeSmith>
but it depends on what your goal is
16:44
<MikeSmith>
if for example your goal is to have good error messages, then don't use RelaxNG
16:45
<MikeSmith>
or don't use any grammar-based schema
16:45
<MikeSmith>
me wonders what you'd want to programmatically generate a schema for
16:46
<MikeSmith>
for validating something, obviously
16:48
<MikeSmith>
but unless you have some large complex set of contraints to check I wouldn't think using RelaxNG will buy you anything
16:49
<darobin>
gsnedders: yes, required ancestors and exclusions are possible but extremely painful
16:49
SamB
contemplates trying to add support for good error messages to relaxng ...
16:49
<darobin>
for instance I wrote the initial RNG for SVG, and doing something like that changing the content model of something because it was inside <a> was pretty painful
16:50
<SamB>
and yeah, not every constraint fits well at the grammatical level
16:51
<MikeSmith>
which is where assertions-based checking like schematron can be a win
16:54
<MikeSmith>
gsnedders: anyway would be fun to hear what you're needing to do
16:55
<SamB>
RelaxNG Compact schemas are of course handy for editing XML in Emacs
16:56
<SamB>
even if they don't capture every constraint
16:57
<MikeSmith>
SamB: you should add that quote to http://twoproblems.com/
16:57
<SamB>
what, you don't like completion?
16:58
<darobin>
any sentence involving Emacs makes for a good twoproblems :)
17:11
<gsnedders>
MikeSmith: In principle we want to trigger a DataLossWarning in html5lib.HTMLSerializer whenever we won't roundtrip, which is most malformed trees, but not all… I'm not really sure what I want to do. :)
17:46
<MikeSmith>
gsnedders: that sounds great but seems like it'd have no use at all for a schema
17:48
<gsnedders>
MikeSmith: Yeah, indeed. I want the API contract to be "only guarantees roundtrip for trees that match the content model restrictions"
17:49
<gsnedders>
MikeSmith: But the error messages should probably be better
17:49
SamB
ponders how you you could implmement such that you can be confident in its correctness, without actually reparsing as you go
17:49
SamB
bemoans his spelling/word-dropping ...
17:51
<gsnedders>
I'm not aiming to prove this correct. :)
17:51
<MikeSmith>
I wonder what's the value of "only guarantees roundtrip for trees that match the content model restrictions"
17:52
<MikeSmith>
what do you gain from that
17:52
<MikeSmith>
oh
17:52
<SamB>
gsnedders: well, by "be confident" I mean be reasonably sure you haven't screwed it up
17:52
<MikeSmith>
yeah for some subset of the content-model restrictions I can see why
17:53
<gsnedders>
SamB: We have enough parser tests that if they roundtrip or error correctly I'd be confident.
17:53
<gsnedders>
MikeSmith: omitting tags is especially hard in that case
17:55
<SamB>
anyway, I certainly wasn't suggesting that you actually, say, convince coq that you'd either emit a warning or achieve = trees ;-P
17:55
<MikeSmith>
oh you having it roundtripping the absence of end tags and start tags?
17:56
<gsnedders>
MikeSmith: Well, it'd be good for the ommitabletags filter to raise DataLossWarning when it cannot guarantee safety
18:38
<MikeSmith>
"whatwag" overheard just now at extensible web summit
18:38
<MikeSmith>
I like it
18:41
<hober>
i pronounce it "what wee gee", i think due to annevk saying it that way
18:41
<Ms2ger>
I thought whatwig was canonical?
18:41
<annevk>
yes, per Hixie that's the way
18:41
<annevk>
no
18:44
<jgraham>
WHAT WAG? WebHypertext Application Technology Wifes and Girlfriends? (c.f. http://en.wikipedia.org/wiki/WAGs in case someone was about to suggest that was sexist)
18:44
<Ms2ger>
That's sexi... Oh
18:44
<TabAtkins>
I do What-Wig.
18:45
<MikeSmith>
I was thinking "wag" more like in the Elizabethan-era sense
18:48
<MikeSmith>
meaning a wit
18:58
<jgraham>
MikeSmith: Well that's just silly