03:38
<GPHemsley>
hmm... it seems I may not be the only one to allow "application/octet-stream" fall victim to find and replace
03:39
<GPHemsley>
"there actually are hardware architectures where one
03:39
<GPHemsley>
byte is 32 bits"
03:40
<GPHemsley>
his hardware architecture has a big byte, if you know what I mean
03:41
<zewt>
do i have to make the sir mixalot joke
03:49
<JonathanNeal>
What are the pros and cons of dropping the <ul> and </li> in an HTML structure like <nav><a href="1">1</a><a href="2">2</a><a href="3">3</a></nav>?
03:51
<GPHemsley>
Hixie: Does HTML use xref? I see MIMESNIFF in HTML, but it's not in references.json.
03:55
<GPHemsley>
Hixie: Also, what is this? http://www.whatwg.org/specs/web-apps/current-work/epub.html
06:25
<annevk>
GPHemsley: it doesn't
06:43
<annevk>
anyone awake that can help sketching a path API?
06:43
<annevk>
it's effectively a list of strings, but you also want to set them
06:46
<annevk>
URLPath -> list of URLPathSegment -> URLPathSegment.segment; .path[0].segment = ? it seems so cumbersome
06:47
<annevk>
and I guess you want a shortcut for the last segment, should we call that filename? ...
06:47
<annevk>
meh
06:50
<annevk>
could also do getPath and setPath
06:50
<annevk>
getPath returns a list, setPath takes the ... thingie
06:50
<annevk>
(keep forgetting the name)
06:54
<SimonSapin>
annevk: the last segment could be basename
06:55
<annevk>
ah interesting
06:57
<SimonSapin>
that’s how it’s called at least in unix and python
06:59
<SimonSapin>
annevk: what else do you want in this API? Set individual segments?
06:59
<annevk>
yeah, just give people some more control over URLs
07:00
<annevk>
I think the get/setPath thingie might be good enough, but if there's a better alternative...
07:01
<SimonSapin>
dirname is the complement to basename; not sure if it is useful
07:01
<SimonSapin>
you can already parse ".." with the previous URL as the base
07:03
<annevk>
relPathname might be interesting, resolves the new path against the old path, returns the same as pathname
07:05
<annevk>
I emailed a bit with Rodney from http://medialize.github.com/URI.js/
07:05
<annevk>
he suggested that on the server you want to have an API for the "effective TLD" (publicsuffix.org)
07:06
<annevk>
not sure how much usage that would have client-side
07:06
<annevk>
e.g. that if you have "amazon.co.uk" in your URL and you do something like URL.effectiveTLD = "com"
07:06
<annevk>
and you end up with amazon.com
07:07
<annevk>
I guess I can note it down
07:08
<annevk>
prolly also useful for cookies
07:22
<annevk>
arv: feedback welcome on https://github.com/whatwg/url/commit/e68d34885b7ec3dc44d7124e93de5bd5f1718e4e
07:22
<annevk>
abarth: ^^
07:28
<abarth>
annevk: please no effectiveTLD
07:28
<abarth>
annevk: that needs to die in a fire
07:29
<abarth>
its not interoperable
07:29
<annevk>
how can it die without security problems?
07:29
<abarth>
we shouldn't expose it in the API
07:29
<abarth>
its bad enough that we use it for cookies
07:29
<abarth>
we shouldn't be using it in new places in the platform
07:31
<annevk>
okay
07:31
<annevk>
noted
08:04
<annevk>
found a way to support parsing file: URLs without introducing extra state: https://github.com/whatwg/url/commit/c7e2c5fbc63710b5421889add11add6f7d19f67d
08:05
<annevk>
that's prolly enough for today
08:46
<asmodai>
You guys seen http://www.firstmenonthemoon.com/ ?
08:48
<WeirdAl>
anyone awake? I'm wondering about why TreeWalker.nextNode doesn't honor FILTER_REJECT.
08:52
<Ms2ger>
Blame... jst, I guess
09:58
<divya>
TabAtkins: JOIN #testtwf
10:13
<divya>
TabAtkins: https://www.w3.org/Bugs/Public/show_bug.cgi?id=19736
14:34
<GPHemsley>
So, do we talk about flags as set or unset? Or can we talk about them as set to true or set to false?
14:34
<Ms2ger>
set/unset
14:37
<zcorpan_>
GPHemsley: http://trac.tools.ietf.org/wg/websec/trac/report/1?asc=0&sort=component lists issues with mimesniff
14:37
<GPHemsley>
zcorpan_: Woah. OK, thanks.
15:00
<zcorpan_>
annevk: is the input to "parse URL" a sequence of unicode characters? or utf-16 code units?
15:09
<zewt>
unsurprised at the name of the person that reopened 12510 ...
15:10
<Ms2ger>
annevk, yep, there's a bug on IDL waiting for heycam to come back
15:11
<annevk5>
zcorpan_: code points, see public-script-coord for my UTFString proposal
15:12
<zcorpan_>
annevk5: thanks
15:12
<annevk5>
(no need to make the IETF more upset than strictly necessary
15:12
<annevk5>
))
15:13
<zewt>
somehow seems like an odd thing to define at the IDL level, though I'm not sure I could say exactly why
15:14
<zewt>
seems more like a part of apis than a language binding thing
15:17
<annevk5>
IDL does a lot of conversion, this is not really new and besides, it's a lot simpler than the alternative
15:18
<annevk5>
WeirdAl: I'm not opposed to extending TreeWalker more, but if there's just a few people using it, is it really worth it?
15:18
WeirdAl
shrugs
15:19
<WeirdAl>
actually I wanted to ask you about nextNode - why does it not honor FILTER_REJECT?
15:19
<WeirdAl>
previousNode does at least glance at that
15:20
<zewt>
isn't it pretty simple to define "execute a callback with error handling" and just use that where wanted
15:20
<zewt>
i mean, wouldn't doing this in webidl mean webidl would have to know about "window" in the first place? seems like webidl is lower level than that
15:21
<annevk5>
oh you meant the exception thing
15:21
<WeirdAl>
dammit, I hate control-Q
15:21
GPHemsley
worries that notes within a list are indented too much and may be confused with lists within lists.
15:21
<WeirdAl>
I missed whatever you said :(
15:21
<annevk5>
WeirdAl: that might be a bug actually
15:21
<annevk5>
WeirdAl: do browsers agree or disagree with the spec?
15:22
<annevk5>
WeirdAl: if they disagree it's a bug for sure...
15:22
<WeirdAl>
honestly I usually think only of Mozilla ;0
15:22
<WeirdAl>
;)
15:22
<WeirdAl>
I think their impl matches the spec. Again, I was looking for an explanation of why FILTER_REJECT doesn't factor in.
15:22
<annevk5>
so the new spec is based on Gecko, might get circular
15:23
<WeirdAl>
figures :p
15:24
<annevk5>
the spec has a
15:24
<zewt>
n unwanted newline
15:24
<WeirdAl>
and actually I wasn't referring to any exception - just that when I am done iterating over currentNode's descendants, later nodes might ignore FILTER_REJECT
15:25
<annevk5>
`<!-- XXX WTF? -->` for sibling traversal
15:27
<WeirdAl>
hehe
15:27
<annevk5>
zewt: dunno about the exception thing, but DOM does not know about the global object, IDL does I think at the binding level
15:28
<annevk5>
WeirdAl: bug + testcase would help, but if the other browsers agree with Gecko there's not much we can do I think
15:28
WeirdAl
frets
15:28
GPHemsley
wonders how annevk uses flags and <b>
15:29
<WeirdAl>
annevk - maybe I should ask smaug - he's familiar enough to know the why's
15:30
<annevk5>
GPHemsley: unset/set
15:30
<annevk5>
GPHemsley: <b> is when I don't want to use <dfn>
15:31
<GPHemsley>
annevk5: In what conditions does that occur?
15:33
<annevk5>
GPHemsley: lazy, or when <dfn> feels like overkill
15:33
<zewt>
annevk: fyi, onerror is also fired as a real DOM event
15:33
<annevk5>
mostly the latter I guess
15:35
<GPHemsley>
annevk5: What's the SOP for using words like perform/execute/rules/steps, etc.?
15:35
<zewt>
(much more usefully, too, at least in chrome; the DOM event gives separate "message", "file", "lineno" properties, where window.onerror is just a string)
15:37
<annevk5>
zewt: window.onerror is a callback with multiple arguments
15:37
<annevk5>
zewt: all of those...
15:37
<annevk5>
GPHemsley: whatever fits the context
15:38
<zewt>
oh yeah, can never remember that, but anyway the point is webidl surely shouldn't be firing events
15:38
<GPHemsley>
annevk5: Alright, well, you let me know if it doesn't fit the context :P
15:38
<annevk5>
zewt: that's not the suggestion
15:39
<annevk5>
HTML defines window.onerror
15:41
<annevk5>
IDL defines invoking callbacks and it should define where their exceptions go
15:41
<annevk5>
or provide some hook for HTML to hook into, but this is not a matter for the DOM
16:33
GPHemsley
wonders what to do with a lot of ands and ors
16:38
<Philip`>
GPHemsley: Disjunctively normalise them
16:39
<GPHemsley>
Philip`: Can you give me an example? I have ( x && ( y || ( z && w ) ) )
16:40
<Philip`>
(x && y) || (x && z && w)
16:41
<GPHemsley>
aw, boo
16:41
<GPHemsley>
alright
16:41
Philip`
has no idea what the context of the question was, and wasn't attempting to give a serious answer
16:42
<GPHemsley>
well, it solves the problem
16:42
<GPHemsley>
it just duplicates wordage
16:44
<Philip`>
Define short terms to refer to the wordy concepts
16:48
<GPHemsley>
it's alright
16:48
<GPHemsley>
they're fairly short already
16:53
<hsivonen>
SGML reverse Huffman coded the terminology and gave the longest names to the most commonly talked about concepts
16:53
<hsivonen>
as in external document type definition subset
17:03
<Hixie>
GPHemsley: my references system is to put all the references at the bottom of the source file, and then have code to filter out the ones that aren't used
17:06
<GPHemsley>
Hixie: You like to buck every system, huh? :P
17:12
<Hixie>
to be fair, the html spec predates most of these systems :-P
17:51
<GPHemsley>
Comments welcome on my latest mimesniff commit
17:56
GPHemsley
wonders if he is the only one who ever uses the wiki
19:08
<annevk>
GPHemsley: x and either y or z and w
19:09
<annevk>
(answering your question above, not saying you have to write it like that)
19:11
<annevk>
GPHemsley: convention is to lowercase RFC2119 terms
19:12
<annevk>
GPHemsley: convention is to end flag terms with " flag"
19:12
<annevk>
GPHemsley: convention is to lowercase user agent
19:12
<annevk>
GPHemsley: convention is to put <code title> around strings, such as "<code title>text/xml</code>"
19:14
<annevk>
GPHemsley: I would combine the "Begin reading" and "Continue reading" steps as "Read bytes from the resource into /buffer/ until one of ..."
19:17
<annevk>
GPHemsley: as for your XXX comments, did you check whether e.g. "image/blahblah" is sniffed?
19:17
<annevk>
GPHemsley: because if that's not going to be snfifed, there's your reason ;)
19:18
<annevk>
GPHemsley: oh, and I use the wiki :)
20:28
<JonathanNeal>
Surely <nav> handles this now? http://www.w3.org/TR/WCAG10-HTML-TECHS/#group-bypass
20:45
<annevk_>
Transmission.app software update: "Fix crash when adding magnet links with malformed webseeds"; "Fix handling of magnet links' webseed URLs that contain whitespace"; but but but URI implementations are interoperable!!1!
20:47
<zewt>
i thought the argument was "so what if implementations aren't interoperable, that's not the spec's problem"
20:47
<zewt>
not that the arguments are necessarily coherent
20:55
<stevefaulkner>
JonathanNeal: how does <nav> handle it?
20:55
<annevk_>
Yeah dunno, I guess I'll hear a few more next week
20:57
<JonathanNeal>
stevefaulkner: i missed the <img>, I was looking into an accessibility issue, where I was told to include non-link, printable characters (surrounded by spaces) between adjacent links.
20:57
<JonathanNeal>
I want to do something like <nav><a href="#foo">Foo</a> <a href="#bar">Bar</a></nav>
20:59
<stevefaulkner>
JonathanNeal: <nav><a href="#foo">Foo</a>, <a href="#bar">Bar</a></nav> ?
20:59
<JonathanNeal>
Yea, dropping the ul and li's.
21:01
<stevefaulkner>
JonathonNeal: i wouldn't use wist for 2 links but if you have a set of related links a list is the best way to group them
21:01
<JonathanNeal>
How would a <nav> fail to do this?
21:02
<stevefaulkner>
JonathanNeal: it does not convey a a list semantic in any browser
21:03
<JonathanNeal>
The nav element represents a section of a page that links to other pages or to parts within the page: a section with navigation links.
21:03
<JonathanNeal>
I don't see how sub-organizing those links into a list is necessary. It could be done, certainly, but it's no longer necessary.
21:04
<stevefaulkner>
JonathanNeal: what it says it does in the spec is a fiction unless the semantics are expressed via browsers
21:05
<JonathanNeal>
Well, when I tested it with VoiceOver and Jaws I did not experience a loss.
21:06
<JonathanNeal>
It certainly doesn't break any semantics.
21:06
<stevefaulkner>
JonathanNeal: a list conveys explicitly a set of items, <nav> or any other grouping semantic does not
21:10
<stevefaulkner>
JonathanNeal: <nav> provides less structural information as implemented and in some browsers (IE for example) does not even convey the fact that its navigation links
21:11
<stevefaulkner>
<nav role="navigation> has better support across browsers and AT
21:22
<JonathanNeal>
Sure, though isn't <nav> supposed to inherently have that role?
21:30
<zcorpan>
annevk: is there a place already where url test cases should live?
21:33
<annevk>
prolly the same place AryehGregor ends up putting the encoding tests?
21:33
<annevk>
https://github.com/annevk/url has some tests fwiw, mostly copied from WebKit with some changes
21:33
<annevk>
(changes are annotated)
21:34
<annevk>
(but that parser does not yet have file URL parsing)
21:35
<annevk>
Why does http://www.whatwg.org/specs/web-apps/current-work/multipage/syntax.html not mention the Unicode restrictions made in the parser?
21:35
<annevk>
E.g. disallowing U+3FFFF
21:49
<annevk>
Context: https://www.w3.org/Bugs/Public/show_bug.cgi?id=19743
21:53
<jgraham>
Things that make me sad: specs without dfn.js
21:54
<GPHemsley>
annevk: Ah, convention! That's the word I've been looking for!
21:55
<GPHemsley>
annevk: Regarding my XXX comments: What if an image is tagged "image/blahblah" but is actually "image/gif"?
21:57
<jgraham>
Hmm, so why is the index argumetn of HTMLTableElement.insertRow(index) optional?
21:57
<GPHemsley>
annevk: Would <code> also be appropriate for the "Bytes in ASCII" values?
21:57
<jgraham>
Does it match IE?
21:57
jgraham
doesn't have IE
21:57
<jgraham>
Apparently it doesn't match anything else
21:58
<jgraham>
Oh, maybe it matches WebKit
22:04
<GPHemsley>
annevk: Why <code title> rather than <code>?
22:13
<stevefaulkner>
JonathanNeal: yes, but when it will be implemented across browsers and platforms is anybodies guess > http://html5accessibility.com
22:18
<JonathanNeal>
stevefaulkner: that's true, and Firefox only forwards the role right now, but what accessibility software is using ARIA roles like nav?
22:18
<JonathanNeal>
That's not a remark meant to be contrary, I'm really curious.
22:21
<GPHemsley>
annevk: So, are you saying it's not necessary to explicitly state that byte-reading should begin with byte 0?
22:22
<stevefaulkner>
JonathanNeal: all main screen readers support navigation via ARIA landmark roles, here is some data from a almost a year ago http://www.html5accessibility.com/tests/landmarks.html support has improved further since then
22:22
<annevk>
GPHemsley: I'm saying that, and byte 0 is poor wording as one might take it to mean byte 0x00
22:23
<annevk>
GPHemsley: <code title> to avoid accidental linking
22:23
<GPHemsley>
annevk: How do I ensure that reading does not begin at some arbitrary point within the resource?
22:24
<GPHemsley>
annevk: I have not seen <code> be linked or otherwise marked up differently anywhere (unlike, e.g., <span>). In what situation does that happen?
22:24
<annevk>
GPHemsley: Read from resource into /buffer/, starting at the first byte, and run these steps:, maybe?
22:24
<annevk>
GPHemsley: <dfn>x</dfn> <code>x</code>
22:24
<GPHemsley>
that's confusing
22:24
<GPHemsley>
(the latter)
22:25
<annevk>
GPHemsley: Did you test Content-Type:image/somethingunsupported with an image whose actual type is image/gif?
22:25
<annevk>
GPHemsley: <code> links just like <span>
22:26
<GPHemsley>
I haven't tested anything yet
22:26
<GPHemsley>
I'm just working with the existing wording at this point
22:27
<JonathanNeal>
stevefaulkner: have experience with JAWS? I can't get it to read Internet Explorer pages to me.
22:27
<annevk>
GPHemsley: I'm pretty sure that's why the wording is that way
22:27
<annevk>
GPHemsley: because it's not sniffed in that case
22:29
<stevefaulkner>
JonathanNeal: yeah, but could be many reasons, try refreshing the page, also try insert+z to turn virtual cursor off and the same to turn back on after pause, that refreshes JAWS virtual buffer may help...
22:29
<GPHemsley>
annevk: Frankly, I'm somewhat confused as to why images are sniffed at all. Basically, the algorithm is only sniffing around the arbitrary image types that the UA supports.
22:30
<GPHemsley>
but I'll look into it further later
22:30
<annevk>
because people put a PNG on the web with an image/gif media type
22:30
<annevk>
or a PNG with a image/x-icon media type (I've done that and took me years to notice because it just worked)
22:31
<annevk>
and don't try to find logic, that's not the way this works
22:39
<annevk>
GPHemsley: anyway, have fun :) bedtime here
22:39
<GPHemsley>
ah, alright
22:39
<GPHemsley>
g'night
22:46
<JonathanNeal>
JAWS 14 reports <nav> in IE8 as "navigation region", without role="navigation".