| 00:06 | <gsnedders> | I can't find anything, at all? |
| 00:06 | <gsnedders> | Did this get dropped with pre-processing the input stream? |
| 00:08 | <SimonSapin> | gsnedders: in CSS? It’s in "consume an escaped code point" |
| 00:08 | <gsnedders> | SimonSapin: HTML. |
| 00:08 | <gsnedders> | I thought this would be assumed given #whatwg :P |
| 00:08 | <SimonSapin> | character encodings’ decoders never emit them |
| 00:10 | <SimonSapin> | http://www.whatwg.org/specs/web-apps/current-work/multipage/tokenization.html#consume-a-character-reference does the rest |
| 00:11 | <gsnedders> | SimonSapin: the encoding spec doesn't say that? |
| 00:12 | <SimonSapin> | gsnedders: not explicitly, but none of the decoders emit them |
| 00:12 | <gsnedders> | UTF-8 appears to be able to? |
| 00:13 | <gsnedders> | Unless I'm being blind? |
| 00:14 | <gsnedders> | And noncharacteres? |
| 00:14 | <gsnedders> | s/noncharacteres/perm. noncharacters/ |
| 00:16 | <SimonSapin> | "If byte is 0xED, set utf-8 upper boundary to 0x9F. " and the like prevent surrogates |
| 00:16 | <gsnedders> | Okay, I'd tried to quickly reason with that in my head but failed. :P |
| 00:17 | <SimonSapin> | u'\uD7FF'.encode('utf8') == b'\xed\x9f\xbf' |
| 00:17 | <gsnedders> | Well, that's a nice bug. |
| 00:17 | <gsnedders> | Given that's never been legal. |
| 00:17 | <SimonSapin> | \uD800 would be \xed\xa0\x80 |
| 00:18 | <gsnedders> | Oh, wait, U+d800 is the first. |
| 00:18 | <gsnedders> | Duh. |
| 00:18 | <SimonSapin> | which is forbidden by the "boundaries" |
| 00:18 | <gsnedders> | This is why I cannot read. |
| 00:18 | <gsnedders> | read? |
| 00:18 | <gsnedders> | reason. |
| 00:18 | <gsnedders> | I'm blaming jetlag for my inability to make sense, despite not really being jetlagged at all. :P |
| 00:19 | <SimonSapin> | the decoders emit "scalar values" http://www.unicode.org/glossary/#unicode_scalar_value , which exclude surrogates but not other non-characters |
| 00:19 | <gsnedders> | Right. |
| 00:19 | <gsnedders> | So the former handling of permament non-characters is gone? |
| 00:19 | <SimonSapin> | I don’t know |
| 00:20 | <SimonSapin> | it’s not in the Encoding layer at least |
| 00:20 | <gsnedders> | Doesn't appear to be in HTML either. |
| 00:20 | <gsnedders> | Of that former section only U+0000 still exists, by virtue of it now being handled inline in the parser. |
| 03:44 | <gsnedders> | So, the input stream is defined in terms of a series of the HTML speeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeec`. |
| 03:45 | <gsnedders> | Um, okay. Something went wrong there, and I couldn't stop that. |
| 03:48 | <gsnedders> | s/of .*$/of what the HTML spec defines a "Unicode codepoint" means — do we really want lone surrogates to be valid?/ |
| 03:53 | <zewt> | valid where? |
| 03:58 | <gsnedders> | A valid HTML document. |
| 04:01 | <gsnedders> | Or a valid fragment. Say from innerHTML. |
| 04:02 | <Hixie_> | "valid" in what sense? |
| 04:02 | <gsnedders> | Produces no parse-errors. |
| 04:03 | <gsnedders> | Which is a conformance requirement. |
| 04:03 | <zewt> | iirc you can't get lone surrogates in an HTML document, since the charset decoders won't emit them |
| 04:04 | <gsnedders> | zewt: Yes, you need to be going from an abstract source, like the innerHTML case from the DOM. |
| 04:05 | <zewt> | not sure what you're asking; you can insert them with script, but an actual html document won't generate them |
| 04:09 | <zewt> | are you saying div.textContent = "\ud800" should throw an exception or something? |
| 04:09 | <gsnedders> | Right, but that doesn't exempt you from conformance, merely from conformance-checkers having to check it :P |
| 04:09 | <gsnedders> | No, be non-conforming and trigger a parse error. |
| 04:09 | <zewt> | "parse error" where? |
| 04:10 | <zewt> | i mean, an error in which parser? |
| 04:14 | <gsnedders> | zewt: Most browsers can report parse errors to dev tools |
| 04:14 | <gsnedders> | From the fragment case from innerHTML. |
| 04:15 | <Hixie_> | gsnedders: i don't think it cares about it in the parser, but there's separate conformance errors about there not being lone surrogates |
| 04:15 | <gsnedders> | Where? |
| 04:16 | <Hixie_> | a fascinating question |
| 04:16 | <gsnedders> | There *used* to be in the clause prohibiting perm. noncharacters |
| 04:18 | <Hixie_> | "Text nodes and attribute values must consist of Unicode characters, must not contain U+0000 characters, must not contain permanently undefined Unicode characters (noncharacters), and must not contain control characters other than space characters." |
| 04:18 | <Hixie_> | + "The term Unicode character is used to mean a Unicode scalar value (i.e. any Unicode code point that is not a surrogate code point)." |
| 04:22 | <gsnedders> | Where's that? |
| 04:23 | <gsnedders> | http://www.whatwg.org/specs/web-apps/current-work/multipage/infrastructure.html#unicode-code-point gives the definition of codepoint differently? |
| 04:23 | <Hixie_> | what i pasted doesn't use the term "code point" |
| 04:23 | <zewt> | (it does a little) |
| 04:26 | <gsnedders> | Blarf! |
| 04:27 | <gsnedders> | Hixie_: So it's a conformance requirement yet not a parse error? |
| 04:28 | <Hixie_> | zewt: not normatively :-) |
| 04:29 | <gsnedders> | Hixie_: Although everything else in that list of what they cannot contain is a parse rror? |
| 04:29 | <gsnedders> | *error |
| 04:29 | <gsnedders> | AFAIK you removed it on grounds that encodings cannot output lone surrogates. |
| 04:31 | <Hixie_> | removed from the parser? |
| 04:32 | <gsnedders> | From the preprocessing the input stream clause that forbids null, noncharacters, etc. |
| 09:16 | <Ms2ger> | Hm, the box at http://www.whatwg.org/specs/web-apps/current-work/multipage/workers.html#WorkerGlobalScope-partial lists a bug that has been closed for months |
| 09:16 | Ms2ger | wonders who maintains that |
| 10:48 | <Ms2ger> | Anyone around who understands the parser? |
| 10:50 | <Ms2ger> | Maybe zcorpan |
| 11:28 | <zcorpan> | Ms2ger: what parser? |
| 11:30 | zcorpan | isn't actually here |
| 17:04 | <Ms2ger> | zcorpan-not-here: html |
| 17:17 | <jgraham> | I love XHTML. I love the way I can get a parser error in gecko loading a page over a bad connection |
| 17:18 | <tantek> | nice |
| 17:18 | tantek | presumes jgraham means "XML" in general |
| 17:18 | <Ms2ger> | jgraham, you know a thing or two about the html parser, right? |
| 17:18 | <jgraham> | Well I suppose the other alternative is that the connection isn't the problem and the content is broken |
| 17:19 | <Ms2ger> | I can't figure out what "If the stack of open elements has a button element in scope, then run these substeps:" means |
| 17:19 | <jgraham> | But the fowc (Flash of Working Content) before the YSOD is... irritating |
| 17:19 | <Ms2ger> | "has a button element in scope" links to "The stack of open elements is said to have an element in scope when it has an element in the specific scope consisting of the following element types:" |
| 17:19 | <jgraham> | Ms2ger: What so you mean "means"? |
| 17:19 | <Ms2ger> | Note no reference to the "button" part |
| 17:20 | <Ms2ger> | My stack is [html, body] |
| 17:20 | <Ms2ger> | Do I have a button element in scope? |
| 17:22 | <jgraham> | Ms2ger: No. The spec is confusing here because there are two different things |
| 17:22 | <jgraham> | The element that may or may not be in scope |
| 17:22 | <jgraham> | and the set of elements that delimit the scopes |
| 17:23 | <jgraham> | Having a button element in scope is not the same as having an element in button scope |
| 17:23 | <Ms2ger> | Right |
| 17:23 | <Ms2ger> | I suspected that was the case |
| 17:23 | <jgraham> | http://www.whatwg.org/specs/web-apps/current-work/multipage/parsing.html#has-an-element-in-the-specific-scope |
| 17:23 | <Ms2ger> | But I can't find a clear definition of having a button element in scope |
| 17:24 | <jgraham> | "target node" in that algorithm is a html:button element |
| 17:24 | <Ms2ger> | (I can find a definition of "having an element in button scope", but I don't understand the definition) |
| 17:24 | <jgraham> | But I don't think it actually says that explicitly |
| 17:24 | <jgraham> | File a bug |
| 17:25 | <Ms2ger> | Filed https://www.w3.org/Bugs/Public/show_bug.cgi?id=23119 |
| 17:26 | <jgraham> | I think instead of talking about "target node" it should be clear that it is an algorithm taking a list and an element name (or an namespace, name pair; I'm not sure) |
| 17:26 | <jgraham> | And say "if node is an /element/ element" or something |
| 17:26 | <Ms2ger> | This code has been implemented interoperably at least four times, and it still isn't clear \o/ |
| 17:26 | <jgraham> | I think this part of the spec changed |
| 17:26 | <Ms2ger> | Yeah, it might have |
| 17:27 | <jgraham> | But the namespace part was never clear |
| 17:27 | <Ms2ger> | Okay, that helps |
| 17:27 | Ms2ger | goes back to reviewing this one-line test case |
| 17:28 | <jgraham> | In general it would be good if Hixie was explict for each algorithm what the imputs and ouptputs are |
| 17:28 | <jgraham> | Even if it meant having some special convention to represent it |
| 17:28 | <Ms2ger> | (And I've consumed the first token, yay) |
| 17:29 | <Ms2ger> | Also, the order of the switch cases in the parser seems to be random |
| 17:31 | <jgraham> | ElementInScope :: Stack Element -> LocalName -> [LocalName] -> Bool |
| 17:33 | <Ms2ger> | Or ElementInScope :: Stack Element -> (LocalName, NS) -> [(LocalName, NS)] -> Bool? |
| 17:33 | <jgraham> | Right |
| 17:33 | <jgraham> | (dunno if tuples can be represented like that in Haskell, but this is at best pseudo-Haskell) |
| 17:34 | <Ms2ger> | Yeah, that's fine |
| 17:34 | Ms2ger | wrote some Haskell this year |
| 17:35 | <Ms2ger> | (Oh, the list is even explicitly [(LocalName, NS)] now |
| 17:35 | jgraham | should probably learn enough to do something useful |
| 17:36 | <Ms2ger> | I dunno |
| 18:07 | <Ms2ger> | gsnedders, land https://critic.hoppipolla.co.uk/r/193 already :) |
| 19:46 | <Yuhong> | https://news.ycombinator.com/item?id=6307219 |
| 19:46 | <Yuhong> | Someone wishes that conforming HTML required closing tags. |