| 00:01 | <Hixie> | apparently the magic answer is "Apply the Unicode Collation Algorithm, using the Default Unicode Collation Element Table as customized for the <span>language</span> of the element <var title="">e</var> in the Common Locale Data Repository. |
| 00:01 | <Hixie> | " |
| 00:01 | <jamesr> | who are you expecting to do this? |
| 00:01 | <TabAtkins> | Presumably table sorters. |
| 00:01 | <Hixie> | jamesr: ICU implements the UCA already |
| 00:01 | <Hixie> | so that should be off-the-shelf |
| 00:02 | <jamesr> | and you want browsers to do this whenever table cell values change or something like that? |
| 00:02 | <Hixie> | i wouldn't say _I_ want that. but that's what people have been clamouring for, yeah. |
| 00:03 | <jamesr> | and what lang do we use for this? the lang of the table or the page or something? |
| 00:03 | <Philip`> | TabAtkins: That does seem conventional when comparing strings, but the phrase "lexicographic sort by codepoint" seems to be trying to be precise enough that I think it's not entirely unreasonable to expect it to say how lengths are handled instead of requiring you to infer it |
| 00:04 | <Hixie> | jamesr: language of the <th> column's, currently, but that's a minor detail |
| 00:04 | Philip` | might be overly sensitive to the term "lexicographic" though |
| 00:05 | <Hixie> | Philip`: i ended up just using the UCA with the defaults in that case |
| 01:00 | <zewt> | Hixie: why do you want to do locale-specific case-insensitivity? unicode case folding exists exactly to avoid that mess (for locale-insensitive case-insensitivty) |
| 01:21 | <Hixie> | zewt: eh? how does locale-specific case-insensitivity differ from unicode case folding? |
| 01:23 | <zewt> | case folding is locale-insensitive by definition |
| 01:24 | <zewt> | afaik |
| 01:25 | <Hixie> | zewt: you know wrong :-) |
| 01:25 | <zewt> | http://www.unicode.org/faq/casemap_charprop.html By contrast, case folding, which is based on the lowercase mapping, is intended to be language-neutral. Since the case folding rules do not vary by language or context, this makes them unsuitable as the basis for displaying or transforming text for human consumption. |
| 01:25 | <Hixie> | e.g. look up "turkish i" |
| 01:26 | <zewt> | i maps to i in all languages, including turkish, when you're using case folding |
| 01:26 | <zewt> | if you want language-specific lowercasing you want case mapping, not case folding |
| 01:27 | <Hixie> | all i was is a sort which when in turkish puts lowercase dotted-i and uppercase dotted-i together, and in english does not. |
| 01:27 | <zewt> | that's not case folding then |
| 01:28 | <Hixie> | which i believe is "the Unicode Collation Algorithm using the Default Unicode Collation Element Table as customized for the language /foo/ in the Common Locale Data Repository, ignoring case" |
| 05:22 | <MikeSmith> | wow an actual switch statement in the spec |
| 05:25 | <MikeSmith> | oh only in the source |
| 05:26 | <MikeSmith> | so Hixie you have some tool for converting the swtich-statement stuff in the source into prose? |
| 06:44 | <MikeSmith> | jgraham: please ping me if you are around today and have to chat about the websocket tests. I'm also getting intermittent Timeout errors for them. Maybe 25% or so of the time. |
| 09:21 | <krijnh> | teers |
| 09:21 | <krijnh> | Err, right |
| 09:22 | <krijnh> | Of joy, of course |
| 09:59 | <jgraham> | MikeSmith: Sure I am around today |
| 10:00 | <MikeSmith> | hey jgraham |
| 10:00 | <MikeSmith> | any clues on the timeouts errors? |
| 10:01 | <jgraham> | Well |
| 10:01 | <jgraham> | Nothing more than "maybe the tests are just a bit close to the timeout" |
| 10:02 | <MikeSmith> | is it possible to increase the timeout per-test? |
| 10:02 | <jgraham> | Yeah |
| 10:02 | <jgraham> | So, iirc, the default is like 1s |
| 10:03 | <jgraham> | and the default timeout for the whole file is like 5s |
| 10:03 | <jgraham> | To increase the former you pass an options dict like {timeout:timeout-in-ms} to the test constructor |
| 10:03 | <MikeSmith> | ok |
| 10:04 | <jgraham> | For the latter you do the same to the setup function |
| 10:04 | <MikeSmith> | ok |
| 10:04 | <jgraham> | Do you have a browser/test combination that gives random behaviour that I can try? |
| 10:05 | <MikeSmith> | I could try looking it at it from the server side but really that server has very llight load on it |
| 10:05 | <MikeSmith> | jgraham: no they all give me timeout errors maybe 1 out of 4 times or so |
| 10:07 | <MikeSmith> | hmm except for Opera right now |
| 10:07 | <jgraham> | MikeSmith: Well I tried http://w3c-test.org/webapps/WebSockets/tests/submissions/Opera/binary/001.html and couldn't see any timeouts |
| 10:07 | <jgraham> | in Firefox/Chrome/Opera |
| 10:08 | <jgraham> | OTOH 002.html appears to timeout in all but Chrome, but 100% of the time |
| 10:09 | <jgraham> | Maybe you just have more latency to w3c-test.org than me |
| 10:09 | <MikeSmith> | 002.html times out for me in Chrome too |
| 10:09 | <MikeSmith> | but only somtimes |
| 10:09 | <MikeSmith> | not always |
| 10:09 | <jgraham> | (this is why friends don't let friends run tests on infrastructure they don't control) |
| 10:10 | <MikeSmith> | heh |
| 10:10 | <jgraham> | What's your ping time to w3c-test.org? |
| 10:12 | <MikeSmith> | PING w3c-test.org (128.30.52.118): 56 data bytes |
| 10:12 | <MikeSmith> | 64 bytes from 128.30.52.118: icmp_seq=0 ttl=35 time=274.241 ms |
| 10:12 | <MikeSmith> | 64 bytes from 128.30.52.118: icmp_seq=1 ttl=36 time=371.074 ms |
| 10:12 | <MikeSmith> | 64 bytes from 128.30.52.118: icmp_seq=2 ttl=35 time=291.444 ms |
| 10:12 | <MikeSmith> | Request timeout for icmp_seq 3 |
| 10:13 | <MikeSmith> | 64 bytes from 128.30.52.118: icmp_seq=3 ttl=36 time=1144.780 ms |
| 10:13 | <MikeSmith> | 64 bytes from 128.30.52.118: icmp_seq=4 ttl=36 time=288.365 ms |
| 10:13 | <MikeSmith> | but hey I finally had the sense to actually look at the logs on the server and I'm seeing an error |
| 10:14 | <MikeSmith> | "mod_pywebsocket: web_socket_transfer_data raised exception for /ws/echo: Connection read error" |
| 10:16 | <jgraham> | Yeah, well if you are geting 1 in 4 packets taking > 1s that is a problem on its own :) |
| 10:16 | <MikeSmith> | yeah |
| 22:03 | <GPHemsley> | Hixie: Don't know if this is still applicable or what, but the Mac OS X sorting algorithm seems to take each embedded series of digits as a separate number, meaning hex numbers sort very unintuitively. |