[02:33:38] Bjoern Hoehrmann leaves the room [09:56:54] Julian Reschke joins the room [11:32:44] Julian Reschke leaves the room [11:33:01] Julian Reschke joins the room [14:11:14] Bjoern Hoehrmann joins the room [14:12:15] In addition to your mail, the grepping didn't work quite out. There are several rules where the name is on one line, and the body on the next, so stripping the prose rules generated a number of rules without right hand side [14:12:22] currently fixing that... [14:13:05] the grepping should be applied to a variant of the ABNF that isn't line folded [14:13:14] indeed [14:16:05] Other than that the grammar parses okay so far [14:16:26] the more ugly issue are the duplicate productions that are not prose rules [14:16:32] ...referring to another part [14:17:36] -> http://www.websitedev.de/temp/httpabnf01.xml (note that the uri productions and such are incorrectly marked local references, but that should be the only issue) [14:18:15] what is producing the XML? [14:18:55] http://cpansearch.perl.org/src/BJOERN/Parse-ABNF-0.05/eg/abnf2xlx.pl [14:19:07] ok [14:23:07] I'm currently looking at the various duplicate rules [14:25:09] I see value, Reason-Phrase, parameter, ... those you mean? [14:25:19] Yes. [14:25:26] Some are intended right now. [14:25:34] Some are harmless, as they are identical [14:25:41] the others we'll have to check [14:27:14] that leaves "parameter", which, for some reason, is duplicated but not identical [14:27:23] the scipt I use is: [14:27:48] httpbis.abnf: draft-ietf-httpbis-p?-*-latest.abnf-parsed.txt cat $^ | \ fgrep -v ", defined in [Part" | \ fgrep -v "Method = %x" | \ fgrep -v 'Status-Code = "100"' | \ sort | uniq> $@ [14:31:41] one is probably for media types, the other for transfer-encoding [14:32:01] if they are supposed to be different, then we should rename one of them [14:32:44] Are they supposed to be different? I think we discussed this a few months ago, and decided that WS isn't needed in one of the cases [14:32:54] I would rename them already because they are semantically very different [14:34:50] seems like the Proposal: i67 - quoting charsets thread has some discussion [14:37:02] so in part 3, rename "parameter" to "type-parameter"? [14:37:17] I would rename the te parameter to "te-parameter" [14:37:32] so part 1... [14:37:55] (more often used productions get shorter name, also, the mime rfcs may use the same production) [14:38:04] right [14:38:35] will do right away [14:40:16] Bjoern Hoehrmann wonders what te-params is for currently [14:43:09] oh for qvalues I see [14:44:03] t-e with qvalues? [14:44:38] for TE: chunked, deflate;q=0.9, gzip;q=0.3 [14:45:53] but that has a separate grammar [14:46:01] me confused [14:46:50] "te-params" is a production used for the TE header that encodes normal parameters plus the extra q parameter [14:47:04] understood [14:47:22] but do we really want different parsing rules for qvalues and other parameters...? [14:48:08] no but separate grammars to make sense to indicate the special significance of the q parameter [14:48:19] ok [14:48:25] I am looking at RFC2616 [14:48:38] which re-used the accept-params production [14:51:23] (http://www.websitedev.de/temp/httpabnf02.xml is from the latest you mailed me, now with some pretty printing) [15:20:49] http://greenbytes.de/tech/webdav/httpbis.abnf [15:21:01] will be updated automatically from now [15:21:52] nice [15:22:53] you don't happen to keep around the RFC 2616 grammar converted to ABNF with no other changes? [15:23:05] Natürlich. [15:23:30] But it doesn't compile with BAP because quite some errors. [15:23:42] such as? [15:24:05] Rule collisions because of name matching being case-insensitive [15:24:08] With such a grammar I could diff old to new, giving, say, regular expressions for the differences. [15:24:18] prose rules crossing multiple lines [15:24:27] well that are non-issues really so far [15:24:40] Could you mail it? [15:24:45] We fixed these bugs in earlier drafts, around -02. [15:24:49] Sure. [15:25:01] But also made other changes in -02 I assume? [15:25:40] Would need to check that. [15:26:06] The "changes" Sections should have all the details. [15:26:12] okay [15:27:59] Aww that grammar has #(rules) [15:28:11] Yes. [15:28:22] That's totally not ABNF then! :-) [15:28:23] We only got rid of them last November [15:28:47] Oh well guess I'll have to dig out my own converted grammar then [15:28:58] I did it once but the #(rules) are hideous to convert [15:29:24] And the rule how to convert them is ambiguous itself [15:29:40] I would have said incomprehensible, but okay... [15:32:54] ...notified the WG about the one change, and the experimental collected ABNF [15:33:52] cool [15:34:18] I'll send you the extracted ABNFs for all drafts for diffing [15:35:26] Du hättest auch gleich sagen können du hast Langeweile... [15:35:57] Für sowas hat man Computer. [15:36:04] hehe [15:37:34] Actually I was a long struggle to get the ABNF into the shape it has now, and I'm totally happy if people run it through a real ABNF parser and see whether that actually yields useful results. [15:38:11] ...and now I'll spend some time addressing Alexey's comments on WebDAV BIND... [15:38:28] sounds like fun! [15:38:58] Lucky enough, for now it's more or less about adding a few examples. [15:53:24] Updated http://greenbytes.de/tech/webdav/httpbis.abnf -- there was a problem because one part of my Makefile was using an incorrect version of BAP (which has hard-wired knowledge how to map the # rule to OWS constructs) [18:04:23] Julian Reschke leaves the room [18:06:17] Julian Reschke joins the room [21:36:51] Julian Reschke leaves the room [21:48:11] Bjoern Hoehrmann leaves the room