IETF
precis@jabber.ietf.org
Tuesday, 15 November 2011< ^ >
yone has set the subject to: précis wg :) | http://tools.ietf.org/wg/precis/ | IETF 82 | slides at https://datatracker.ietf.org/meeting/82/materials.html#wg-precis | audio at http://ietf82streaming.dnsalias.net/ietf/ietf824.m3u
Room Configuration

GMT+0
[05:33:11] yone joins the room
[07:01:34] nemo joins the room
[07:03:28] tony.l.hansen joins the room
[07:21:54] linuxwolf joins the room
[07:26:08] Andrew Sullivan joins the room
[07:26:17] KIM Kyongsok joins the room
[07:26:41] Dave Thaler joins the room
[07:26:48] <Dave Thaler> anyone remote?
[07:26:56] stpeter joins the room
[07:27:13] <Dave Thaler> I can jabber scribe but if no one is remote it'll be light...
[07:27:19] kazubu joins the room
[07:27:26] <Dave Thaler> wg deliverables
[07:28:55] <Dave Thaler> problem statement doc
[07:29:06] <Dave Thaler> no updates
[07:29:14] <Dave Thaler> framework doc
[07:29:29] ykjung83 joins the room
[07:30:19] <Dave Thaler> slide 2
[07:31:18] <Dave Thaler> 3
[07:31:38] resnick joins the room
[07:31:44] <resnick> http://xkcd.com/936/
[07:34:50] bluegirlcmh joins the room
[07:45:47] <Dave Thaler> discussion about whether subclass is useful concept or not
[07:47:16] <Dave Thaler> Andrew Sullivan: if it's different rules, coul dbe a different class rather than a subclass [scribe note: not sure I got this right]
[07:47:29] <Dave Thaler> others disagreeing with Andrew
[07:49:53] <Dave Thaler> Andrew: sometimes knowing exactly what the differences from an existing class are can be difficult to enumerate
[07:50:03] <linuxwolf> ʼ
[07:50:43] <Dave Thaler> compare to SHA-1... if you change something, it's a different algorithm
[07:51:27] <Dave Thaler> David Black: and then we'll have Precis Considerations in all our drafts :)
[07:52:20] <Andrew Sullivan> @Dave: yes that's just what I meant
[07:53:10] <Dave Thaler> NFS has both ends of the spectrum. There's a mix, with delimeter substitution
[07:56:44] <Dave Thaler> John Klensin: provide a menu and let apps pick, rather than acting as a precis service. Unicode is a long series of compromises. We just get to layer our compromises on top. Agree with Andrew on not having apps modify rules themselves.
[07:59:49] <Dave Thaler> Pete Resnick: IDNA has a mappings doc, the way people type dots may have various resulting unicode chars. Mappings doc enables conversion between input method and the protocol elements. So David Black probably means a similar mapping.
[08:00:38] <Dave Thaler> you can't have a menu, you need some name classes and some protocol elements
[08:02:22] <Dave Thaler> JK: punctuation-oids are weird. For idna, they're wrong.
[08:02:42] Dowon Kim joins the room
[08:04:47] <Dave Thaler> AS: delimeters in DNS are weird since not in wire format. you don't need large classes, you need small pieces you can build up from. You might need a new class for one piece, but may still be able to reuse pieces for other parts of a full string
[08:06:04] <Dave Thaler> PSE: we have protocols that the only differences are different disallowed characters.
[08:07:16] =JeffH joins the room
[08:08:37] <Dave Thaler> PR: "specials" don't belong in strings. "delimeters" are protocol elements, after the fact.
[08:08:56] <Dave Thaler> DB: delimeters in URIs show up on the wire
[08:09:44] <=JeffH> klensin mentioned (scary) things from the unicode world called "punctuation oids" ?
[08:11:02] <Dave Thaler> DB: sometimes use delimeters to break into pieces, other times (like iscsi) to assemble pieces into unique id
[08:11:09] <Andrew Sullivan> @Jeff: They're not actually so defined in Unicode.
[08:11:45] <Dave Thaler> PSE: what's the harm in letting apps mess up rather than us doing so generically
[08:11:50] <Andrew Sullivan> Consider U+02BC. It's a sort of modifier, in many contexts, so it's not _really_ a "letter" in quite the way you might think of "a" as a letter.
[08:12:06] <Andrew Sullivan> (In some languages, however, it works more like a letter.)
[08:12:35] <=JeffH> what was/is the thing that klensin was mentioning -- are they formally identified / denoted in unicode ?
[08:12:52] <Dave Thaler> JK: doing it generically gets it wrong for everyone, doing it on their own, someone might get it right
[08:13:02] <Andrew Sullivan> No. His point was that Unicode doesn't have a category that would be useful for us in this sort of discussion
[08:13:18] <=JeffH> lovely, thanks.
[08:13:32] <Andrew Sullivan> and for what Unicode was interested in doing, that they don't have this category doesn't matter. It just happens to matter a lot for this purpose.
[08:14:59] <Dave Thaler> Ted Hardie: all bofs are optimistic, except precis's bof wasn't. it's a support group for the horrors.
[08:15:17] <=JeffH> ok, so is there a list of such chars in say unicode 6.0 ? or are these chars that folks are sort of stumblingly realizing bit by bit, "hey, this char is going to be a problem for any apps / protocols that try to handle it in char strings" ?
[08:16:01] <=JeffH> but in any case, how does one spell the name JK gave to them, that I heard perhaps incorrectly as "punctuation oids" ?
[08:16:16] <Dave Thaler> TH: identifiers used in multiple contexts. if not unified then at least understandable is good, if there's enough pixie dust
[08:18:57] <Dave Thaler> PR: define some legit things... say "if you want internationalized ids in your protocol, write down types, and identify specials". Separate problem is how to deal with user input. Maybe make mappings doc be more theoretical.
[08:21:38] <Dave Thaler> AS: email addr, jabber id, and sip id similarly confusion is a user interface problem. they may look alike but have different syntax. expecting ids to be reusable would require a global list of specials. suspect that would fail.
[08:23:03] <Dave Thaler> PSE: subclassing was to deal with specials problem, not to give flexibility about things outside US-ASCII range.
[08:26:45] <linuxwolf> said ugly thing: <http://xmpp.org/extensions/xep-0106.html>
[08:29:35] <Dave Thaler> slide 4
[08:29:45] <Dave Thaler> bullet 3 was about mapping doc topic we discussed
[08:30:40] <Dave Thaler> PR: 2nd & 3rd bullets are user input problems
[08:33:31] <Dave Thaler> PR: new protocols might be able to avoid issue that resulted in TR46
[08:34:34] <Dave Thaler> AS: DNS means one entry in global tree and looked up by entities you have no prior relationship with, so harder than it might be in other cases
[08:35:11] <Dave Thaler> separate mapping doc underway
[08:35:32] <Dave Thaler> Not onto mapping doc preso (draft-yoneya-precis-mapping-00)
[08:35:39] <Dave Thaler> s/Not/Now/
[08:35:44] <Dave Thaler> 2
[08:36:15] <Dave Thaler> 3
[08:37:45] <Dave Thaler> 4
[08:40:00] <Dave Thaler> 5
[08:40:43] <Dave Thaler> DB: punct problem is user input problem, not map on wire. agree with sentiment in last 2 bullets
[08:42:23] <Dave Thaler> PR: punct and half/full width problems are both user input problems. all of this is about managing user expectations. And it is precis work. ambivalent about whether in framework doc.
[08:43:04] <Dave Thaler> half/full applies to all protocols, but delimeters vary by protocol
[08:46:13] <Dave Thaler> AS: in idna, user input filtered to be compatible with protocol via mapping. so not bad to keep as separate document since different phase.
[08:47:16] <Dave Thaler> PSE: limited universe of specials (e.g. things that look like apostrophes)
[08:51:23] <Dave Thaler> can we enumerate common types of mappings
[08:51:45] <Dave Thaler> PR: mappings doc settled on: half/full width, and equivalent-to-dot.
[08:52:14] <Dave Thaler> we might be able to do more, but it'll be examples, not definitive
[08:52:46] <Dave Thaler> Q: shoudl we have text about this? (lots of Yes)
[08:53:17] <Dave Thaler> Q: separate doc? 8-ish framework doc? 2-ish
[08:53:38] <Dave Thaler> Q: is this doc a good doc, if separate. People nodding.
[08:55:18] <Dave Thaler> Jeff Hodges (as protocol designer): if my protocol has a domain name slot, and I want to compare two values. how do I specify in my spec to do that.
[08:56:44] <Dave Thaler> PR: you have a string that's a representation of a domain name. (not same as in DNS where no dots)
[09:00:51] bluegirlcmh is now known as Myung Hee CHOI
[09:00:52] Myung Hee CHOI is now known as bluegirlcmh
[09:00:52] bluegirlcmh is now known as Myung Hee CHOI
[09:01:17] linuxwolf leaves the room
[09:01:31] resnick leaves the room
[09:02:40] Myung Hee CHOI leaves the room
[09:02:40] bluegirlcmh joins the room
[09:05:10] KIM Kyongsok leaves the room
[09:05:13] yone leaves the room
[09:05:25] nemo leaves the room
[09:06:07] kazubu leaves the room
[09:08:32] =JeffH leaves the room
[09:08:45] bluegirlcmh leaves the room
[09:10:48] Dowon Kim leaves the room
[09:14:45] Dave Thaler leaves the room
[09:17:03] ykjung83 leaves the room
[09:44:46] Andrew Sullivan leaves the room
[10:08:01] =JeffH joins the room
[10:10:47] =JeffH leaves the room
[10:21:34] Andrew Sullivan joins the room
[10:26:52] stpeter leaves the room: Disconnected: connection closed
[10:28:05] Andrew Sullivan leaves the room
[13:00:18] KIM Kyongsok joins the room
[13:01:11] KIM Kyongsok leaves the room
[13:02:55] ykjung83 joins the room
[13:05:19] ykjung83 leaves the room
[14:50:32] tony.l.hansen joins the room
Powered by ejabberd Powered by Erlang Valid XHTML 1.0 Transitional Valid CSS!