[02:01:51] frank joins the room [02:02:08] frank leaves the room [10:11:48] Jaeyoun Kim (Peter) joins the room [10:11:58] Jaeyoun Kim (Peter) leaves the room [10:34:46] Lee Hyeong Woo joins the room [10:35:17] Jaeyoun Kim (Peter) joins the room [10:36:01] Lee Hyeong Woo leaves the room [10:36:20] Jaeyoun Kim (Peter) leaves the room [12:56:00] avri joins the room [12:59:48] tonyhansen joins the room [12:59:48] tonyhansen leaves the room [13:01:29] avri leaves the room [13:43:59] frank joins the room [13:45:39] tonyhansen joins the room [13:50:00] frank has set the subject to: BiDi toplabels, binary eszetts, and other rational protocol tables [13:54:45] tonyhansen leaves the room [13:55:44] tonyhansen joins the room [14:12:22] tonyhansen leaves the room: Replaced by new connection [14:12:22] tonyhansen joins the room [14:16:08] simon joins the room [14:19:57] Barry Leiba joins the room [14:20:09] Lisa Dusseault joins the room [14:20:54] Randall Gellens joins the room [14:22:26] yone joins the room [14:24:03] shinta joins the room [14:24:12] hta joins the room [14:24:13] swshin92 joins the room [14:24:46] fujiwara joins the room [14:25:39] Jaeyoun Kim (Peter) joins the room [14:26:02] Chris Newman joins the room [14:27:10] bortzmeyer joins the room [14:27:13] avri joins the room [14:28:21] Andrew Sullivan joins the room [14:30:49] xiaodong.lee joins the room [14:31:07] koji joins the room [14:31:16] ogud: Olafur Gudmundsson joins the room [14:34:01] healthyao joins the room [14:34:11] cary joins the room [14:34:59] Jaeyoun Kim (Peter) leaves the room: Replaced by new connection [14:36:23] is anyone planning on scribing here? [14:37:08] Sorry, I thought we had a jabber scribe [14:37:12] I'll do it [14:37:17] elwynd joins the room [14:38:00] Andrew: Thanks, I get only 30% of the audio :-) [14:38:48] Plan is to make recommendation between now & tomorrow by which decision gets made about how to express the rules and where to express them [14:38:52] simon leaves the room [14:39:32] lynch joins the room [14:39:48] (no scribe: I confess I am confused why we are talking about which document to put the rules in when there seems to be a move afoot to reduce the document set size to 1) [14:40:17] PHB joins the room [14:40:18] andres: I think we are deciding between the format of -protocol appendix A and -protocol appendix B. [14:40:34] andrew: .... and I don't think the movement will succeed. [14:41:07] hta: yes, sorry, but John also started by suggesting the whole thing move to -tables [14:41:23] [scribe hat back on] [14:41:30] .. which I think is the right thing. [14:41:50] (speaking very softly since I didn't get to -protocol in pre-meeting reading) [14:41:52] Paul Hoffman: everyone's going to have to write a 1st-round parser whatever happens, so try not to optimise [14:42:03] Mark Davis @ mic [14:42:22] break down, the expressions get much simpler, then things are more widely supported [14:42:36] John K: yes, but now 1/2 way to the other approach [14:42:54] Vint Cerf: take away, make proposal tomorrow [14:43:05] Mark D., John K, Patrik F. going to do this [14:43:10] anyone else who cares and can help? [14:43:27] WPM01906 joins the room [14:43:42] Vint: other issues to address on protocol? [14:43:55] John: what tests are appropriate for BiDi? [14:44:43] Mark Davis to mic [14:44:48] couple more issues [14:44:52] 1. mapping issue [14:45:10] see list posting on this [14:45:36] 2. like to see normative parts of rationale should be pulled into protocol or tables or bidi [14:45:45] talk about when get to rationale [14:45:53] save for later [14:46:37] Moving to BiDi [14:46:51] Harald Alvestrand to present [14:47:19] Lisa Dusseault leaves the room [14:47:35] Jaeyoun Kim (Peter) joins the room [14:47:51] not much change since last time. No protocol change. Hasn't seen relevant discussion heading towards consensus to justify [14:47:54] some additional text [14:48:28] issues from Philadelphia: [14:49:19] 1. can we accept strings that mix contexts (LTR & RTL)? (I think this is what he said, it's not completely clear to me) [14:49:25] 2. Inter-label tests [14:49:35] now of the opinion that this ought to be a registration rule [14:50:12] Pete Resnick to mic [14:50:28] suggests to punt the whole set of problems [14:50:34] general set : 2 problems [14:50:48] 1. strings in RTL context come out looking different than LTR context [14:50:56] Harald: described as ok in document [14:51:03] edmon joins the room [14:51:08] 2. numbers at end, which become ambiguous [14:51:25] Harald: requirement in document [14:51:49] issue still with numbers on both sides of label [14:53:38] arabic and hebrew too - not just dead languages [14:53:54] Harald: there are exactly 2 requirements [14:54:02] narooter joins the room [14:54:10] he hereby reads from Book of Harald [14:54:31] (reads) [14:55:54] See sec 3 of -bidi- [14:56:08] Pete says he wants to relax them [14:56:11] John Kelnsin [14:56:16] at mic [14:57:01] Argues that checks of more than one label at a time is bad [14:57:18] if to impose a requirement on bidi that it never look at > 1 label at a time [14:57:24] what happens to current reqs? [14:57:46] Harald: can still satisfy if you accept that disallow numbers at both ends of a string with RTL chars [14:59:12] John K: still willing to do because the users may not have a consistent experience, worse than breakage in some small cases [14:59:17] Paul Hoffman: [14:59:32] put rules in the document that work, and take out the why to do them [14:59:34] no rationale [15:01:23] Harald: not having a justification was a fundamental reason why IDNA2003 broke [15:01:36] really wants rationale [15:02:05] tonyhansen leaves the room [15:02:09] xiaodong.lee leaves the room [15:02:13] Paul Hoffman: are you (Harald) happy with the changes proposed? [15:02:28] Harald: yes, feel both necessary [15:02:32] Ted Hardie [15:03:04] idea that we are considering that we might let characters to move across label delimiter is clinically insane [15:03:08] "dots of doom" [15:03:12] "leave me my sanity" [15:03:13] &c [15:03:18] mayhem mayhem, [15:03:19] tonyhansen joins the room [15:03:47] Even if the restrictions might disadvantage certain language communities [15:04:04] the overall stability of individual labels is way more important [15:04:16] strong endorsement of requirement number 2 [15:04:26] John Klensin: principle of least horrification [15:04:29] agrees with Ted [15:04:56] also an argument in favour of rationale [15:05:14] because without reason for strange rule, people will break them almost immediately [15:05:46] more worried about crossing label boundaries than imposing seemingly strange rules which can at least be explained [15:05:55] Mark Davis (?) at mic [15:06:09] Yes, he is Mark Davis [15:06:44] Argument in favour of rationale part [15:06:57] some minor preference that the rationale part be in the rationale document [15:07:19] argues for rules in document right now [15:07:42] Chris Newman leaves the room [15:07:46] can live with either more stringent single-label rule or nulti-label rule [15:07:49] Pete Resnick again [15:07:54] "sacrosanct dot" [15:08:57] anyone who deals with non-ASCII characters will need to do work on the domain name before they talk to the DNS anyway [15:09:24] How about replacing dot as delimiter with null as delimiter :-) [15:09:48] ctrl-A [15:10:19] Ted Hardie: how come, Pete, you don't share your crack with me? [15:10:47] Just suggested you invent new pretend delimiters to solve the problem we created for ourselves? [15:10:58] Pete: not going to solve at data level, have to solve at presentation level [15:11:41] ok with "no numbers at beginning and end" but bigger set of restrictions not ok [15:11:44] Randall Gellens: because most people do not see the null character :-) [15:11:57] Harald outlines the rules for 1st character [15:13:01] John Klensin: if you change the dots, there are applications and protocols that will break in nasty ways [15:13:41] trick rules to the DNS names working not going to solve the problem. Very high-level presentation layer [15:14:37] argues in favour of rules as written plus change of adding prohibit at end [15:14:46] and eliminate inter-label checking [15:15:03] Harald: [15:15:41] example of how something appeard to him [15:16:12] Pete R: [aleph].3com.com in RTL context, gets something very surprising [15:16:15] Dave Crocker: [15:16:33] doesn't understand the topic and so is going to try to talk about what he does understand [15:16:43] this is at presentation layer [15:17:01] trying to 2d guess presentation layers and their behaviour in presence of "." [15:17:33] Harald: cannot possibly adjust the total amount of insanity, but adjust it to conform to Unicode standard [15:18:22] Dave: understood PEte's proposal to be, "the dot is the problem, get rid of it at the presentation layer." [15:18:26] Is that close? [15:18:32] Pete: [15:18:56] no. "Use something different than dot" is the only way to _solve_ the problem [15:19:16] and [aleph].3com.com will be legal even if 3d rule in place [15:19:28] no need for things to be displayed identically in RTL and LTR contexts [15:19:35] Harald: not the goal, & not achieved [15:19:53] when Unicode bidi alg. applied to string [15:20:01] the result doesn't cause characters to move over dots [15:20:15] Pete: dots are stationary in memory on the machine [15:20:36] Name is [unclear] [15:20:40] at mic [15:20:46] He is Alirezah Saleh [15:20:52] oh, thanks [15:21:39] removing the rules on RTL char. at end of label doesn't solve problem of [aleph].3com.com, as Pete says [15:21:47] need to test this [15:21:54] WPM01906 leaves the room: Replaced by new connection [15:21:59] WPM01906 joins the room [15:22:25] Vint: problem with testing [15:22:45] at protocol, mapping label by label [15:22:56] at DNS level, ascii . separator [15:23:19] only as emerge out of punycode and into unicode, processing all label by label [15:23:34] issue is all within a label [15:23:40] issue of aleph.3com.com [15:23:47] is not label by label [15:24:03] tonyhansen leaves the room [15:24:19] xiaodong.lee joins the room [15:24:26] Alex McMahon joins the room [15:24:47] tonyhansen joins the room [15:24:53] problem right now that software doesn't know the string in question supposed to be a domain name [15:25:20] swshin92 leaves the room [15:25:24] Harald has problem [15:25:32] domain names tend to occur in text [15:25:49] and not going to see text processing software pieces that treat domain names inside the text specially [15:26:23] need to behave rationally when doms presented in text [15:26:33] Paul Hoffman [15:26:40] procedural problem [15:27:14] consensus that 3com is a problem, and not accepted, going to have to start over [15:27:46] Harald: [15:28:04] SNOWBALL602C9192 joins the room [15:28:17] current text just means 3com doesn't allow subsidiary RTL labels [15:28:18] tonyhansen leaves the room [15:28:20] xiaodong.lee leaves the room [15:28:23] SNOWBALL602C9192 leaves the room [15:28:28] disagreement on this point [15:28:38] Julian joins the room [15:29:06] refer to -01 draft [15:29:17] tonyhansen joins the room [15:29:23] Harald, fixed in 1.1 and not 6.1 [15:29:50] paul: so look 1 label before or after? [15:30:46] Harald: yes, but still we have the objection of inter-label comparisons [15:30:51] John Klensin: yes [15:31:24] tonyhansen leaves the room [15:31:29] also domain names in the running paragraph, then the heuristic is bits of texts on 2 sides of dots with no spaces [15:31:38] tonyhansen joins the room [15:31:46] xiaodong.lee joins the room [15:31:50] anything we do that starts using funny dots will cause some current things to stop working [15:32:00] and security people will be annoued [15:32:03] annoyed [15:32:08] also [15:33:03] difficulty with testing on current devices and current systems may cause problems [15:33:39] bert joins the room [15:33:40] Vint would like people to be crisp in comments [15:33:48] Stuart [missed] [15:34:12] have been attempting experiements [15:34:21] get surprising behavior [15:35:10] suggesting a presentation-layer anchor to set off the dots [15:35:29] Alirezah Saleh: also have the problem with "@" [15:35:50] if you use some substitution of dot, and start making things work that way [15:35:56] will be hard on RTL readers [15:36:08] because it will force whole domain name to be LTR [15:36:08] mohsen joins the room [15:36:33] Harald: [15:36:51] do you think it reasonable if you have RTL script, can't use numbers at end? [15:37:04] ALirezah: not a big issue compared to the problems [15:37:10] already adapted to IDNA2003 [15:37:42] Mark Davis at mic [15:37:43] tonyhansen leaves the room [15:37:44] tonyhansen joins the room [15:37:46] tonyhansen leaves the room [15:37:51] xiaodong.lee leaves the room [15:38:08] He has put something I can't read on a flipchart [15:38:22] (I mean, I can't see that far) [15:38:38] points: [15:38:46] It is a variant of the .3com.com problem [15:38:57] 1. Impossible to put constraint on people that all software has to recognize domain names [15:39:04] Many word processors would make characters cross the inter-label boundary [15:39:46] 2. if there are reasonably good rules that prevent a lot of confusion, take them even if they're not 100% [15:39:52] tonyhansen joins the room [15:39:55] 3. Need ot have inter-label constraint [15:40:28] xiaodong.lee joins the room [15:40:40] 1AB2.X.3CD4 displays as 1AB2.3.XCD4, where X is some arabic char [15:42:10] Little-endian is evil [15:42:37] [missed name] notion of which order you write the domain names in may be a mistake [15:43:23] It was Stuart CHeshire [15:43:26] 2008xyz as zyx8002 would be *too* odd [15:43:32] so maybe you just insist that the domains always be in the same order, and no context differences [15:43:52] This is just a rendering problem. ON the wire, there is zero amibguity. [15:44:03] And rendering issues are typically out-of-scope for the IETF [15:44:08] tonyhansen leaves the room [15:44:13] xiaodong.lee leaves the room [15:44:25] tonyhansen joins the room [15:45:55] Klensin: part of the problem here is that the editors that are doing the moving things about are doing so because they're designed for text input and do things automatically. So this is a corner case [15:46:19] what if you had a different rule: [15:47:01] put market around domain with any RTL character in it [15:47:05] doesn't work either [15:47:47] Vint: nobody at mic, not enough conclusion [15:48:32] asks Harald, if were to adopt three rules, is it correct that (e.g.) "3com" would be invalidated? I.e. already-registered domain names that will become invalid? Or only on context of RTL? [15:48:43] Harald: only in context [15:48:48] Suz joins the room [15:48:58] Ted Hardie: upshot: must compare across labels [15:49:06] Julian leaves the room [15:50:52] Harald: this is ok for a rule, though, because there's no way of enforcing anyway [15:51:27] Vint: only in the case with mixed RTL and LTR labels that there is a problem [15:51:36] Thomas Narten: [15:51:47] this is going to be registry guidelines, no protocol? [15:51:56] Vint: no, it's part of the specification [15:52:00] it's not just advice [15:52:05] Pete Resnick: [15:52:16] if you did inter-label checking, it'd satisfy all rules [15:52:30] but if you got rid of 3d rule, would solve the problem without inter-label check [15:53:04] pk joins the room [15:53:12] Harald: IRIs are defined as being in RTL context [15:54:07] Klensin: "if you do something that violates the standard, bad things will happen to you" is ok. But the problem is if different people do individually reasonable things, someone else will have bad results [15:56:33] tonyhansen leaves the room [15:56:40] WPM01906 leaves the room [15:56:48] Vint: what to do to proposed rules to elim concern? [15:57:00] xiaodong.lee joins the room [15:57:02] Klensin: drop inter label [15:57:03] tonyhansen joins the room [15:57:09] and try to warn people [15:57:19] "odd things may happen" [15:58:10] koji leaves the room [15:58:27] Philip Hallam Baker [15:58:36] koji joins the room [15:58:36] chasing heuristics [15:59:10] existing editors are not interesting, because they'll change [15:59:27] inter-label rules are unsustainable [16:00:01] do not have LTR in DNS [16:00:09] just sequence of bits [16:01:05] Vint: what is conclusion? [16:02:53] resnick joins the room [16:02:55] PHB: inter-label checking not going to work [16:02:59] Vint: [16:03:33] checking for opinion of room [16:03:52] 1. Anyone who feels strongly about removing inter-label comparisons? [16:03:56] a couple [16:04:00] 2. How many don't care? [16:04:05] no [16:04:29] 3. eliminate inter-label checking? [16:04:42] hearing no objections (advantage to being deaf guy), rule gone [16:04:54] Harald [16:05:15] will add a section saying "in following cases, things will go haywire" [16:05:35] Carey Karp: how certain do we need to be? [16:08:24] Patrik Faltstrom about the tables document [16:08:54] Email today, from Korea, with a large number of characters that are DISALLOWED [16:09:11] please look at that email [16:09:27] Lisa Dusseault joins the room [16:09:41] changes to the tables document hasn't been great for a long time [16:09:57] prepared to introduce proposed changes to -03 document [16:10:08] should make this change or not? [16:10:11] Mark Davis: [16:10:16] change is on road to perdition [16:10:34] character by character through Unicode, we'll never get done [16:10:46] resnick leaves the room [16:10:48] some characters are already restricted because would be disallowed [16:10:53] by NFC [16:11:04] dangerous to look at each character [16:11:13] Patrik: reason prepared to do it [16:11:22] is because it's a change which is a rule change [16:11:33] not a character-by-character analysis [16:11:52] Mark Davis: [16:12:10] on IDNA list, was proposal to eliminate some 30 scripts, and was rejected [16:12:27] on the grounds that it was too picky [16:12:59] Vint: source of the request comes from group of native language speakers [16:13:18] if it is coming from such a group, hate to discard out of hand [16:13:27] Mark Davis: can of worms? [16:13:35] Patrik: let's have a look & discuss tomorrow [16:13:54] John Klensin: [16:14:22] disallowing characters that are chars in which a lang is or was written [16:14:30] is a big deal in the original principles [16:14:45] the question is, "Can these be harmful?" [16:15:26] if the answer is no, and these are language characters, then these go in the registry restriction not protocol issue [16:15:35] Stephane Bortzmeyer [16:15:45] similar document re: Arabic [16:15:47] recently [16:15:51] it's the same problem [16:15:51] resnick joins the room [16:16:24] Patrik: difference is where in the process [16:17:15] John Klensin [16:17:38] in the arabic script case, there's been a finding of harmfulness in the case of certain characters [16:17:48] avri leaves the room [16:17:53] tonyhansen leaves the room [16:17:56] xiaodong.lee leaves the room [16:18:03] so in one sense it's further ahead, and in another sense behind [16:18:09] tonyhansen joins the room [16:18:20] Please read the document, especially if you read Korean [16:18:28] Second question [16:18:41] one objection has been IANA considerations not clear [16:18:41] Randall Gellens leaves the room [16:18:56] this because the document does 2 things [16:19:03] healthyao leaves the room [16:19:06] bert leaves the room: Computer went to sleep [16:19:07] main (normative) docs and (non-normative) table [16:19:17] request to IANA to keep non-normative part updates [16:19:19] updated [16:19:45] does not ask for IANA to keep track of rules, &c. [16:19:50] that's the source of the confusion [16:20:04] have a new approach [16:20:21] keeping track of normative and non-normative part [16:20:30] please think about this until tomorrow [16:20:46] current plan [16:20:58] 1. non-normative table of code points updated by IANA and kept [16:21:09] do this with appointed expert [16:21:30] 2. Normative part, and update to it, need IESG action [16:21:37] see comments about this on list [16:21:44] Alex McMahon leaves the room [16:21:57] Lisa Dusseault leaves the room [16:22:12] resnick leaves the room [16:22:12] narooter leaves the room: Computer went to sleep [16:22:16] that brings the meeting to the end [16:22:18] adjourned [16:22:22] Andrew Sullivan leaves the room [16:22:31] cary leaves the room [16:22:43] hta leaves the room [16:22:48] frank leaves the room [16:22:49] shinta leaves the room: Computer went to sleep [16:22:58] PHB leaves the room [16:24:22] pk leaves the room: Computer went to sleep [16:24:32] koji leaves the room [16:25:50] Barry Leiba leaves the room [16:31:09] Barry Leiba joins the room [16:32:01] Barry Leiba leaves the room [16:34:36] Barry Leiba joins the room [16:35:01] mohsen leaves the room: Computer went to sleep [16:36:07] yone leaves the room [16:36:16] lynch leaves the room [16:36:31] elwynd leaves the room: Replaced by new connection [16:36:35] elwynd joins the room [16:39:57] edmon leaves the room [16:40:43] ogud: Olafur Gudmundsson leaves the room [16:40:46] tonyhansen leaves the room [16:40:50] Barry Leiba leaves the room [16:41:02] tonyhansen joins the room [16:44:22] bortzmeyer leaves the room [16:45:14] hta joins the room [16:46:07] xiaodong.lee joins the room [16:48:01] xiaodong.lee leaves the room [16:48:42] Lisa Dusseault joins the room [16:52:18] Suz leaves the room [16:53:48] elwynd leaves the room: Replaced by new connection [16:53:51] elwynd joins the room [16:55:06] Lisa Dusseault leaves the room [16:59:39] xiaodong.lee joins the room [16:59:55] xiaodong.lee leaves the room [17:00:08] tonyhansen leaves the room [17:04:42] elwynd leaves the room: Replaced by new connection [17:04:45] elwynd joins the room [17:15:21] bert joins the room [17:19:50] bert leaves the room: Replaced by new connection [17:20:23] mohsen joins the room [17:20:32] elwynd leaves the room: Replaced by new connection [17:20:34] elwynd joins the room [17:34:09] Jaeyoun Kim (Peter) leaves the room [17:53:27] xiaodong.lee joins the room [17:53:36] xiaodong.lee leaves the room [18:09:40] hta leaves the room [18:46:17] elwynd leaves the room [19:28:22] fujiwara leaves the room [20:51:08] avri joins the room [21:04:53] avri leaves the room [22:21:08] PHB joins the room [22:24:48] mohsen leaves the room [22:43:43] PHB leaves the room