[05:48:35] rgonzalez joins the room [05:49:06] rgonzalez leaves the room [06:02:54] hardaker joins the room [07:28:24] ogud joins the room [08:12:43] Doug Barton joins the room [08:14:10] Carsten Strotmann joins the room [08:16:50] sandoche joins the room [08:18:50] Chris Griffiths joins the room [08:19:51] Patrik Wallström joins the room [08:20:40] niall joins the room [08:21:50] mwm joins the room [08:22:32] mwm leaves the room [08:22:57] mwm joins the room [08:23:07] Zhen Tsao joins the room [08:23:45] Exodus joins the room [08:24:02] mwm leaves the room [08:24:12] Exodus is now known as Dürst [08:24:14] fujiwara joins the room [08:24:19] Frederico Neves joins the room [08:24:20] Jelte joins the room [08:24:30] Hi [08:24:30] danny leaves the room [08:24:33] Larissa Shapiro joins the room [08:24:42] I'll be your jabber scribe for the day! [08:24:56] elmi4711 joins the room [08:24:59] stainlesskim joins the room [08:25:10] rhe@wiki.ja.net joins the room [08:25:11] sconte joins the room [08:25:43] This will be session two of the dnsext working group, for those that weren't aware [08:25:46] ogud has set the subject to: IETF-78 DNSEXT second seesion [08:26:26] What if we WERE aware, what will it be for us? :) [08:28:15] ogud has set the subject to: IETF-78 DNSEXT second seesion "Aliasing Requirements" [08:28:36] pure hell ;) [08:29:14] [08:29:20] mwm joins the room [08:29:32] mwm leaves the room [08:29:37] Andrew sullivan is starting the meeting [08:29:45] mm joins the room [08:29:50] goal of this meeting is a little different than session 1 [08:30:06] SUN Guonian joins the room [08:30:19] try to ensure that we are on the right track with aliasing work [08:30:24] hallam@gmail.com joins the room [08:30:25] note well on screen [08:30:56] cary joins the room [08:31:02] mm leaves the room [08:31:06] skudou joins the room [08:31:41] plan: 1. what we have 2. what proposals 3. problem description 4. description of impossibilities 5. discussion [08:32:15] 'what we have today' slide; cname/dname [08:32:28] Jacky Yao (Health Yao) joins the room [08:32:31] some people think we already have what we are discussing here [08:32:33] Is there a convention for requesting proxy microphone input? [08:32:52] Jim Galvin joins the room [08:32:57] mtcarrasco joins the room [08:32:59] ... Or NS record [08:33:04] howard_eland joins the room [08:33:17] niall: i haven't seen a volunteer yet, but add 'for mic' or something to your statement/questions [08:33:26] joseph.yee joins the room [08:33:33] i could proxy but i'm also scribing :) [08:33:36] dankjewel jelte [08:33:49] J.D. Falk joins the room [08:33:50] I will repeat things for the mic [08:33:58] things proposed 1: cname+dname [08:34:09] jinmei joins the room [08:34:53] needs signalling/might have side effects [08:34:53] Slide: Things Proposed 2: BNAME [08:35:04] new type, no side effects from existing records [08:35:18] but does need backwards compatibilty (trhough cname) [08:35:33] unknown (ie. BIG) deployment barrier [08:35:57] woolf joins the room [08:35:58] Sllide: Things proposed 3: Shadow zones [08:36:12] basically a trick for in-zone provisioning [08:36:52] needs modification of provisioning systems, but the rest of the dns doesn't really need to know about it [08:36:55] Shadow zone also has '2nd-class' effect: problem with sub-delegation [08:37:05] so from a deployment point of view its much simpler [08:37:30] (i personally have more issues with shadow, but we're just enumerating here ;) ) [08:37:32] sandoche leaves the room [08:37:41] Patrik Wallström leaves the room [08:37:47] Patrik Wallström joins the room [08:38:22] shadow might have nasty side effects for things like dynamic update [08:38:34] Slide: What these would solve [08:38:55] I really need to find the time to write an I-D for my CLONE idea [08:39:05] getting 'aliasing' functionality (which people think we already do) [08:39:16] g.e.montenegro joins the room [08:39:21] some management help for operators of many zones [08:39:57] Alireza Saleh joins the room [08:40:22] these things could be solved by provisioning systems, but that's not feasible for some people [08:40:41] for instance that doesn't work across administrative boundaries [08:40:59] hardaker leaves the room [08:40:59] Slide: what will not get solved [08:41:27] still lots of host names [08:41:53] no automatic char-to-char maps [08:42:59] idna2008 does not allow certain kind of case operations; people want all uppercase and all lowercase to match (like in ascii), but that doesn't work and it's unknown if there is a way thaat *we* can solve that [08:43:30] no automatic solution to any language issues [08:43:57] c+d and b still have 'first class citizen' problem; you still need canonical name for certain operations [08:43:58] hta joins the room [08:44:05] same goes for shadow [08:44:27] No that is an implementation choice [08:44:37] there's also no suggestion for how to prove what aliases a host is known by [08:44:43] Implemenation of shaddow is a vendor option, there is no standard [08:44:58] I could write a DNS server and make shadow work any way I choose [08:45:08] Slide: Discussion [08:45:42] Rule: We are collecting use-cases, not ruling work out [08:46:01] We might end up with 'cannot be done in DNS, need something else' [08:46:03] jpc@jabber joins the room [08:46:43] sandoche joins the room [08:46:56] i was thinking if 'don't do anything' falls under that header [08:46:57] Olaf needs to be nearer to mic! [08:47:08] (olaf needs his own mic ;) ) [08:47:16] yone joins the room [08:47:21] i want another entry on the list of use cases [08:47:30] EBADSOUND [08:48:22] ywang830 joins the room [08:49:02] in this case what's needed is a pair of names that have symmetric relationships [08:49:09] Requirement for symmetry is not proven!! [08:49:25] occur at any level and any combination of levels [08:49:41] wouter joins the room [08:49:45] Benno Overeinder joins the room [08:49:55] would require both provisioning and cnames pointing both ways [08:50:20] so this might be for the not possible liist, but there *is* demand for it [08:50:32] are remote folks having any trouble hearing the remote mic speakers? [08:50:41] Jim Galvin: YES [08:50:43] yes [08:50:47] May be single characters or group of characters. Second is more general, includes first. [08:50:51] could be louder, yes [08:51:02] did this help ? [08:51:04] indeed could be louder, yes [08:51:13] helped a tiny bit [08:51:14] If JK is standing at a mic, I doubt it's actually on [08:51:23] he sounds amplified here in the room [08:51:26] I have to keep adjusting volume whenever source chsnges between room and top table [08:51:26] Bernie joins the room [08:51:30] it sounds like he is being picked up ambient from the other mic(s) [08:51:54] mixer desk team (if any) could give better attention to balance, please [08:51:57] want to step back a little bit [08:52:06] antoin's mic sounds fine [08:52:08] very clear and lound now () [08:52:11] antoin using different remote mic [08:52:12] to the design of DNS itself; it's a tree with branches and nodes [08:52:28] ask everybody to use that mic! [08:52:48] would be very worried if we were going to design something that had aliases for nodes; that would make it a web, not a tree [08:53:01] where you cannot predict the path anymore [08:53:04] Note that Olafur just asked people to use the working mic [08:53:10] tlr joins the room [08:53:21] There is a lot between Tree and Web. I guess mostly, what we are trying to do is a DAG (directed acyclic graph) [08:53:22] (interesting that both mics work fine in the room) [08:53:24] perhaps a bramble [08:53:32] JcK joins the room [08:53:59] Antoin joins the room [08:54:02] Chris Griffiths leaves the room [08:54:10] technical clarification: i believe BNAME needs signaling, not C+DNAME [08:54:17] not sure, think they both do [08:54:30] mesmerizing screensaver on screen now [08:54:48] ok, i think you are right [08:54:52] question for john klensin, what is the use case for label symmetry, and where can more information about it be found? [08:55:04] Thanks, , for also jabber-scribing the visual aspects of the meeting :-) [08:55:27] for mic: DNS has already considered a similar problem to address wildcards, we should look for the same solution, re the authoritative name server synthesizes records on the fly and the only time a RR needs to appear on the wire is to support zone transfer or DNSSEC. Ie we need to do the shadow approach. Attempting to change the legacy DNS is hard, the same fixes that are required to support shadow will be required to support legacy zones. The only place that we have any real maneuver room is in the area of DNSSEC where there are few servers consuming records and those that exist are fluid. Would commercial support for such services remove the need for this work item? [08:55:40] the most common use-case is for the web [08:55:56] Benno Overeinder leaves the room [08:55:58] ppl want to type address and magically get to the right place [08:56:00] koji joins the room [08:56:01] question for john klensin: +1 [08:56:02] Benno Overeinder joins the room [08:56:09] jinjian joins the room [08:56:25] label symmetry: in politics, there is a lot of need for "officially being exactly equal". [08:57:08] Hallam your comment is in the queue for the mic [08:57:31] there a classes of badness that can happen where its easier to have a record at the top saying 'these are the same' than to have everything configured correctly for multiples at a lower level [08:58:05] I'm aware of a certain level of political drama about these issues, especially at the IDN TLD level, but the question for US is, what are the technical requirements? [08:58:16] Patrik Wallström leaves the room [08:58:22] Patrik Wallström joins the room [08:58:32] so for one of the most occuring usecases (web), aliases are helpful; it's neither necessary nor sufficient, but it makes things a lot easier [08:58:44] tlr leaves the room [08:58:49] and point to john: i have no idea what 'equal' means [08:58:53] ogud, thanks [08:59:11] Chris Griffiths joins the room [08:59:18] olafur reading comments from jabber now [08:59:45] peter koch joins the room [08:59:50] mccreary joins the room [09:00:50] on phil's comment: everything is true, but there are assumptions on apps here and i'm not sure about those [09:01:11] Ning KONG joins the room [09:01:27] comment on haralds comment: i think that if we optimize for web, we make a systems design mistake [09:02:03] wouter leaves the room [09:02:12] about use cases: lots of weird encoding idiosyncrasies in unicode [09:02:44] strange things happened when combining scripts [09:02:54] e.g. zero-width joiners [09:02:57] wouter joins the room [09:03:13] joao joins the room [09:03:36] Jaap Akkerhuis joins the room [09:04:12] if you have two users using the same script, if they type the same name, those should match. but that does not necessarily happen [09:04:43] same even goes for spelling [09:05:53] normal humans see things as the same that we computer people don't [09:06:05] I can't spell "Ameericn".... [09:06:07] ed8f1d0b4b9f174c joins the room [09:06:51] And how does that lead specifically to a symmetry requirement? [09:07:13] we need more of those examples [09:07:23] Isn't a 'contraction mapping' to a subjacent canonical form good enough? [09:07:49] if you have examples (and not just hypothetical cases), please talk to us [09:08:07] ywang830 leaves the room [09:08:18] niall: what's a subjacent canonical form? [09:08:28] and is it language dependent? [09:08:34] ywang830 joins the room [09:08:40] underlying 1st-class name [09:08:42] i think equivalency of labels does not have to mean it has to be equivalent in dns, might be impossible, and it could work, but we might need something else [09:08:54] are you suggesting a layer on top of dns? [09:08:57] some examples: Half-width versus full-width: DNS vs. ＤＮＳ, ヒラガナ vs. ﾋﾗｶﾞﾅ [09:09:26] a little bit, you kind of have this already (for instance in apache) [09:10:14] First (rough) draft of my CLONE idea: http://ops.ietf.org/lists/namedroppers/namedroppers.2009/msg03106.html [09:10:28] (discussion of our reach; can we redefine email rules?) [09:11:10] for MX it's problem of the target, not the owner, and aliasing is owner, so just need a little care [09:11:59] people way down in an aliased tree might not even know that [09:12:13] (that they are in an aliased tree) [09:12:15] Arabic vs. Eastern-Arabic digits: ٠١٢٣٤٦٦٧٨٩ vs. ۱۲۳۴۵۶۷۸۹ (three of them look different, the others look the same, just in case you don't have a font configured) [09:12:32] Bernie leaves the room [09:12:52] Patrik Wallström leaves the room [09:12:58] Having an alias at the apex AND below the cut is bad. Having the apex alias override the "real" data is very bad. [09:12:59] Bernie joins the room [09:13:01] Patrik Wallström joins the room [09:13:42] please put the word "mic" in the comments you want repeated at the microphone [09:13:45] Don't we need to distinguish between 'zone operators' and 'users'? [09:13:45] i don't see this as a problem, enough places to break stuff already [09:13:59] Doug, is your "clone" idea functionally different from Paul Vixie's "shadow"? [09:14:04] yes [09:14:08] very different [09:14:32] (very long sentence) [09:14:48] Dougb: The current problem statement draft (identical-names) surveys proposals that exist as I-Ds, since it's not clear to me how to evaluate ideas that haven't been put in at least that much detail against the requirements. Do you plan to turn your idea into a proposal in an i-d? [09:14:52] Doug please write up RSN so we can consider it in our big selection process in early Sept. [09:15:09] conclusion at first idna about some of these problems was that current gen DNS cannot solve them [09:15:11] Michael Graff joins the room [09:15:21] woolf: no harm, no foul :) [09:15:30] need dns-ng, or level on top of dns [09:15:32] Has even one person who is not a DNS expert stood up and talked? [09:15:50] I tried [09:15:55] I know. [09:16:04] mccreary leaves the room [09:16:14] But no offense, you're at least 50% of a DNS expert in this context :) [09:16:22] ogud : yes, I need to do that, honestly I was hoping that $SOMEONE else would come up with the same/better idea [09:16:32] mccreary joins the room [09:16:36] My primary hat is not "DNS expert". I just play one at IETFs. [09:16:58] Maybe we should remove case-differences from DNS, and make it 8bit clean. [09:16:59] I'm terribly time-challenged atm between trying to find work, and doing the work I have :-/ [09:17:03] there is a point at which we cannot abuse dns any more to solve new usecases [09:17:08] hta: there's a roomful of people with similar hats today [09:17:28] DNS has a major complicating factor that other systems don't have in that there are authority delegations. It is going to be very difficult to make something work reliably if the aliasing is happening at the TLD stage. [09:17:46] we looked at don't do it in dns, we got no traction at all [09:17:56] way too much attachment to the DNS [09:18:05] awwww, people love us [09:18:26] Various Yeh variants in Arabic (all in various connecting positions): ىىى ى ; ييي ي; ئئئ ئ;ٸٸٸ ٸ; ییی ی; ۍۍۍ ۍ; ۍۍۍ ۍ; ێێێ ێ; [09:19:06] Patrik Wallström leaves the room [09:19:15] Patrik Wallström joins the room [09:19:32] Jabber is great; RFCs are terrible for non-ASCII [09:19:36] I still do not see how this is a DNS topic... [09:19:57] Durst: +1 [09:20:09] if we say 'x aliases y', what users hear is 'x and y are completely interchangeable, everywhere, and if i have x i can get y' [09:20:50] pure tree are a nice way of searching knowledge, but a terrible way to organize it [09:20:55] s/tree/trees/ [09:22:39] followup to paul's comment: we may be getting closer to the point where the perception is that we do need big changes [09:22:50] might still get no traction at all, but perhaps we need to try again [09:23:01] i will not be coherent here [09:23:03] RFCs are indeed terrible for non-ascii, which should embarrass the IETF a bit more than it does (for all that I understand how it happened). But I appreciate them. [09:23:15] what are the timescales that we need solutions here? [09:23:26] roessler joins the room [09:23:36] can we have a kludge or two, or can we take the time to go back all the way to the drawing board? [09:24:17] roessler leaves the room [09:24:48] wilmer joins the room [09:24:59] timescale: we heard 'yesterday' 'not at all' and everything in between [09:25:31] problem is not what an alias is, or what names are equal [09:25:50] sorry missed the last part of that [09:26:11] roessler joins the room [09:26:28] = [09:26:32] Zhen Tsao leaves the room [09:26:33] ah, thanks [09:26:53] the question is not if we allow equivalent zones, we cannot disallow them, they will exist [09:27:04] the question is do we need standards work for them [09:27:24] "the internet is life" [09:27:31] the structure of the dns database is a tree, and we cannot change that [09:27:45] Patrik Wallström leaves the room [09:27:51] with equivalence we will always have first- and second-class citizens [09:27:54] Patrik Wallström joins the room [09:28:39] stainlesskim leaves the room [09:28:42] I'm fine if the problem statement document ends up saying "there's no clear situation in which a problem we know about matches a solution we've come up with." If there's work to do here, I'm fine with doing it. If there isn't, let's stop changing the DNS. [09:29:05] we might be able to store the information about what is desired to be equivalent, and perhaps carry that information through the DNS, but we cannot get dns to do that automatically in every case [09:29:09] Some Kanji examples: 国, 國; 鴎, 鷗 [09:29:28] ® and (R) [09:29:34] we simply need canonical forms, for caching, to prevent state explosions, etc. [09:29:36] © and (C) [09:29:40] +1 canonical form [09:30:12] don't expect perfect equivalence, please accept need for canonical form, and please accept the need that we need to carry information about equivalence through the system [09:30:19] g.e.montenegro leaves the room [09:30:29] canonical may be different for the same script but different languages [09:30:32] weiler joins the room [09:30:36] i have some questions [09:30:51] do you have problems that the proposals here would help you with? [09:30:53] More Kanji examles (numbers): 一, 壱; 二, 弐; [09:30:53] rhe@wiki.ja.net leaves the room [09:31:02] I agree with alfred's comments, and hta's [09:31:03] me: Yes [09:31:05] is anyone sure that these proposal do not help you? [09:31:08] (i.e. would help) [09:31:09] And the canonical form is "Hönes", "Hones", or "Hoenes"??? [09:31:15] nobody raised hand on first question [09:31:16] hta leaves the room [09:31:18] a few did on second one [09:31:25] d'oh [09:31:41] third question: need more information? [09:31:43] hta joins the room [09:31:45] about same number as for 2 [09:31:55] rhe@wiki.ja.net joins the room [09:32:00] Harald, stop hacking people :) [09:32:00] timeline to look at this is the next few weeks [09:32:00] stainlesskim joins the room [09:32:10] rhe@wiki.ja.net leaves the room [09:32:10] PLEASE POST YOUR USE-CASES AND PROBLEMS [09:32:17] sconte leaves the room [09:32:24] JcK: canonical form has to be by administrative 'fiat' [09:32:25] wouter leaves the room [09:32:25] Antoin leaves the room [09:32:27] we can't do anything if we don't know what we need to solve :) [09:32:32] Benno Overeinder leaves the room [09:32:33] Unless there is an intermediate translator/matcher, beyond a certain point, "canonical form" is very nearly equivalent to "everyone should just learn to speak MY language and use MY spellings" [09:32:39] koji leaves the room [09:32:43] problem: email. next question? heh [09:32:48] joao leaves the room [09:32:57] and would also like better answers to olaf's timing question [09:33:02] skudou leaves the room [09:33:02] jinmei leaves the room [09:33:03] weiler leaves the room [09:33:04] Michael Graff leaves the room [09:33:05] Carsten Strotmann leaves the room [09:33:08] i think the meeting closed [09:33:13] yone leaves the room [09:33:14] Patrik Wallström leaves the room [09:33:14] Define what will be the canonical form. Probably "Hönes", where "ö" is defined as some U+ [09:33:16] i hope my jabber scribe services have been of value [09:33:18] yep, meeting has ended [09:33:22] mtcarrasco leaves the room [09:33:22] yes, thanks [09:33:24] Yes, if you have a perspective on what you'd like to see working, please please please give us specific, concrete examples ... we cannot know if we have solved a problem if we don't know what the problem is [09:33:27] hta leaves the room [09:33:29] you are welcome [09:33:33] stainlesskim leaves the room [09:33:37] hallam@gmail.com leaves the room [09:33:41] Jelte: excellent work, and thanks :) [09:33:45] woolf leaves the room [09:33:51] Larissa Shapiro leaves the room [09:33:54] jpc@jabber leaves the room [09:33:55] thanks for the thanks :) [09:33:55] Alireza Saleh leaves the room [09:34:08] J.D. Falk leaves the room [09:34:12] Jim Galvin leaves the room [09:34:15] cary leaves the room [09:34:18] howard_eland leaves the room [09:34:30] joseph.yee leaves the room [09:34:31] The admin fiat isn't the problem --I agree with that. The problems are getting everyone to follow that fiat (it is going to work for some zones and not others and, unless it is universal, one will have different conventions within the same FQDN) and getting all lusers who want to look up names to understand the particular fiats for the zones in which those names appear [09:34:32] Jaap Akkerhuis leaves the room [09:34:56] Alireza Saleh joins the room [09:34:58] Alireza Saleh leaves the room [09:35:02] Frederico Neves leaves the room [09:35:02] ogud leaves the room [09:35:55] mccreary leaves the room [09:36:50] hah .... I knew that mic was off :) [09:37:13] Ning KONG leaves the room [09:37:47] JcK: choice of canonical form is indeed arbitrary and restrictive. However, it is a per-case choice, imposed by the responsible registry (or perhaps, delegation parent). Greece, China, US, Ireland (my country) can each have a local policy at the ccTLD level. gTLDs are more troublesome, of course. The root is probably less so, as politics takes the problem away from the technical realm. 8-) [09:38:32] Bernie leaves the room [09:38:37] Larissa Shapiro joins the room [09:39:02] niall: agree about the root. It's politics, but also the fact that stuff can be discussed on a one-by-one basis [09:39:12] Larissa Shapiro leaves the room [09:39:27] elmi4711 leaves the room: I'm happy Miranda IM user. Get it at http://miranda-im.org/. [09:39:42] Dürst: more on mailing list? [09:40:38] A typical example is that Russia might have wanted .ру (reading 'ru' in Cyrillic), but that is too close to .py (Paraguay) [09:42:10] Benno Overeinder joins the room [09:42:23] Benno Overeinder leaves the room [09:43:07] Chris Griffiths leaves the room [09:43:36] niall leaves the room: Logged out [09:43:54] so they went for .рф (rf, russian federation), which is easy to distinguish from everything else there is at the top level [09:44:23] I'm not on the mailing list, and don't want to subscribe because I don't have the time to read it. [09:44:23] niall joins the room [09:44:33] niall leaves the room [09:44:35] Please feel free to send what I wrote to the mailing list. [09:45:48] roessler leaves the room [09:45:50] jinjian leaves the room [09:45:51] Jaap Akkerhuis joins the room [09:46:20] Jaap Akkerhuis: You're a bit late my friend :) [09:46:21] JcK leaves the room [09:46:32] peter koch leaves the room [09:48:32] Jaap Akkerhuis leaves the room [09:50:02] sandoche leaves the room [09:50:32] Jelte leaves the room [09:50:40] Doug Barton leaves the room [09:52:32] Jacky Yao (Health Yao) leaves the room [09:58:16] ogud joins the room [10:01:17] hardaker joins the room [10:02:38] ed8f1d0b4b9f174c leaves the room [10:02:40] ed8f1d0b4b9f174c joins the room [10:06:46] ed8f1d0b4b9f174c leaves the room [10:07:40] ywang830 leaves the room [10:15:10] jinmei joins the room [10:20:14] jinmei leaves the room [10:36:26] mccreary joins the room [10:36:30] SUN Guonian leaves the room [10:37:11] ywang830 joins the room [10:40:29] Patrik Wallström joins the room [10:42:20] Patrik Wallström leaves the room [10:42:21] Patrik Wallström joins the room [10:42:32] mccreary leaves the room [10:49:26] Bernie joins the room [10:52:09] Bernie leaves the room [11:03:12] Jacky Yao (Health Yao) joins the room [11:03:38] Jacky Yao (Health Yao) leaves the room [11:06:23] fujiwara leaves the room [11:06:28] J.D. Falk joins the room [11:06:36] J.D. Falk leaves the room [11:07:09] ogud leaves the room [11:12:03] Michael Graff joins the room [11:13:09] Dürst leaves the room [11:16:16] Michael Graff leaves the room [11:43:40] ywang830 leaves the room [12:50:51] Patrik Wallström leaves the room [13:05:41] ogud joins the room [13:21:06] Patrik Wallström joins the room [13:28:43] ogud leaves the room [13:47:23] Patrik Wallström leaves the room [13:57:06] hardaker leaves the room [14:15:24] Patrik Wallström joins the room [14:31:20] hardaker joins the room [16:13:57] Patrik Wallström leaves the room [16:43:39] hardaker leaves the room [17:26:26] hardaker joins the room [18:27:21] hardaker leaves the room [21:20:49] hardaker joins the room [23:02:56] hardaker leaves the room