[05:00] <sabdfl> hey all
[05:00] <jdub> morning
[05:00] <sabdfl> ish
[05:01] <haggai> uh tigger
[05:01] <mdz> there rae no new agenda items
[05:01] <mdz> are
[05:01] <sabdfl> the wonderful thing about tiggers
[05:02] <mdz> is tiggers are wonderful things
[05:02] <sabdfl> is tigger are wonderful tings
[05:02] <jdub> aww so cute ;)
[05:02] <sabdfl> ... was trying to think of something rhyming with kubuntu :-)
[05:02] <jdub> sabdfl: ... ubuntu?
[05:02] <sabdfl> thatwuntoo
[05:02] <haggai> tiggerunto?
[05:03] <sabdfl> should we begin
[05:03] <sabdfl> nothing on the agenda
[05:03] <sabdfl> mdz, any other business?
[05:03] <jdub> there are two items on the agenda
[05:03] <jdub> - libnns-mdns
[05:03] <jdub> - language packs
[05:04] <jdub> (libnss-mdns)
[05:04] <sabdfl> ah right
[05:04] <mdz> sabdfl: none from me
[05:04] <sabdfl> that's howl right?
[05:04] <jdub> no
[05:04] <sabdfl> mdns?
[05:04] <mdz> language packs were discussed at the previous tech board meeting
[05:04] <mdz> that one should have been moved down to old business
[05:04] <jdub> mdns == multicast dns resolution (not part of howl, but related to zeroconf support)
[05:04] <mdz> there is no bullet point next to mdns
[05:04] <mdz> and therefore it is clearly invalid
[05:04] <jdub> haha
[05:05] <mdz> so the matter with mdns is that it wants to listen
[05:05] <sabdfl> mdz: might be worth pointing folks at the notes from the languagepak bof at the conf
[05:05] <jdub> it doesn't.
[05:05] <jdub> mdnsresponder from howl does
[05:05] <jdub> libnss-mdns is just a resolver
[05:05] <mdz> http://www.ubuntulinux.org/wiki/LanguagePacksBOF
[05:05] <jdub> you punted it to tech board because you felt that nss modules were dangerous
[05:05] <mdz> so is the proposal just to add libnss-mdns to supported?
[05:06] <jdub> to desktop
[05:06] <mdz> or to add it to the default nsswitch configuration?
[05:06] <jdub> yeah, and the default configuration
[05:06] <jdub> we have the option between an nss module or a glibc patch. personally, i think the nss option is slightly saner (and easy to disable). :-)
[05:07] <jdub> i'm not really clear on your POV about nss module fear though
[05:07] <sabdfl> i've no preference
[05:07] <mdz> it's no scarier than a glibc patch
[05:07] <mdz> but glibc patches are, well, SCARY
[05:08] <mdz> what does it buy us in terms of functionality?
[05:08] <mdz> does it let applications discover mdns things?
[05:08] <lamont_r> Keybuk: 20 pushups.
[05:08] <sabdfl> is it worth having a mataro session on mdns / zeroconf in general?
[05:09] <jdub> mdz: no, it just resolves .local names
[05:09] <jdub> sabdfl: potentially, i punted it because it was pretty airy-fairy.
[05:09] <sabdfl> it seems zeroconf / rendesvous has big potential, but also brings risks
[05:09] <mdz> if it's vaporous, should we be adding it to the default code path for every application which uses DNS?
[05:10] <jdub> libnss-mdns -> vaporous?
[05:10] <sabdfl> apple is pushing rendesvous pretty hard, is anything happening in the windows world about it?
[05:10] <sabdfl> if it's going to be widely supported and standard, then we will need to find a sane way to integrate it
[05:10] <mdz> jdub: airy-fairy?
[05:10] <jdub> MS backed UPnP, but will probably stumble on zeroconf at some stage.
[05:11] <sabdfl> much like cups, smb browsing etc
[05:11] <jdub> mdz: oh, a single bof about "zeroconf" in general.
[05:11] <sabdfl> yes
[05:11] <jdub> sabdfl: the big question, wrt risk, is whether mdnsresponder listens by default
[05:11] <sabdfl> what about a single-check "turn it on" option?
[05:11] <haggai> if it's not really in wide use yet, I'd say turn it off for now
[05:11] <mdz> that's my preference
[05:11] <jdub> sabdfl: automatic answer is "no", but that makes it useless and non-zeroconf. :)
[05:11] <mdz> ("turn it on" option)
[05:11] <mdz> jdub: "oneconf"
[05:11] <jdub> given policy, me too.
[05:12] <jdub> uniconf! (.sf.net -> fear)
[05:12] <sabdfl> what about a way in the system in general for things to say "i need to do some conf, can we turn it on.... ok, done"
[05:12] <Keybuk> mobile-esque "Discoverable for 5 minutes" ?
[05:12] <pitti> sabdfl: we could also need this for cups browsing etc.
[05:12] <haggai> hmm.  How about turn it on, if we get a zeroconf network config?
[05:12] <sabdfl> this is really something that's useful while the user is at the console, right?
[05:12] <mdz> cups browsing is broken by design
[05:12] <mdz> it should use zeroconf
[05:12] <jdub> rh are doing some good user-level cups / hal / howl integration
[05:13] <sabdfl> mdz: thereby deferring the problem?
[05:13] <sabdfl> "metadata!"
[05:13] <mdz> it reverses the problem
[05:13] <mdz> the cups server would need mdns-responder, but not clients
[05:13] <pitti> mdz: upstream is not going to change it
[05:13] <jdub> "who?" rather than "me!"
[05:13] <sabdfl> right
[05:13] <sabdfl> so it's more like dns
[05:13] <jdub> pitti: but it's fairly likely that everyone else will be using those patches.
[05:13] <mdz> yes
[05:13] <jdub> sabdfl: it, um, is dns.
[05:14] <sabdfl> is howl == zeroconf == libnss-mdns (client side)?
[05:14] <jdub> they're all related
[05:15] <jdub> zeroconf covers autoconfiguration of local link IP addresses and names, as well as dns-based service discovery
[05:15] <jdub> libnss-mdns is the resolver side of ip autoconfiguration
[05:15] <Keybuk> so zeroconf is "where are my printers" ?
[05:15] <jdub> howl is service discovery (uni and multicast, and publishing)
[05:15] <jdub> we're not doing ip autoconfiguration yet, but we can defer that for network manager, etc.
[05:16] <jdub> i should make a wiki page to explain all of this :)
[05:16] <mdz> yes, you should
[05:16] <sabdfl> what's the relationship between libnss-mdns and dhcp?
[05:16] <jdub> sabdfl: nothing, really
[05:16] <sabdfl> ip autoconf?
[05:16] <mdz> libnss-mdns needs an IP to send from
[05:16] <jdub> ip autoconfiguraiton and dhcp are related in that they both dynamically assign ips
[05:16] <jdub> dhcp is server based, zcip (zeroconf ip) is local lan non routable, etc.
[05:17] <sabdfl> ok
[05:17] <sabdfl> so this would be useful if, say, three of us were sitting on a bus and just wanted to throw a wifi netowrk up together?
[05:17] <mdz> right
[05:17] <mdz> or if you were all at an Ubuntu conference and wanted to talk to each other
[05:17] <sabdfl> or on a plane to australia
[05:17] <jdub> sabdfl: yes
[05:17] <sabdfl> give it to me now!
[05:17] <jdub> sabdfl: like the ichat 'rendezvous' window
[05:18] <jdub> at a conference
[05:18] <mdz> I think the flight crew frown on ad-hoc wireless networks
[05:18] <jdub> flight crews always smile
[05:18] <sivang> *not
[05:18] <sivang> (while in air)
[05:18] <sabdfl> jdub: if you smile first
[05:18] <mdz> so the question is, should we integrate nss-mdns and start exercising it?
[05:18] <Keybuk> does libnss-mdns have an always-listening component?
[05:19] <jdub> Keybuk: no, it's an nss module only
[05:19] <mdz> is there anything planned for hoary which would actually use it?
[05:19] <jdub> just a resolver
[05:19] <mdz> (which would rely on it, I should sya)
[05:19] <mdz> say
[05:19] <Keybuk> so to discover things with it, you need a listener elsewhere on the network ?
[05:19] <jdub> mdz: "is there anything in hoary that relies on dns resolution?"
[05:19] <sabdfl> jdub: when you try to resolve, does it ping the network and listen for responses?
[05:19] <jdub> Keybuk: well, yeah - mdnsresponder automatically publishes the hostname
[05:20] <Keybuk> ah, so it's a dns responder -- and nothing to do with service discovery ?
[05:20] <jdub> sabdfl: i don't believe so, no.
[05:20] <sabdfl> what's the status of utopia / dbus stack that would let a user see a dialog giving the status of a howl discovery session happening at system level?
[05:20] <mdz> jdub: that is not the same thing and you know it :-P
[05:20] <jdub> Keybuk: yes, see above.
[05:20] <Keybuk> it's *literally* multicast dns
[05:20] <mdz> depending on where we put it in nsswitch.conf, it'll be queried in order with other name resolution services
[05:21] <jdub> sabdfl: i can't see why that would be desireable or useful
[05:21] <mdz> currently we just try /etc/hosts and then dns
[05:21] <pitti> sabdfl: right now it is not integrated in hal AFAIK
[05:21] <Keybuk> the nss equivalent of "Is there a Hugh Janus in the room?  Telephone call for a Hugh Janus"
[05:21] <sabdfl> can that be used to redirect traffic?
[05:21] <mdz> yes
[05:22] <mdz> if we add it to the end, and you try to go to www.goggle.com, I can answer and send you somewhere, e.g.
[05:22] <sabdfl> nice
[05:22] <sabdfl> ok
[05:22] <mdz> adding it to the start would be, as jeff would say, BONG
[05:22] <sabdfl> dns says "nope" then mdns says "sure, try this one"
[05:22] <mdz> exactly
[05:23] <mdz> every single failed DNS request would try mdns as a fallback
[05:23] <jdub> only for .local
[05:23] <mdz> eh?
[05:24] <jdub> it wouldn't go looking for www.google.com with libnss-mdns
[05:24] <jdub> well, it would, but mdns would say no very quickly
[05:24] <jdub> as in, "not .local? kthxbye"
[05:24] <Keybuk> wouldn't we need to configure things to search .local then?
[05:25] <mdz> jdub: where is the code which does that bit?
[05:25] <jdub> ah, well, that's a separate issue :)
[05:25] <Keybuk> and then wouldn't it resolve www.google.com.local ?
[05:26] <haggai> so it's the dns equivalent of M$'s netbios?  Flood your local network instead of having a sane DNS server?
[05:27] <mdz> it's somewhat less braindead in that it uses multicast rather than broadcast
[05:27] <crimsun> the difference being that at least mdns is explanable
[05:27] <mdz> it includes its own DNS implementation, cute
[05:27] <crimsun> not even MS engineers can really fathom out netbios
[05:28] <mdz> Keybuk: it seems to be irrelevant
[05:28] <mdz> Keybuk: because jdub is full of shit about it only resolving .local :-P
[05:28] <jdub> heh
[05:28] <sabdfl> so, we really have to choose between pioneering and being conservative
[05:28] <mdz> that's easily fixable, though
[05:28] <sabdfl> i don't see the big win in breaking down this particular wall
[05:28] <jdub> so i think lennart meant it did it in svn, not the released version
[05:29] <jdub> but yeah, not a huge issue
[05:29] <Keybuk> if we do $zeroconf, we should to it in one big strategy-led punch; rather than in drips
[05:29] <sabdfl> seems that ms, apple, others will battle out the usability issues, we can follow on this one
[05:29] <sabdfl> we can't pioneer on every front
[05:29] <jdub> we don't have to
[05:29] <jdub> supporting the standard is enough
[05:29] <jdub> as more stuff up the stack supports it, we win
[05:29] <haggai> you can do that without turning it on by default everywhere
[05:30] <sabdfl> "standard" is one thing, figuring out what normal sane regular behaviour is is quite different
[05:30] <mdz> add it to supported and write a ZeroConfHowto?
[05:30] <jdub> Keybuk: i don't think that's a problem
[05:30] <jdub> sabdfl: we know what that behaviour is
[05:30] <haggai> it would be nice to have a single knob that would turn on all the zeroconf bits
[05:30] <sabdfl> i don't think we do!
[05:30] <sabdfl> we don't know how it should integraqte with dns and that's pretty fundamental
[05:31] <jdub> 1) we should always have a local lan ip
[05:31] <jdub> 2) we should only mdns query for .local
[05:31] <jdub> 3) we shouldn't search on .local
[05:31] <Keybuk> why #1 ?
[05:31] <mdz> jdub: we already do #1 :-)
[05:31] <Keybuk> we do?
[05:31] <jdub> mdz: no we don't
[05:31] <mdz> "ip addr" sometime
[05:32] <sabdfl> m gut feel is it will take a while before there's clear "best practice"
[05:32] <jdub> mdz: that's inet6, not inet4
[05:32] <sabdfl> and if we forge ahead now,  we have to support upgrades and old behaviour etc
[05:32] <Keybuk> does libnss-mdns support ip6 networks?
[05:32] <jdub> we're really not forging ahead here
[05:32] <sabdfl> bluetooth, however, might be an easier point
[05:32] <jdub> windows does zcip, mac os x does zcip
[05:33] <sabdfl> do they do it in the same way?
[05:33] <sabdfl> always?
[05:33] <jdub> yes
[05:33] <sabdfl> so bring up a win box, and a ma, on the same network, and it interoperates?
[05:33] <jdub> (for #1)
[05:33] <mdz> http://files.zeroconf.org/draft-ietf-zeroconf-ipv4-linklocal.txt
[05:33] <mdz> "standard"
[05:33] <jdub> sabdfl: no, they both get zcips
[05:33] <jdub> well, you can ping each other
[05:34] <sabdfl> demonstrably?
[05:34] <jdub> that's the zeroconf ip part of the standard
[05:34] <jdub> as far as i've seen, yes
[05:34] <jdub> Keybuk: planned, but not fully implemented
[05:34] <mdz> zcip doesn't sound so scary
[05:34] <haggai> "Microsoft Windows 98 (and later) and Mac OS 8.5 (and later) already support this capability.  This document standardizes usage,"
[05:34] <haggai> so we're definately not leading
[05:34] <jdub> on zcip, no
[05:35] <jdub> the second point of resolving .local stuff is not exactly rocket science
[05:35] <mdz> BUGS
[05:35] <mdz>        zcip  currently assumes that the target interface is down and reconfig
[05:35] <mdz>        ures it. The IETF draft seems to suggest using multihoming,  which  we
[05:35] <mdz>        dont do yet.
[05:35] <mdz> that sounds problematic
[05:35] <jdub> os x already does it
[05:35] <jdub> zcip is one particular implementation
[05:35] <jdub> howl includes an implementation
[05:35] <Keybuk> so is libnss-mdns useful without a $zcip ?
[05:35] <jdub> Keybuk: yes
[05:35] <sabdfl> yes, this looks like mutual dhcp
[05:35] <Keybuk> if you get dhcp, chances are you get dns too
[05:36] <haggai> sabdfl: you say we can't forge ahead on all fronts.  When you look at cross-OS stuff we are forging ahead on very little..
[05:36] <mdz> Keybuk: mdns serves a different purpose
[05:36] <sabdfl> ok, have to head off to barcelona, enjoy
[05:36] <mdz> it lets me advertise something to you
[05:36] <mdz> rather than having the network admin add it to the DNS server
[05:37] <Keybuk> that's not what libnss-mdns looks up though, it just does name lookups ?
[05:37] <jdub> i think mdz's original concern,
[05:37] <jdub> new code in a very active codepath,
[05:37] <jdub> was probably the most relevant issue
[05:37] <jdub> Keybuk: yes
[05:38] <jdub> i'm also going to barcelona
[05:38] <Keybuk> but unless you have an IP, you can't do name lookups anyway
[05:38] <Keybuk> you need $fu to co-operate on IP and ensure "domain local" is the order of the day
[05:39] <jdub> Keybuk: "domain local" as in resolv.conf?
[05:39] <Keybuk> yeah
[05:39] <jdub> nah
[05:39] <Keybuk> or equiv
[05:40] <jdub> ok, i think i'm going
[05:41] <mdz> me too
[05:41] <mdz> we'll defer mdns stuff until something interesting comes along
[05:41] <mdz> any urgent business before we adjourn?
[05:41] <mdz> adjourned, bye
[05:42] <mdz> thanks, everyone
[05:42] <Keybuk> "<mdz> kthxbye"
[05:49] <Simira> http://www.simira.net/UbuntuWomen.html - My article about Ubuntu and women in open source
[06:17] <Simira> hi lulu
[06:39] <Treenaks> Simira: there's  a '\ ' in your name.. don't you mean a '/' ?
[06:39] <Simira> oh no, not again... yes, thanks, Treenaks
[06:39] <Simira> so, there
[06:40] <Treenaks> Simira: oh, and s/Mataro/Matar/ :P
[06:41] <Simira> Treenaks: that looked to me like an A with a ~ and a 3 over  :p
[06:41] <Treenaks> Simira: in which editor?
[06:42] <Treenaks> Simira: and are you using an UTF-8 locale?
[06:42] <Simira> in mIRC
[06:42] <Treenaks> mirc.. ah :)
[06:43] <Simira> :p
[06:43] <Treenaks> yes.. that's how an UTF-8 o-with-accent looks if you try to parse it as Latin1