[20:30:39] --- Seung Yi has joined
[20:44:31] --- yowada has joined
[20:45:22] --- dthaler has joined
[20:45:25] --- thomas.heide.clausen@gmail.com has joined
[20:46:08] --- bebemaster has joined
[20:46:23] * bebemaster has changed the subject to: Agenda Bashing
[20:46:59] --- monden has joined
[20:47:09] --- monden has left
[20:47:31] --- Yangwoo Ko has joined
[20:48:11] * bebemaster has changed the subject to: MANET Architecture presentation
[20:48:53] <bebemaster> Thomas Clausen presenting
[20:48:58] --- ks has joined
[20:49:25] --- monden has joined
[20:50:21] <bebemaster> Clausen draft produced by ian chakares was first autoconf architecture document this is a continuation of this and discussions with others
[20:50:34] --- jasso1 has joined
[20:51:00] <bebemaster> Clausen: Misperception host have prefix's on interfaces
[20:51:24] <bebemaster> Clausen: this breaks MANET for various reasons and is a bad thing
[20:52:22] <bebemaster> Clausen: MANET routing protocols behave better without prefixes, example icmp redirects do no occur when no prefixes are present on the interface
[20:52:53] <bebemaster> Macker: some poeple have been doing this but its a good thing to say that its not a good thing
[20:53:29] --- igarashi has joined
[20:54:07] <bebemaster> The way the picture is drawn has two interfaces on the router
[20:54:26] <bebemaster> Clausen: I don't think there should be different in a MANET
[20:54:47] <bebemaster> I have confusion over the routing/host term and what they apply to
[20:55:24] --- petrescu7 has joined
[20:55:59] <bebemaster> Thalar: if those devices are in all host then if the link on top has some wierdness then the applications don't see the wierdness because of this abstraction sepeartion of router/host on one node
[20:56:48] <bebemaster> Having a dotted line around the router/host would help conceptualize that this can be one box
[20:57:13] <bebemaster> Thalar: can a host be the top of of the box (interface) and can the router be at the bottom of the box (no host)
[20:57:41] <bebemaster> Macker: Had the same question with the solid line and in the slide
[20:58:12] <bebemaster> Clausen: another example with one router and multiple hosts hanging off the back end
[20:58:45] <bebemaster> Clausen: because this is autoconf we are conserned with settting the prefix/addresses of this setup
[20:59:25] <petrescu7> Jari Arkko: so these are regular hosts, nothing special to do
[20:59:40] <bebemaster> Clausen: With this example we can use normal address assignments for the host addresses on the back end
[21:00:13] <bebemaster> Jari Arkko: so these hosts are normal hosts?
[21:00:16] <bebemaster> Clausen: yes
[21:01:07] <bebemaster> Clausen this "On-Link" address can be link-local address, unnumbered, /32 /128
[21:01:34] <bebemaster> Clausen: interface has multicast cabibilities
[21:02:28] <bebemaster> Clausen: the wierd interface is only seen by MANET protocols and hosts only do not see the wierdness
[21:03:21] <bebemaster> Clausen: the link needs on-link address assignment and or prefix assignment
[21:04:00] <bebemaster> Clausen: the "hosts" can be assigned with other non-manet auto configuration procedures
[21:04:54] <bebemaster> Clausen: any quesitons regarding this Architectural Model
[21:06:25] <bebemaster> Narten: I don't fully understand how you are setting this up.
[21:07:00] <bebemaster> Bob Hinden: this is just an architecturel model not a solution
[21:07:32] <bebemaster> Narten: then we should specifiy if these interfaces are virtural or real
[21:07:48] <bebemaster> Bob Hinden: so it is possible to build it physicly this way?
[21:07:55] <bebemaster> Clausen: yes you can build it this way
[21:08:38] <bebemaster> Thalar: typically mulitple "hosts" can be an ethernet hanging off the router and the single host can be a loopback interface
[21:08:47] <bebemaster> Clausen: yes
[21:09:24] <bebemaster> Templen: it could be many ethernets or whatever hanging off the back not just one
[21:09:56] <bebemaster> Arkko: what we need to do for the MANET interface
[21:10:36] <bebemaster> Narten: the more I understand this the more I am getting confused because if there isn't just one address how do we do duplicate detection on the interface
[21:11:10] <bebemaster> Clausen: we are conserned with configuring the upper routing interface
[21:11:45] <bebemaster> Clausen: prefix is related to the host end on-link address is related to the MANET side
[21:12:55] <bebemaster> Arkko: so duplicate address detection...if all neighbors are located then nothing is needed if they move around then what are we going to solve.
[21:13:32] <bebemaster> Clausen: I think we need to clarify what the problems and entities and tasks are before taking on solustions
[21:14:34] <bebemaster> Perkins: How much MANET is this architectural Model applies to. What about the disconnected already assigned nodes?
[21:14:53] <bebemaster> Clausen: if they have addresses then your in the wrong wg.
[21:14:53] --- petrescu7 has left
[21:14:59] --- petrescu7 has joined
[21:15:31] <bebemaster> Perkins: you would like to have this applied to manet's which some nodes have addresses and others do not have address?
[21:15:55] <bebemaster> Clausen: I think we are entering a solution area here and outside the scope of this discussion
[21:16:17] <bebemaster> Perkins: I can go with that as long as you aren't disallowing this type of behaviors
[21:17:06] <bebemaster> Arkko: I feel its a fine requirement that we support this type of behavior with some nodes needing autoconf and others not
[21:17:37] <bebemaster> Macker: I think this is consistant with manually configured manets
[21:18:02] <bebemaster> Macker: the link local thing is okay I am unsure about requiring DAD in all cases but it should be something that we explore
[21:19:29] <bebemaster> Clausen: reiterating that the "hosts" we are not solving dad as it behaves in a normal way but we will be looking at the DAD solution at the mulithop top part
[21:19:49] <bebemaster> Perkins: Manually configured is not the same as existing addresses
[21:20:51] <bebemaster> Taylor: I think we should use another term for DAD because we can have other solutions for setting unique addresses
[21:21:59] <bebemaster> Perkins: if you wanted to do /128 to model one of the things going on in manet but..
[21:22:18] <bebemaster> Perkins: that means you don't have much on the bottom then
[21:22:45] <bebemaster> Clausen: the top comes out of a different space than the bottom space
[21:23:26] <bebemaster> Perkins: I haven't seen many un-numbered links in manet
[21:23:57] <bebemaster> Macker: manet is scoped at looking at v6 so we currently aren't looking at this but this stuff has been done in previous work
[21:24:45] <bebemaster> Templeton: Suggestion for bottom part "classical link model"
[21:25:13] <bebemaster> Macker: where do you expect this to go? the architecture document?
[21:25:39] <bebemaster> Clausen: do we feel this is somethign we can put into the existing architecture document?
[21:25:45] <Seung Yi> any way you can make the slides available online?
[21:26:23] <bebemaster> they may be located at http://www.lix.polytechnique.fr/hipercom ill check...
[21:26:39] <bebemaster> Perkins: I think I am missing something...
[21:26:52] --- becho has joined
[21:28:13] <Seung Yi> can't hear all the voices. Could all speak to the microphone, please?
[21:30:08] <bebemaster> the slides are currently up on the web but I will bug thomas after to get them put up and provide a link where they will exist as soon as I can get it
[21:34:21] <Seung Yi> I think the business of configuring the bottom/classical IP link and the business of configuring upper/MANET interface are orthogonal. The only thing to be careful about is not to restrict the solution for the upper part so that the lower part cannot happen.
[21:37:31] <Seung Yi> can't hear the voice from the back..
[21:38:35] <bebemaster> Bob Hinden: I am for this model.
[21:39:50] <bebemaster> Taylor: There is one thing missing hosts may be wanting "normal" things such as dhcp server which you can relay
[21:40:26] --- petrescu7 has left: Replaced by new connection
[21:40:45] --- petrescu7 has joined
[21:40:54] <bebemaster> Macker: depending on stateless or stateful things happening you may or may not need dhcp
[21:41:28] <bebemaster> Clausen: its too early to talk about requiring dhcp
[21:42:10] <bebemaster> Taylor: modifiation of my statement is... not requiring dhcp but providing enough information to the "router" to support autoconfiguring of the hosts
[21:43:52] <bebemaster> Taylor: the hosts may want more information than just addresses and we shouldn't disallow this
[21:45:24] <petrescu7> Taylor is Dave Thaler
[21:47:38] <bebemaster> thank you I am bad with names and thomas accosted me to do this....
[21:48:35] <bebemaster> apologies for the confusion
[21:49:57] <petrescu7> it's fine, you do great.
[21:52:22] <bebemaster> wolfgang: would like examples of this instead of just an abstract outline
[21:52:46] <dthaler> slight correction to the characterization of my statment about 9 lines up... hosts don't want a dhcp server per se, hosts want non-addressing info such as DNS server addresses, etc. It's the router (which may be acting as a dhcp server/relay for H's on the bottom) that needs an address of a dhcp server
[21:53:02] <dthaler> in order to relay the request to it
[21:53:16] <bebemaster> Clausen: is this a good model? the consensus was yes
[21:53:17] <dthaler> which is normal behavior for an edge router in the Classic IP Link Model :)
[21:53:18] <petrescu7> wondering how will this turn into _not_ becoming DHCP
[21:53:38] <Seung Yi> I agree, Dave. Autoconf should address more than just getting the router an address.
[21:54:37] <dthaler> the manet router itself doesn't need to know anything about DNS server information or other non-addressing information (it could but need not), just like a normal edge router
[21:54:52] * bebemaster has changed the subject to: AUTOCONF problem statement
[21:55:38] <bebemaster> Berc:00 published shortly after ietf66th 01 version is out
[21:56:12] <bebemaster> Berc: wg consensus required on key issues to go forward
[21:56:19] <Yangwoo Ko> Having DHCP relay functionality is *NOT* part of standard, is it?
[21:56:30] <petrescu7> maybe they'll need a means for auto-selecting the MANET router that becomes _the_ DHCP Server.
[21:56:45] <Yangwoo Ko> I mean whethter it is edge router's mandatory funcationality.
[21:56:53] <dthaler> not part of RFC standard (that I know of, maybe Ralph would correct me), but standard practice for edge routers
[21:57:26] <Yangwoo Ko> Sometimes, it confuses me whether is just common pratice or part of standard. :-(
[21:57:26] <bebemaster> my understanding is that dhcp relay functinality is not goin to be required but is may be supported...dhcp relay for example would not be needed with /32 /128 addresses
[21:57:57] <dthaler> I'm talking about the case where you have a subnet with multiple hosts behind the manet router
[21:58:08] <dthaler> e.g. on an ethernet
[21:58:33] <thomas.heide.clausen@gmail.com> Folks, this is an interesting discussion, which we should keep up - however I just want to point out that at this point, we're not yet developing specific solution-space guidelines - we've been talking purely an architectural model at this state.
[21:59:58] <petrescu7> Thomas, what MANET-specific criterion makes this MANET Router not become a DHCP Relay or Server? JohnM mentioned fundamtental things, CharlieP mentioned fundamental needs of wireless links unreliable, etc. But what?
[22:00:45] <thomas.heide.clausen@gmail.com> Alex: I'd like to discuss that, but right now I have to follow the PS presentation, so can we take this over a beer in the bar? Or post-meeting in general?
[22:01:36] <petrescu7> yes, sorry, I understand that.
[22:02:55] <bebemaster> Taylor: the full-DAD model is not a problem statement but a solution to a problem so the text might want to be moved to another document which would be a solution
[22:03:48] <bebemaster> Berc: what do you think of the other diagrams? they are meant to be problem statements.
[22:04:32] <bebemaster> Taylor: I am not sure about the others but the Full-DAD model is more of solution than problem statement.
[22:05:53] <bebemaster> Berc: one of the main goals of the problem statement document is to align with the archeture document
[22:06:38] <bebemaster> Clausen: we need to enumerate clearly the problems without getting into the solution guidelines
[22:07:30] <bebemaster> Clausen: a second order would be looking at the solution guidelines after clearly stating the problems
[22:09:06] <bebemaster> Clausen: we should try and stay away from making multiple documents but it should be seperated out into two documents to make it clearer
[22:10:28] <bebemaster> Macker: the solutions don't belong in this document but more explaining the context and rational of the problems
[22:11:29] <bebemaster> Macker: I would like to hear consensus on sepeation of this document into two seperate documents
[22:12:38] <bebemaster> Berc: the last wg was that the document was focing dad so we added problems which were non-dad or partical dad to the document
[22:13:00] --- petrescu7 has left: Replaced by new connection
[22:13:27] --- petrescu7 has joined
[22:19:05] --- becho has left
[22:19:56] <bebemaster> berc: consensus on removing the problem guidelines from the problem statement document
[22:20:18] <bebemaster> Thalar: another issue what are we configuring
[22:22:41] <dthaler> i.e. "the problem" is not just about prefixes/addresses.
[22:23:14] <bebemaster> Clausen: important to keep in mind is that the current problem statement has been written before this discussion and we do appreciate the work put into the document and look forward to the next iteration
[22:23:19] <dthaler> it's about making sure the gaps get filled so we end up with a complete solution when combining all the protocols/WGs involved, such that any hosts you might attach behind the router work normally.
[22:26:30] --- igarashi has left: Computer went to sleep
[22:26:33] <bebemaster> templtin: layer 3 a manet routing allows for normal autoconfiguration as the manet looks like an ethernet to higher layers
[22:27:01] <bebemaster> Thomas: are you proposing an alternitive model?
[22:27:53] <bebemaster> Thomas: if the manet looks like an ethernet then we are done we can go home
[22:28:18] <bebemaster> Templtin: yes but only if you are willing to accept the tunneling overhead
[22:28:38] * bebemaster has changed the subject to: solutions document
[22:28:57] <bebemaster> Templtin using dhcp for autoconf in manet
[22:30:34] --- peetu has joined
[22:30:41] --- igarashi has joined
[22:40:37] --- petrescu7 has left: Replaced by new connection
[22:40:44] --- petrescu7 has joined
[22:44:33] <bebemaster> for those that asked thomas' slides will be updated to the IETF proceedings for autoconf this evening
[22:44:41] <thomas.heide.clausen@gmail.com> https://datatracker.ietf.org/public/meeting_materials.cgi?meeting_num=67
[22:45:01] <thomas.heide.clausen@gmail.com> (is where they'll be as soon as the session is over and I've gotten time to pdfify them)
[22:46:36] --- monden has left
[22:46:39] --- ks has left
[22:46:40] --- petrescu7 has left
[22:46:41] <bebemaster> thank everyone
[22:46:45] <bebemaster> thanks even
[22:46:52] --- Yangwoo Ko has left: Computer went to sleep
[22:46:56] --- jasso1 has left
[22:46:58] --- bebemaster has left
[22:47:49] --- igarashi has left
[22:47:50] --- thomas.heide.clausen@gmail.com has left: Logged out
[22:47:55] --- Seung Yi has left
[22:48:09] --- yowada has left
[23:35:06] --- dthaler has left