IETF
roll@jabber.ietf.org
Sunday, July 10, 2016< ^ >
mcr/credil has set the subject to: IETF96 ROLL WG meeting
Room Configuration
Room Occupants

GMT+0
[00:13:51] Michael Richardson leaves the room
[08:56:55] eintopf joins the room
[08:57:44] eintopf leaves the room
[08:57:48] eintopf joins the room
[14:30:57] Michael Richardson joins the room
[14:44:39] <eintopf> hi michael
[14:45:27] <eintopf> yesterday at some beer festival I maybe found a solution why short address works and we don't need to care about "how do broadcast that node has no short address anymore"
[14:46:01] <eintopf> I don't know the rfc buzzwords (because I am not a ndisc expert) but I may knowing why
[14:47:08] <eintopf> the procedure is: ifup -> node A sends NS - node B receivs NS and answers with NA, if this node isn't soliciated with node A before then it will set the override flag in ICMPv6 message
[14:47:46] <eintopf> after that all NA's from node B will not contain address option fields anymore (until some timer runs out, I suppose or some sepcial NUD_STATE
[14:47:48] <eintopf> )
[14:49:01] <eintopf> but then everything should be fine, because if you change the short address, then the next NA which is an answer for NS will be also set the override flag (and A will parse address options again and see "no short address field is there")
[14:49:28] <eintopf> it will set the NA override flag because it's not soliciated (I think that's the right word) anymore
[15:10:16] Michael Richardson leaves the room: Disconnected: closed
[16:02:42] Michael Richardson joins the room
[16:07:04] <Michael Richardson> hi.
[16:07:48] <Michael Richardson> what you write makes sense. I hadn't finished the email about the problem before.
[16:08:03] <Michael Richardson> beer festivals are good to promote clear thinking :-0
[16:41:27] <eintopf> good
[16:41:33] <eintopf> then I need to fix that mainline
[16:42:09] <eintopf> http://git.kernel.org/cgit/linux/kernel/git/bluetooth/bluetooth-next.git/tree/net/6lowpan/ndisc.c#n104
[16:42:13] <eintopf> remove this else branch
[16:42:35] <eintopf> should be fix that, I should test it with changing short address on one node with previous communication
[16:56:22] Michael Richardson leaves the room: Disconnected: closed
[17:02:40] Michael Richardson joins the room
[17:05:32] <Michael Richardson> I think that in general, the short address will "go away" because it times out, or because it is replaced.
[17:06:15] <eintopf> okay
[17:06:52] <eintopf> currently on destination address we do a lookup in ndisc if short address is there for L3 destination address
[17:06:56] <eintopf> if yes we use it
[17:07:33] <eintopf> but this lookup is much more complex, I already told you that you could compress more bytes depends on L3 and fragmentation to prefer extended address than short
[17:07:41] <eintopf> but I didn't implemented that stuff yet :-)
[17:08:10] <eintopf> otherwise 2 address types with different lengths would make not much sense
[17:08:40] <eintopf> then I would always use the short address and forget the extended, but extended is unique
[17:09:04] <eintopf> short address is unique in combination with pan, but then it's also not an EUI
[17:09:29] <eintopf> =-O I hate this two address types stuff
[17:10:25] <eintopf> "(07:07:33 PM) eintopf: but this lookup is much more complex, I already told you that you could compress more bytes depends on L3 and fragmentation to prefer extended address than short" - is a improvement only, good stacks should do that :-)
[17:11:23] <eintopf> on rx side it doesn't matter if you don't follow that or not, but... we making this stuff to compress bytes and then on tx we should do everything which is possible
[17:12:01] <eintopf> the bad thing is, RFCs doesn't write such recommondations
[17:12:18] <eintopf> you can see it only if you think about the whole standard
[17:12:45] <eintopf> I want to have pseudo code for each RFC
[17:40:34] mcr/soho joins the room
[17:44:30] Michael Richardson leaves the room
[17:46:34] Michael Richardson joins the room
[17:48:30] <eintopf> there is some riot people in berlin
[17:48:41] <eintopf> they don't want to leave some house there
[17:48:49] <eintopf> it's not their house :D
[17:49:48] <eintopf> some punks/political lef-ortiented people
[17:50:01] <eintopf> don't visit "friedrichsheim" :-/
[17:50:21] <eintopf> friedrichshain :D
[19:00:40] Michael Richardson leaves the room
[19:10:51] Michael Richardson joins the room
[19:37:01] Michael Richardson leaves the room: Disconnected: closed
[19:37:21] Michael Richardson joins the room
[19:44:29] <Michael Richardson> yes, so we always want to use the short addresses if we can, because they are shorter, and we want to use them for the L3 address if we can, because we can also compress more.
[19:44:44] <Michael Richardson> by "riot people", I don't think you mean RIOT-OS people :-)
[19:45:02] <Michael Richardson> I've been looking at pictures of the old east/west border fences...
[19:45:18] <Michael Richardson> We want to use the long addresses during bootstrap....
[19:50:20] Michael Richardson leaves the room: Disconnected: closed
[21:42:41] eintopf leaves the room: Disconnected: closed
[23:32:33] Michael Richardson joins the room
[23:37:32] Michael Richardson leaves the room
Powered by ejabberd - robust, scalable and extensible XMPP server Powered by Erlang Valid XHTML 1.0 Transitional Valid CSS!