IETF
roll@jabber.ietf.org
Wednesday, December 16, 2015< ^ >
Thomas Watteyne has set the subject to: IETF93 ROLL WG meeting
Room Configuration
Room Occupants

GMT+0
[12:51:46] eintopf joins the room
[16:04:53] <eintopf> I am done with the new context entries direction.
[16:05:11] <eintopf> s/direction/directories/
[16:05:40] <eintopf> and I dropped the context table managment from radvd, except somebody make entries in radvd.conf
[16:05:48] <eintopf> then this should be 6LBR
[16:06:02] <eintopf> I suppose
[16:06:52] <eintopf> What I meant is "AdvSendAdvert off;" and declare context inside radvd, then this is wrong.
[16:07:21] <eintopf> and I also don't "zero" the context table when somebody starts radvd
[16:08:23] <eintopf> int set_interface_ctx_active(const char *iface, uint8_t id, uint8_t val);
int set_interface_ctx_compression(const char *iface, uint8_t id, uint8_t val);
int set_interface_ctx_plen(const char *iface, uint8_t id, uint8_t plen);
int set_interface_ctx_pfx(const char *iface, uint8_t id, struct in6_addr pfx);
int setup_allrouters_membership(int sock, struct Interface *);
[16:08:35] <eintopf> this is what radvd now offers for setting
[16:08:50] <eintopf> but I still not sure if radvd can also handle the lifetime management
[16:10:08] <eintopf> because when delete one entry(makes it inactive), it is not deleted directly (they are still some time inside the table). Not sure if putting this logic inside kernel or userspace
[16:10:19] <eintopf> maybe I should try it into userspace
[16:10:23] <eintopf> and then I know maybe issues
[20:55:41] eintopf leaves the room: Disconnected: closed
[21:54:05] eintopf joins the room
[22:24:16] <Michael Richardson> lifetime. hmm.  in theory, one should mark an entry as expired to the kernel, so that it no longer uses it for *transmit*, but still can decode using it.
[22:26:17] <Michael Richardson> the order should be: 1) fill new entry for new CID, mark it as not-yet-alive. 2) announce new one to network, wait appropriate time for propogation [a clue that it's gotten out is that the previous CID usage goes to zero for receive. 3) mark new CID as active, mark old CID as inactive. 4) announce old CID expired and/or reuse it.
[22:26:40] <Michael Richardson> that says that we need some usage counters, and a way to enable/disable the entries in the kernel for receive/xmit seperately.
[23:06:19] Michael Richardson leaves the room: Disconnected: No route to host
[23:30:45] Michael Richardson joins the room
Powered by ejabberd Powered by Erlang Valid XHTML 1.0 Transitional Valid CSS!