Op 8 nov. 2012, om 16:56 heeft Acee Lindem het volgende geschreven: > I noticed it had been reduced from minutes to 30 seconds in this version. I > guess that rules out RIPng. Since this is a new specification we'll take > lower hello/dead under advisement. However, I doubt we go as low as 1 and 4.
Values similar to RSTP would be fine; hello=2, dead=6. Dead=8 would be more robust, for lost or delayed hello. Wait doesn't need an adjustment if we go for less slow hellos. Teco > Acee > On Nov 8, 2012, at 9:24 AM, Teco Boot wrote: > >> >> Op 8 nov. 2012, om 14:03 heeft Acee Lindem het volgende geschreven: >> >>> >>> On Nov 7, 2012, at 11:18 PM, Teco Boot wrote: >>> >>>> I checked the draft-ietf-ospf-ospfv3-autoconfig-00. I think the proposal >>>> doesn't meet expectations of users, with regard of protocol convergence. >>>> The default timers are far too conservative. First reconfig on OSPF router >>>> in my hands is adjust timers on high speed interfaces to hello=1 and >>>> dead=4. This is a bit better than what old days bridges with STP do. >>> >>> This was not considered a requirement in the home. Refer to the section 3.5 >>> in the homenet architecture document. >> >> Right. Dead=40 does *not* meet the requirement. >> >> See also http://www.ietf.org/mail-archive/web/homenet/current/msg01245.html >> >> Teco >> >>> >>>> Minor: the timers are not defaults. These are sample values. I have seen >>>> routers with different defaults than RFC 5340, for certain interface types. >>> >>> I'd expect homenet ethernet and WiFi interfaces to default to the broadcast >>> type. >>> >>>> >>>> For autoconfig OSPF, I prefer removal of the restriction that all routers >>>> on a link MUST use same timers. That provides flexibility like OLSR. If >>>> this is not possible because of reasons for DR selection, I suggest >>>> removal of DR completely and configure all interfaces in P2MP. Or MANET >>>> interface type, if we want to. Question is: which MANET extension. >>> >>> I'm not sure if there is a real requirement. Routers on the link need to >>> agree on the timer values lest we impact more of the protocol than >>> necessary. Hence, if this is required, the way forward would be for routers >>> on shared links to adopt the timers from their neighbor(s). For simplicity, >>> the same rules used for DR preference could be used. >>> >>> Thanks, >>> Acee >>> >>>> >>>> [posted here, because requirements come from Homenet] >>>> >>>> Teco >>>> _______________________________________________ >>>> homenet mailing list >>>> [email protected] >>>> https://www.ietf.org/mailman/listinfo/homenet >>> >> > _______________________________________________ homenet mailing list [email protected] https://www.ietf.org/mailman/listinfo/homenet
