On Tue, Aug 24, 2010 at 8:29 PM, Hemant Singh (shemant) <shem...@cisco.com> wrote: > Shane, > > > -----Original Message----- > From: Shane Amante [mailto:sh...@castlepoint.net] > Sent: Tuesday, August 24, 2010 3:35 PM > To: Hemant Singh (shemant) > Cc: sth...@nethelp.no; adur...@juniper.net; > dtha...@wollive.windowsmedia.com.akadns.net; ipv6@ietf.org > Subject: Re: 6man discussion on /127 document @ IETF78 > >>Hemant, > >>We don't statically configure, (nor would we even consider doing so), > default routes or want "hidden" default routes in our Core routers. If > those default routes are installed into the FIB of the >routers, > (intentionally or accidentally during "implementation" of your > proposal), it would have substantial negative repercussions on our > operations, namely traffic forwarding loops (whee!) in our >network that > would likely cause congestion, etc. Furthermore, there are way too many > 'hidden' configurations/nuances inside every router vendor's > implementation that I don't want to keep track of >another one ... > >>Please, as an operator, I really just want IPv6 /127's to be as KISS as > /31's are in IPv4. > > I agree that simple is always desirable. We are just discussing > internals of a router so that one gets around the genuine problems > raised with /127 which is to not have the router start anycast > forwarding on an interface configured with a /127. As for the routing > loop, let's see the cases when routers are configured for a global /127 > address. Router A sends a packet destined to Router B in a p-p network > between the routers. The packet gets sent out Router A and Router B > terminates/consumes this packet since the packet is destined to B. > Likewise for B sending to A. The only other case is when A sends a > packet to B where the packet destination is not B but C, so the router B > looks up the routing table and finds a route to C and ships the packet > out the interface that the routing table found to egress on. Only if > router B does not have a route to C and then the router has no choice > but to use the default route, then a loop situation arises, but then see
shane, level3 has a default route in their core? since when? Hemant, shane's point is that there is no default route in his network, there are no default routes in lots of networks. having one suddenly appear is a super extra bad thing, trust me, i've done it... people get very angry when a few million packets per second arrive all of a sudden... Also, in the scenario, what makes routerA the 'default' ? what makes routerB? who decides? how is that decision made? is it deterministic? across all vendors and across all implementations inside a single vendor? (I'm looking at you XR|IOS]CatOS|wifi-crap-os... and to be fair: JunOS|E) to echo 3 other people... this is a very bad plan. If you put a /127 on a ptp link all it ever is is 2 addresses that are part of a very small (2 host-addr) subnet. it doesn't need nor want anything like RA or ND on it. -chris -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
