>Gert Doering wrote: >> Hi, >> >> On Mon, Jan 05, 2009 at 02:38:17PM +0000, David Freedman wrote: >>> Even when you get it, it is only implemented for link-local addresses >>> so you have to use RA or static routes :( >> >> Unfortunate, indeed. Do you know whether there is work in progress to >> get it fixed/improved to handle "global" router IP addresses? >> >> While it might not follow the IETF's vision of "how things should be", >> we prefer to configure our servers' default route towards well-known >> router addresses (<subnet>::1), and have them ignore RAs...
And nothing stops manual configuration, with perhaps a minor tweak in the address of that default gateway (IMHO fe80::1 is not the right approach!). > >There is significant difference of opinion in the IETF about the value >and future of RA in IPv6, as I found out recently when I tried to figure >out how IPv6 DHCP was supposed to work (answer: the spec is broken) IMHO, and I could be wrong, I think that was really a very vocal minority ... the vast majority of people I deal with, clients I work with, etc. have no problem with RAs being used, for the default route host configuration anyway. (FWIW - SLAAC vs (stateful) DHCPv6 both have advantages for the rest of the host-config pieces, and I see no reason to not have both be available) > >This thread (if you have an afternoon free) and surrounding threads are >worth reading: > >http://marc.info/?l=ipng&m=122391355810549&w=2 > >As far as I could tell: > > * IPv6 and the RA mechanism were spec'ed back when DHCPv4 was not >widely used, and IPX autoconfig was the model they were aiming for > > * Virtually no work has been done in the field since then, so the >hard-learnt lessons of the last decade in IPv4 are simply not there in IPv6 Or perhaps consensus continues to favor IPv6's current approach? > > * There are a lot of architecture astronauts in the IETF (IPSec for ND >- whose idea was that?) Well, SEND has promise ... > > * The stateful-address/stateful-other bits in the RA are junk. > >The posts by David Hankins of the ISC agree with my personal position, >including this one: > >http://marc.info/?l=ipng&m=122406652232186&w=2 Again, I think we are far from having consensus on RA deprecation ... while the current handling of M & O bits are far from optimal, I don't see the RAs themselves going away. /TJ _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
