ipv6 next-hop link-local
nanog at 85d5b20a518b8f6864949bd940457dc124746ddc.nosense.org
Tue Feb 15 22:01:42 CET 2011
On Tue, 15 Feb 2011 21:35:33 +0100
Roger Wiklund <roger.wiklund at gmail.com> wrote:
> My understanding is that all routing protocols that support IPv6 are
> using link-local as next-hop.
I think the idea behind that is that link locals are always present and
independent of any other addressing and it's/their preferred or valid
lifetimes. This isn't a new idea, IPX, Appletalk and IPv4 have
had "this network" prefixes before. In IPv4, it is the node/host
address with all zeros in the network/subnet portion.
> But when I configure a static route I usually use the global unicast
> as next-hop. When I issue a show route, there is an inconsistency
> where dynamically learned routes have link-local as next-hop and my
> static route has the global unicast.
> Are you making any efforts to keeping it consistent? I.E specifying
> interface and link-local as next-hop even for static routes?
It would make sense to try to keep your routing protocols' next-hops
decoupled from your global or other addresses. To make supporting
static routes easier, you could configure static link local addresses
e.g. fe80::1/64 etc. It probably wouldn't hurt if the SLAAC derived
link-local addresses were left there, although you may wish to disable
them so that all protocols that want to use link locals will use your
statically configured ones.
More information about the ipv6-ops