ipv6 next-hop link-local

[Mark Smith]

On Wed, 23 Feb 2011 08:40:48 +0100
"S.P.Zeidler" <spz at serpens.de> wrote:

> Thus wrote Mark Smith (nanog at 85d5b20a518b8f6864949bd940457dc124746ddc.nosense.org):
> 
> > On Mon, 21 Feb 2011 11:02:28 +0100
> > "S.P.Zeidler" <spz at serpens.de> wrote:
> > 
> > > This is either not stateless, or you have trouble after all your systems
> > > have been off (UPS outage, anyone?),
> > 
> > Just need to define a tie breaking algorithm to select the new seed to
> > then choose the new subnet id. Highest or lowest IID would probably do
> > after DAD has been completed. I think most of the "packet functionality"
> > exists in RAs - PIOs for SLAAC can be announced by an RA without the
> > device issuing an RA being considered a default router, by setting the
> > router lifetime in the RA to zero.
> 
> So you have a unique local prefix, for what exactly?

For the same purposes as link locals are or can be used for today, with
the difference being that the inbound/outbound interface doesn't have
to be specified, as the ambiguity of which interface the address
exists on has been eliminated.

> Since you will come up with something totally different after a power
> outage, when you need extra hassles least, you cannot use it to
> identify a fileserver, appserver, etc; basically, you can not -use-
> these addresses for anything that is not based on dynamic polling.
> 

Dynamic or Multicast DNS (a.k.a. Bonjour, Avahi etc.) would provide name
resolution for them if you where using them for more general
applications communication. It would be possible to make generating and
configuring the unique link local subnet ID manual, however I think
it'd be worth making it automated and agreed, so that IPv6 on the local
link autoconfigures itself (to easily facilitate the the IPv6 "dentist's
office" scenario).

They'd be functionally like ULAs, except that they wouldn't be routable
off-link (a useful security property), and have a separately designated
prefix so that if necessary they can be identified in places where the
identifying the type of prefix and it's reachability is useful or
necessary.

Regards,
Mark.