Static vs SLAAC - Static expected to be preferred?

[Mark Smith]

On Wed, 27 Apr 2011 15:38:08 +0200
"Steinar H. Gunderson" <sesse at google.com> wrote:

> Den 27. april 2011 15:14 skrev Mark Smith
> <nanog at 85d5b20a518b8f6864949bd940457dc124746ddc.nosense.org> følgende:
> > Do you have any comments on your general static address
> > expectations?
> 
> Well… one thing is expectations, another thing is what would be useful.
> 

I'd say usually they're the same. My expectations of static addresses
overriding dynamic addressing are based on my dynamic verses static
routing experience, and my similar experience with IPv4 static and
dynamic addressing. I'd be very surprised if those are uncommon
expectations, but that is why I'm asking.

> In Linux (at least Debian, which I'm used to), it's actually pretty
> hard to get servers not to take RA—thus, using static addresses by
> default would certainly be useful. It sort of feels like the wrong
> knob for the issue, though; I can certainly see situations where I
> would wish that a server contact external hosts over a less “official”
> address, and always preferring static addresses would seem to make
> that impossible.
> 

With my suggested mechanism, statics would be preferred by default,
because they have infinite preferred and valid lifetimes. If you want a
host to use non-static addresses for outbound connections, but have a
static address for an inbound connections, you'd configure the static
address with a zero second valid lifetime, and leave the valid lifetime
as infinite. In other words, the static address is permanently
deprecated, and therefore not preferred for outbound connections.

> /* Steinar */
> -- 
> Software Engineer, Google Switzerland