On 8/23/13 9:38 AM, Truman Boyes wrote:
> Keying for network resources is highly complicated and will likely
> have little adoption. 

There have been problems with adoption of security technologies
for routing, but I'm not sure it's because of difficulty around
keying - we see plenty of other instances of key provisioning
for things like VPN tunnels, management consoles, etc.  Granted
that they're almost always done badly.

And of course that's not a reason not to try to secure an
nvo3 tunnel - if nothing else you've got to protect the control
plane from things like spoofing, replays, etc.  That almost
always requires cryptographic protections (pace people who
run everything through an ssh tunnel).

> We could do it, but today's trust boundaries typically live on the 
> NVE/PE/ToR/etc, and are not truly multi-org, but rather multiple 
> projects that can be authorized at the application level.

I think there's a proper middle ground here.  The control plane
cannot go unprotected, period, and ssh tunnels don't provide
the kind of granularity or key management facilities that are
needed.  I'm always (well, almost always) happy to see group keys
being considered but there are some relatively subtle issues
that they introduce that need discussion.

Melinda
_______________________________________________
nvo3 mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/nvo3

Reply via email to