On 7/11/12 8:51 AM, Larry Kreeger (kreeger) wrote:
For example, I assume that the switches/routers providing the underlying
network are owned/secured/administrated by the trusted data center
administrators who also control what ports on those devices connect to
their own trusted devices vs tenant devices which would be considered to
not be trusted.

I think that's a critical distinction when thinking about how
to secure this mechanism.

Based on that, I suppose you could extrapolate to say that the outer
headers sent across the underlying network are trusted since the data
center administrator can control what devices connect to the underlying
network.

This is pretty far outside my comfort zone but it seems to be a
common assumption in actual deployments, so let me ask this:  Do you
want to put that particular constraint (that it always be deployed
in "trusted" environments) on the protocol?  My feeling is that
since the concept of a "trusted data center network" and a "trusted
data center administrator" are, in general (there are exceptions),
pretty bogus, that's an unrealistic constraint, and because it can
be anticipated that there will be some number of deployments that
ignore that constraint anyway, nvo3 needs to be somewhat more
responsible about security than "it's all on a trusted network,
anyway" would suggest.

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

Reply via email to