Joe Touch <[email protected]> writes:
> FWIW, why not just add the entropy in the IPv6 flow ID rather than
> expecting it at the transport layer? Intermediate network devices should
> be relying only on the flow ID for that entropy anyway.
>
> (and yes, that doesn't solve the problem for IPv4, but perhaps that's a
> good reason to encourage use of IPv6)

Given that we're talking about encapsulations, I see that Geneve has a
"Virtual Network Identifier" field that can be treated as part of the
entropy.  If it isn't needed to identifiy a virtual network, it can
loaded with entropy from the encapsulated packet.  In any case, packets
with different NVI values can be freely reordered.

And there is an 8-bit reserved field which could be defined to be
entropy.

These fields are at a fixed location relative to the containing headers,
so they're easy to find.

It might also be useful to define an option entirely for containing flow
identifiers or entropy -- the essential definition being that packet
routing is allowed to treat it as entropy, two packets with different
values for this field may be reordered freely.  And recommend that if
this option is used, it should be the first one, so routers could find
it easily if it is present.

Dale

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

Reply via email to