> The caveat to this is that intermediate devices may also need to parse
> packets but not necessarily participate in the protocol control plane.
> Consider a stateless firewall that parses LISP packets to filter on
> encapsulated IP destination address. If the P-bit is implemented in
> the end hosts, but we've neglected to update all the firewalls in the
> path-- the packet will be misinterpreted if say the firewall sees a
> LISP packet with an encapsulated Ethernet frame. In this case,
> hopefully the firewall will drop the packet, but there's no guarantee
> of that-- the behavior is non-deterministic. Maintaining compatibility
> with such devices is a hard problem and might imply constraints on new
> options that could fundamental change parsing or interpretation of the
> packet (adding protocol type is a good example case).

I think your point here is (and if I interpreted incorrectly, I'll make a new 
point then), that if the encapsulated packet is L3 or L2 and described with a 
port number versus a bit in the payload after the UDP header, it would be 
easier for firewalls to decide when to filter one versus the other?

Dino

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

Reply via email to