Joel,

> On Oct 29, 2019, at 10:08 PM, Joel M. Halpern <[email protected]> wrote:
> 
> I presume that most silicon implementations of VxLAN VTEPs do not have any 
> logic for trapping out BFD packets under any circumstances.  While some may 
> have been built anticipating this draft, we have to assume that many will not 
> be able to support this.  So it goes when you add features to a protocol.

Just to be clear, BFD itself (RFC 5880) doesn't necessarily require the IP 
portion of the stack. What's required is sufficient information for an 
implementation to manage the BFD session for the transport in question.

Things that tend to run IP usually end up with a bit of programming that tells 
it that the forwarding layer can trap on a given 5-tuple signature.  Things 
like BFD on LAG will catch a layer deeper in some circumstances such as startup.

This discussion is running rather similar (albeit more publicly) to the BFD on 
LAG discussion as various vendors try to figure out how they can successfully 
trap BFD for the testing point in question.

-- Jeff

Reply via email to