Jeff/Albert - Given the MTU issue is associated with a link coming up - and the use of Echo would allow the problem to be detected and prevent the BFD session from coming up - and you are acknowledging that the protocol allows padded Echo packets today ...
is there really a need to do anything more? Les > -----Original Message----- > From: Jeffrey Haas <[email protected]> > Sent: Monday, October 29, 2018 8:36 AM > To: Reshad Rahman (rrahman) <[email protected]> > Cc: Les Ginsberg (ginsberg) <[email protected]>; Naiming Shen (naiming) > <[email protected]>; [email protected] > Subject: Re: BFD WG adoption for draft-haas-bfd-large-packets > > Reshad, > > On Fri, Oct 26, 2018 at 06:32:26PM +0000, Reshad Rahman (rrahman) wrote: > >On 2018-10-25, 11:38 AM, "Jeffrey Haas" <[email protected]> wrote: > > > The draft I had previously worked on with Xiao Min discussing probing > using > > > BFD Echo had the concept of probes that would happen wherein the > session is > > > not torn down. The example carries similarly with the "send > occasional". > > > > <RR> We discussed use of echo at IETF102. The large-packets draft > mentions > > that echo can only be used for single-hop, hence the need for padding the > > control packets. But isn't single-hop Albert's main use-case? > > It's Albert's primary use case. And, I think a common related one is > protecting tunnels of various flavors; e.g. GRE or IPsec. > > > I believe we > > should add the echo option in the large-packets draft, it has the benefit > > that you get the desired functionality even if only 1 side of the WAN link > > supports echo. I realize not all implementations support echo so they > > might have to do pad control packets instead. > > While I don't think Albert or I would have any objections to adding Echo > discussion in the existing document, we perhaps risk running into one of the > issues Xiao and I had briefly discussed. Echo is intentionally > under-specified in RFC 5880 et seq. While it's possible that we can simply > put in a discussion section that says "if you use Echo mode with similar > padding, you can get similar benefit", I think that may be as far as we want > to go. > > The related observation is that nothing stops an Echo implementation from > doing this with no changes to the protocol. :-) > > -- Jeff
