On Wed, Mar 30, 2016 at 02:16:18PM -0400, Danny Mayer wrote: > On 3/29/2016 10:44 AM, Sharon Goldberg wrote: > > As such, I'm not convinced that sending these KE messages according to > > the NTP polling interval makes sense. I'm further concerned because I > > don't understand how the NTS KE will deal with lost packets. > > Some of this was discussed elsewhere. There are two cases to look at: > > 1) Client sends a packet and the server doesn't receive it > This is the simple case. If the client doesn't get a response from the > server it can reissue the same request and hope that the server gets it > the second time. Open issue is how many retries should there be before > it gives up?
I think this is handled by NTP. If it's a persistent association, the client doesn't give up, it will just increase the polling interval up to the maxpoll. If it's an ephemeral association, it will be demobilized at some point. I'm not sure if NTS needs to change anything in that. > 2) The client sends the packet and the server receives it and responds > but the client doesn't get the response. If the client now resends the > request, what should the server do? Resend? Ignore? Does it need to > remember what it last sent? This is a much harder case to resolve. I'd expect the server to respond to all requests it can or is configured to (e.g. with rate limiting), like an NTP server normally does. Should it be different with NTS? -- Miroslav Lichvar _______________________________________________ TICTOC mailing list [email protected] https://www.ietf.org/mailman/listinfo/tictoc
