Resumption performance about draft-westerlund-tsvwg-dtls-over-sctp-bis

[gloinul]

So from the expected use cases I think the benefit is limited, but you are strictly right that this is resource consumption save. But, if you rekey approximately once an hour, the cost of that rekeying is fairly insignificant, but we should write incorrect things. We will address the text to be factual correct, and if we want to give any recommendation we will motivate it better.

[emanjon]

Let's write that resumption has performance benefits for the Web but that these are not really important when setting up the parallel connections in draft-ietf-tsvwg-dtls-over-sctp-bis. The intention was not intended to imply that resumption has no advantages in general.

A more important aspect based on this comment is that I think the text in draft-ietf-tsvwg-dtls-over-sctp-bis seems to assume that resumption is only used for the parallel connections. Resuption could also be used for the initial DTLS connection. E.g., if the association is shut down for some reson and has to be restarted. Here resumption could have performance benefits for draft-ietf-tsvwg-dtls-over-sctp-bis

[emanjon]

I still think the old text was technically correct for parallel connections. Parallel connection are setup relatively infrequently, the nodes will have a single connection, and latency is not important when setting up a parallel connections. I have a hard time see any significant benefits. However I don't think this sentence need to be in the document at all.

I removed the sentence and replaced it with a sentence describing that "Resumption can have significant latency benefits for quickly restarting a broken DTLS/SCTP association."