On Mon, Aug 26, 2013 at 3:04 PM, Laurent Alebarde <[email protected]> wrote:
> 9) Separing curvezmq_codec_execute in two parts, one for the client and the > other for the server would quite not take more lines of code, and would > provide better clarity and maintainability IMO. Hmm, yes, indeed. > 10) In my understanding, the respect of nonce sequences, for long nonces > that use them, should be controlled by the peers. Let's say the client uses > 1, 3, 5, etc. and the server 2, 4, 6, etc, the client should control that > the nonce is 2 when it processes the welcome. In case of sequence rupture, > the connection should be closed. Not particularly; the sender chooses a nonce that only has to be unique. Using an incrementing counter is one strategy but the recipient shouldn't be validating that. -Pieter _______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
