ietf-nntp Notes from the IETF41 meeting

[John Hascall]

> Would anyone care to share with me their notes from this meeting? It would
> be much appreciated so I don't miss anything in the "official" summary.

    I didn't take any notes, but here are my comments on the "dynfeed"
    extension.  My suggestion was that the protocol be (roughly):

      IHAVE xxx   ------------->
                 <-------------   "DONT WANT THAT"
      WHYNOT      ------------->
                 <-------------   REASON(S)

    And here is my reasoning.

       (0) We don't even want to consider a change to the
           basic send-receive... state machine.

       (1) The receiver does not know everything he doesn't want
           initially (e.g., it may get a message for a group it
           doesn't want but has never heard off) so we can't just
           do a protocol exchange at the start of the connection.

       (2) You could change the protocol such that the "DONT WANT"
           includes (say as a multiline) the REASONS but that end
           doesn't know if the other end will understand that.
           Worst case something breaks, else just wasted bandwidth
           as ignored REASONS are included in the replies over and
           over.

       (3) With the exchange above the sender can keep a flag
           like the following pseudocode:


           enum whynot_states { unknown, yes, no } whynot_understood = unknown;

              /* if we get a dont want, see why not if we can */
              if (whynot_understood != no) {
                 reply = send_whynot();
                 if (reply == unknown_verb) whynot_understood = no;
                 if (reply == reasons) {
                    whynot_understood = yes;
                    process_reasons();
                 }
              }

           which would cause only 1 WHYNOT to be sent if the
           receiver end didn't understand that verb and which
           would be really unlikely to break any implementations
           which didn't understand it.

John Hascall