Imported from previous forum
[ original email was from Bob Lamoureux - ]
One of the most poorly documented tenets of the FIX spec has been the fact that both parties must be able to re-transmit all messages pertaining to a particular session. Since sessions span across Logout boundaries, the sequence number series must be unique across all messages until a new session is established.
I like the idea of a flag indicating a new session but I think it should be in the LOGON message. People are not very religious about sending a LOGOUT to terminate the session (or the process gets killed by some scheduler program before the LOGOUT is sent.) However, EVERYBODY demands a successful LOGON message and a flag indicating this is a new session might work better here
[ original email was from Scott Atwell - ]
I like the idea of using the Logon message for this as well. In this case, a Logout would never need to be issued since Logons can be exchanged within an existing session for the purpose of PGP/DES/MD5 key change. This could allow systems to run continuously for 24 hours/day.
We would have to recongnize that once a sequence number reset to 1 has been performed by both sides (in the middle of some market some where around the globe), that any messages prior to the reset would be unable to be resent for any reason.
One concern might be how much time people’s sytems might require in order to do whatever they may have to do to change sequence numbers (i.e. flush their log tables, etc). If we establish this feature, we may want to suggest that systems be designed to wait at least X period of time when waiting for the response to the initial logon requesting sequence number reset.