13 Apr
2010
13 Apr
'10
7:46 p.m.
Radek Krzywania wrote:
Regarding race conditions, it's not the role of the protocol to prevent it.
Most protocol purists argue that a protocol is defined as a state machine and a message format that change the state. If race conditions are part of the state machine, it is part of the protocol. In practice there are plenty of protocol specification which only define the message formate. Most IETF protocols do not specify a state machine. That said, race conditions is not an issue right now. If it should be considered from the start or if it is better to add it later as an afterthought, I don't know. Regards, Freek