![]() |
![]() |
|
You're here: Home |
Robert Loomans wrote:
I still believe that msg_ref is enough for replay protection (and synchronization too), and that that is the minimum we need - no need to clock syncs and the rest.If, and only if, the server/parent retains the last msg_ref used indefinitely.
Is it so damn difficult to keep track of one integer per client (parent)?
The timestamps and the time windows are to reduce the burden on the server to remember state.
And, they are also there so that we can introduce a complete algorithm and clock syncs, massively complicating everything.
In the case that the child is rebuilt, it makes re-starting communications easier... otherwise the child would have to do a search for the next msg_ref.
The only thing that's needed to resync is the last msg_ref the parent knows about. The client can find that out as that is already in the protocol. So what "search" would this be?
Robert