A method is disclosed for determining how often a client station in a
network should send keepalive messages. Based on a measure of network
load, a presence server determines a keepalive period, which is a time
interval in which a client station needs to send a keepalive message, and
the presence server reports this keepalive period to the client station.
The client station responsively sends a keepalive message to the presence
server within the determined keepalive period.