Hm... the worst-case-scenario would be that the user comes to the brilliant idea to adjust the time on his computer. So the answer to your question, if you can rely on the internal clock is: no! But to be honest I also don't have a bullet-proof solution. I would probably go with you solution and would add an extra
thread, that would start counting the seconds as soon as the computer is offline and checks, as it becomes online again, if there are dramatic differences between the seconds counted in the thread and the time-difference between offline and reconnect from the internal clock. If there's a huge gap between those two values, you know you can't trust the timestamps.
But I don't know if this method I accurate enough, especially if the cpu has a high load during the offline-period and the thread is not able to count *exactly* one second. If this would give you trouble Suns
Java Real-Time System might be solution (
http://java.sun.com/javase/technologies/realtime/index.jsp).
Due to the fact you haven't mentioned what kind of data you are proceeding a few questions: What happens if the data of one client is lost? What happens if the order is not correct? And is there a chance to increase the intervall to, let's say, 10 seconds?
