A while back I was implementing a distributed transaction coordinator, and in the process also learned some of the pertaining theory. The only thing I can remember semi-vividly was a proof showing that you can't synchronize an action between two parties over an unreliable link, but I can't seem to find it any longer. The final step of the proof was something like "thus, if this was possible, you could shorten every conceivable protocol by one step, which reduces all protocols to the zero-step protocol, which is impossible". I'm pretty sure that it's a standard result of computer science, but I'm unable to find the right keywords to locate it. Help?
posted on Aug 18, 2006 - 5 answers