I''m in a similar position: I''m writing a 2-player Asteroids clone which I''m trying to network. See my homepage for its latest (single-player) incarnation.

As I see it, you don''t need to exactly synchronize both computers. Just halve the average round-trip ping time to make sure both games start at roughly the same time, then store the local start time on each computer. When sending information packets, include the time since the game started. Then, when you receive a "create missile" packet, say, actually create the missile at the point it would be (taking into consideration its velocity) if it were created on your own end at the same time since the game started.

If one computer is actually running a few ms behind the other, it will get slightly better response times.

For each missile, I''ve given the originator control over collisions. Therefore, neither player has an advantage. If I had made one computer the server, the player on that end would have a huge advantage. The only tricky bit is with missile-player and player-player collisions. In these cases, both ends confirm that the same thing''s happened on the other end before proceeding. Since there''s a natural break when a life is lost, this isn''t much of a problem.

Everything has a velocity, but only the players'' ships have any acceleration. Every time an object is created (missiles and broken asteroids) a packet is sent. When the players thrust, a packet is sent with their position, updated velocity, and the time.

I do make sure that packets get through (with ACKs), but for thrust packets, the new packet overrides any old (unacknowledged) packet containing a previous thrust.

Before the game starts, and at the start of a new level, the computers synchronize (roughly) and agree on the position of the asteroids before starting.

Once I''ve finished implementing all of the above (nothing is set in stone yet), I''ll put a 2-player version on my homepage.

Dave