Hi i am looking for in depth knowledge on how replication really works ons SQL2000.

I have 2 servers with different security context and want to implement replication on it.

The problem is the replication should be protected against network outages and the network should be used at least possible.
Their will be no much data replicated so i thought about push subscriptions.

First test was to setup a push subscription. When there was no network and no transactions and the network came up again everything stills works fine automatically.
If during transactions their was a little hickup after 10 times a little hicking cumulative in time (10 network outages of 1 sec in 2 hours) the distribution agent stopped and must be started again manually.

Second test was to setup a push subscription with queued updating. To my surprise When there was no network and no transactions and the distribution agent went in failure . ??? no transactions were done , so distribution normally didnt have to do anything. The queued agent however did not went in a fault phase.

Third test was merge replication. If there was network and no transactions that failed also and must the agents must be restarted manually.

So there is a significant difference between all. and the books online of microsoft dont go in that detail so that you can understand everything what happens.

Does anyone knows good method of replication that is fully automatically and dont need manually intervention when network breaks down and comes back up?

Romaric