Hello, >From time to time the server disconnects clients, usually when there's a longer period of inactivity. The message logged says "child timed out". According to the code, this happens if no "tickles" were received from the client in 2 minutes, i.e. 4 of them. This is more than just annoying because the user is not able to save his work. The volume where file resides is still on the desktop, can't be removed and, effectively, all work done after last save is lost. The server is Linux/Alpha 2.2.14 running netatalk 2.1.4-37b which was compiled with -DUSE_FLOCK_LOCKS. Everybody's connected to a 10/100Mb switch (no errors on the network). This is happening only on clients running MacOS 8.5.1, Open Transport 2.0.2 and Apple Share Client 3.8.1. Clients are synchronized using NTP to the server. Or, at least they should be, but I found the NTP setup on Mac a little bit awkward. I don't know what is exactly the reason for AFP over DSI to treat connections this way. I find this behaviour arguably incorrect or at least in contradiction to TCP. Even so called keep-alive packets (which, if without response, bring the TCP connection down after two hours of inactivity) should be considered harmful. I would really like to hear why there are various timeouts implemented within the DSI code. Cheers. Dejan
