At 09:19 AM 27/01/2009, you wrote:

>Hi,
>
>I use FB 2.1.1 (superserver) on Windows Server 2003, XP, and Vista. It
>appears that dead client connections (e.g. due to client-server
>network failure) are not cleaned up by the FB server. Using isql "show
>users" command, I can see that the dead connections never gets cleaned
>up.

Never say "never". When (eventually) the server knows that the client has gone, the configured TCP/IP timeout will kick in. Once the timeout has elapsed, the network will inform Firebird, and Firebird will do the proper cleanup. Most Windows versions' timeoouts default to 4 hours but you can configure it (and *how* depends on what version of Windows the server runs on.)

>In firebird.conf I see the "#DummyPacketInterval = 0" commented out.
>Although there is a strong warning stating that this setting should
>not be used, can I safely use it under the OS listed above?

There's no point. It's a legacy from very old (winsock 1) bugs that affected Windows 3, Win 95 and NT3 and it never really worked as intended.

>Is there an elegant way to ensure dead connections are cleanup up?

Firebird doesn't control the network in any sense: it just rides on the train. Until the network tells it otherwise, it thinks all its co-travellers are still aboard. As with any networked application - configure the network server appropriately.

./heLen