Short version: Is there a way to shorten the timeout of the "net backend" so that after about 10 seconds without a response from a computer running saned, scanimage -L will move on and not hang around for minutes on end awaiting a response?
Long version (for those who need more context): I have a total of 5 Fujitsu scanners at 5 remote locations. Each Fujitsu is attached to a computer (which I'll call a "scan-computer") running saned. The scan-computers all live on the same VPN so a 6th, central, computer can safely use the net backend to connect to the scanners via the scan-computers. This setup works relatively well. Running a "scanadf -L" to get the list of scanners only takes ~4-5 seconds. That is until one of the VPN connections is severed, either because someone turns off a scan-computer, or someone's DSL is down or something. When that happens, running a "scanadf -L" on my central computer ends up taking 5-6 minutes! This delay is crippling to my setup. I've looked around and I can't seem to find a way to set a shorter timeout value for the net backend. net.conf doesn't seem to provide a way to pass a value and there doesn't seem to be an environment variable I can set. I'm even willing to find some global TCP timeout variable that would shorten the timeout for all TCP connections initiated by my central server. Lastly, I'm willing to patch the net.c file to use the setsockopt call to lower the SO_RCVTIMEO to a value defined in some environment variable, say "SANE_NET_TIMEOUT". The last option seems a bit drastic, so I'm hoping there is a better/easier way. Thanks, Ryan
