The meaning of the "idleTimeout" param is little different than most idle timers, and can cause your readers problems if used incorrectly (it bit me). Basically, if any NNTP command takes longer than idleTimeout to complete, the reader is disconnected. So, if the client is just slow, or the server is just slow, or if the reader has issued a big XOVER or requested a really big article, the connection will dropped, if it takes longer than idleTimeout. I found that 10m is not enough of an idleTimeout considering groups with 60,000 headers in them, and a server that may get a bit slow during peak time. I think nntpcache should disable the idle timer when processing requests, and re-enable it when waiting for a request. I think most of the code is guarded enough so it doesn't need an idle timeer. Or, additional calls to alarm() need to be added to the xover and article retrieval sections. Tom
