On Tue, 3 Mar 1998, Tom wrote: > I'd like to know how effective nntpcache is in this situation. > Particularly on how whether you can get away with leasing less connections > from the upstream server, then you actually need, because of benefits of > caching. I have nntpcached feeding primarily off a site I co-admin in Chicago. Whenever I check during evening peaks, nntpcache has at least a dozen parallel connects open. I'd suggest that if you're doing it to try and reduce the number of connects to the upstream, some code hacking would have to be done to add a "MaxUpstreamConnects" parameter. You can restrict nntpcached to one connect to the upstream server by specifying a port to bind to in nntpcached.access (this caught me out when I was trying to track usage and thought I'd be smart by making nntpcached use one port per upstream, but it was a slow dog when set that way), however there's no intermediate stage available between one and unlimited as far as I can see. I'm surprised a provider is charging per number of concurrent connects. I'd have thought that they'd charge on a "number of IP addresses" or "articles served" or "overall traffic Mb" basis. AB
