At 11:31 22/12/97 +1100, Julian Assange wrote: >The problem here is that nntpcache tries very hard to provide >up-to-the-minute news and the default high-water-mark check timeout >reflects this. I agree having an option to continue serving from the >cache in the face of server errors is probably a good idea >though... though it requires caching of server errors - but this can be >done on an nntpcached wide basis. More importantly, nntpcached shouldn't be telling people that a group doesn't exist if it can't connect to the server which carries it. That screws up many readers badly and causes newshound to remove the group from its precache list. wrt article numbering, all sorts of weird effects occur if a newsgroup is read on one server, which is unavailable when the group is accessable on another server. Local article numbering would be a very "good thing" IMO. AB
