> Actually, this is more or less how nntpcache does it. There is only > one process that read/writes the history database, and all > nntpcache-serverlets communicate with it via ipc. Presumably this means nntpcache is going to throttle prematurely on SMP systems. What about more finely grained locking using the first 8k of cache.mmap for locks/semaphores and giving all processes the ability to read/write to disk? Not that I've ever seen nntpcache CPU bound..... -Chris