Bob Tanner wrote: > > Quoting Ben Lutgens (blutgens at sistina.com): > > > design does not lead to a distributed processing model. The bottleneck will > > > always be the mbox file and the write-exclusive lock to it. > > > > Yeah, too bad there's not a filesystem that supports byte-range locking > > (that I know of) > > A better solution, IMHO, would be to strip the archiving part of mailman out and > make it a seperate process. Thus, you'd be able to distribute the > receiving/delivery and archiving/web processes. > > The hard "hit" on the list server is appending files to the huge mbox files. But > the next hardest "hit" is all the search engines spidering the archives. > Maybe search engine spiders ought to be limited to the middle of the night? Given the P.I.T.A. factor associated with maintaining a fast, working mailing list, "choice B" might be to examine a Mn.tclug usenet newsgroup...