2004-01-09T06:16:24 Eugene Crosser: > I think that memory-based approach will be big win even if it is > limited to parsing multipart/nested messages.
With tmpfs, the biggest performance difference is the latency of the file I/O syscall API -vs- direct mmapped access, and the possibility of avoiding some extra copying. And ext2 (no journal) w/ noatime is but little slower than a real tmpfs for very transient tmp files. Unless these actually prove out to be major performance bottlenecks dominating nclamd, I would definitely leave 'em for "maybe some day". -Bennett
pgp00000.pgp
Description: PGP signature
