One thing I forgot to suggest: when measuring performance, don't forget to
take into account any access control done by your server. If your server
reads extra files (like NCSA's ".htaccess" file), then things will be
slower for files in directories with access control.
If you want to be formal, you could define different classes of benchmarks
for different access control methods:
- no access control
- host-based access control
- basic user/password
- etc, etc
-andy