Scott Howard <[EMAIL PROTECTED]> wrote:
> Start with an uncompressed archive, and try again. You should also be
> careful to make sure that any disk caching/etc doesn't get in the way...
Yes, that is very important. The varying delays caused by writes
to disk can amount to the same dimension as the differences in your
results. If you put an uncompressed tar archive on a tmpfs and also
extract it there (sufficient RAM presumed), you will arrive at a
somewhat different conclusion.
The significance of a test on tmpfs is doubtful, of course, since
the scenario is not completely realistic. But if you test with an
on-disk filesystem, statistical evaluation of a multitude of test
runs for each implementation is indispensable.
Gunnar
_______________________________________________
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org
