From: David Vallner <david@...>
Date: 2006-11-03T06:20:18+09:00
Subject: Re: still more relentless non-repetition

--------------enigAC890B69156F3593D79D893C
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

Giles Bowkett wrote:
> (by the way, Ferret actually runs faster than Lucene, according to the
> Ferret site. I don't know how, but that's pretty cool.)
>=20

A) All benchmarks lie, -especially- ones people use to pimp their own
products.

I'm not claiming this is the cause though, because:
B) Ferret uses a C extension, and the benchmark is probably that
version, not the pure Ruby one; I imagine advanced fulltext search
algorithms do quite a fair share of numbercrunching, where C would win
out. Of course, with Ruby it's entirely possible marshalling of results
between C and Ruby would kill the pure performance gain (speaking
generally, not of specifically Ferret anymore), but it's possible to
skew a benchmark so that doesn't show - a large dataset and a small
result set for an extension that instead tends to be used frequently for
larger result sets. A full processor usage profile on datasets and
result sets of varying sizes would tell books above a simple benchmark.

David Vallner


--------------enigAC890B69156F3593D79D893C
Content-Type: application/pgp-signature; name="signature.asc"
Content-Description: OpenPGP digital signature
Content-Disposition: attachment; filename="signature.asc"

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (MingW32)

iD8DBQFFSmEKy6MhrS8astoRAkc9AJ4gBW1zs+Xq+4wlYnzcWRB3rl6KKwCbBMMo
ciD7d8P4o4hvVLcL53il4p8=
=4LN/
-----END PGP SIGNATURE-----

--------------enigAC890B69156F3593D79D893C--