From: James Rogers (jamesr@best.com)
Date: Tue May 15 2001 - 13:17:36 MDT
At 11:08 AM 5/15/2001 -0700, Adrian Tymes wrote:
>That said, it'd still be good for its stated purpose of database
>searches...*if* special drivers can be written and shipped with some
>big name DB software that would make use of this, as opposed to
>treating the data store like any other.
The usage of the technique for database searches seems contrived.
They are using a quantum O(k) algorithm to replace normal O(k) methods
(e.g. linear hashes), and it wouldn't even address the more useful O(log)
methods (e.g. B*tree) that can be used for things like indexing for range
scans. And in any case, the large collection of indexing algorithms
currently known tend to be extremely scalable on normal hardware for the
most part. The biggest benefit from the quantum technique is that it might
save space normally allocated for indexes. In most real world cases, the
benefit of a quantum database lookup might be a small increase in retrieval
performance, which I suspect is probably not worth the additional cost of
the special hardware (as opposed to just buying faster/bigger ordinary
hardware). I guess I just don't see a real benefit in this type of case;
if the entire computer was already a quantum optical system, then I would
expect the databases to already support this feature where it made sense.
Quantum techniques would seem to be much more useful for exhaustively
sifting key spaces that are too large to actually be stored anywhere, or
for which a physical record doesn't need to exist.
-James Rogers
jamesr@best.com
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:07:40 MST