On Thu, 2003-05-01 at 17:55, Richard Grevers wrote:
I thought the reason we were having this discussion
was because articles
were repeating too often, indicating some fault in the randomizer?
Well, two things:
First, the existing assigned indexes were very un-randomly distributed,
which would may have lead to a somewhat increased chance of duplicates.
I re-randomized the whole set this morning, and I just now ran 100
random pages and got no duplicates.
Second, I'm not 100% sure the randomizer is reliable under heavy load.
We're running MySQL 3.23.54; I notice a line in the changelog for
3.23.56 (released in mid-March, apparently, when I wasn't looking) which
may be relevent:
* Better RAND() initialization for new connections.
Were we not about to replace it with a new machine and MySQL 4.0, I'd
recommend trying an upgrade. :D
-- brion vibber (brion @
pobox.com)