On Tue, 2002-12-17 at 11:35, Tom Lane wrote:
> AFAICS, that is not a misstatement. B-trees are still the only
> recommended index type for concurrent-update situations ...
> hash has deadlock issues and the other two are just plain not
> concurrent.
Yes, but what I disagree with is the recommendation that "concurrent
applications" should use B-trees -- a concurrent application that wishes
to index non-scalar data won't have a lot of success using B-trees,
page-level locks or not...
Cheers,
Neil
--
Neil Conway <neilc(at)samurai(dot)com> || PGP Key ID: DB3C29FC