> Anyway I now think that we might be better off with the other idea of
> abandoning an insertion and retrying if we get a lock conflict. That
> would at least not create any performance penalty for non-concurrent
> scenarios; and even in concurrent cases, I suspect you'd have to be
> rather unlucky to get penalties as bad as the heavyweight-lock solution
> is showing.
Agree, it would be a better workaround for now. I will be able to do this at
this friday.
I considered the idea to forbid placement of child on the same page as parent,
but this implementation a) could significantly increase size of index, b)
doesn't solve Greg's point.
We definetly need new idea of locking protocol and I'll return to this problem
at autumn (sorry, I havn't time in summer to do this research).
--
Teodor Sigaev E-mail: teodor@sigaev.ru
WWW: http://www.sigaev.ru/