> Paul, I'm afraid "hurd" isn't quite English. You can chose between
> herd and heard, as in:
>
> We're off like a herd of turtles!
> A herd of turtles?
> Of course I've heard of turtles!
> Why would I care what a turtle heard? I have no secrets from

Thanks, now I will never forget. Homonyms are hard, and indeed 'hurd'
isn't even a valid word. Perhaps I suffer from Stallmanitis. For
clarity, I meant to write 'herd', not 'heard'.

> The third reason is that result caching works much better than
> statement caching on most competitive benchmarks, even
> if they are next to useless in practice.

Your first two reasons were highly MySQL specific. This one is
general. Does this imply that you don't agree with the observation on
the demand pattern of web apps & alike 3-tier stuff?

> The more modern practice is to use memcached so at least the stale
> data is on your local machine.

Yup, memcached is the band-aid we're trying to get rid of.

> It doesn't always get the right answer, of course, but then neither
> does result caching. Result caching is a disaster with MVCC and
> transactions.

Why? Fundamental issues or shoddy implementations? I'd say Oracle's
solution is correct in view of your concerns, and the MS solution
could be correct, if done right.

Paul