As usual, the MySQL crowd is doing unrealistic benchmarks that will
obviously will favour their "speed is all that matters. who cares about
everything else a database should do?" approach.
If they'd remember their undergraduate computer architecture class,
they'd know that the only benchmarks that truly matter are the one
with real applications.
Just think about it: How many times in the lifetime of a project, do
you insert 100,000 rows all at once? How many times do you do a
count(*) on a table you _know_ is huge? These are not realistic uses
of a database, unless you have no clue of what you're doing, or are
still learning.
Now, how many times do you have users entering bad data and maybe you
forgot to put a check? How many times do you do queries that are not
so trivial as a count(*)? How many times do you use a subselect in
your queries? How many times do you need to write stored procedures to
make your life easier and programming faster?
You want a real benchmark? Go look at the tests Tim Perdue did with
MySQL and PostreSQL this past january (beore the improvementh in 7.1
that Don mentions) at http://www.phpbuilder.com/columns/tim20001112.php3.
THAT is a real benchmark. Look at the results and decide what's more
important: being fast on unrealistic operations or being safe and
faster or realist operations.
