I completely disagree. Relational databases are *meant* to scale like
this.
Your suggestion of using flat files will break a lot quicker than my
relational database. Go ahead and try to dump 250k files into your
filesystem. You won't be able to do it. Every hear of "inodes" ? You'll
need to repartition with a larger set of inodes and a smaller block
size.
I'm more than willing to agree to disagree. I will continue work against
a relational database. It provides much more flexibility, data
integrity, reporting ability, and the ability to scale. If you believe
that you can build a similar system from flat files, then please... go
for it.
XML has nothing to do with filesystem vs. database. That only comes into
play if you attempt to capture arbitrary chunks of XML from the input; I
do not think that is a valid approach. The submitted XML must be parsed
and the relevant data extracted... allowing arbitrary crap from
arbitrary submittors just won't cut it.
Cheers,
-g
-- Greg Stein, http://www.lyra.org/