Please stop spreading FUD. Something doesn't have to have a SQL interface to be powerful. Berkeley DB supports applications with multi-threaded concurrent read-write access to databases of hundreds of terabytes at rates of over 30,000 accesses per second. Much of this native functionality is not available through the DB_File interface, but it is through BerkeleyDB.

Now if you care about the complexity of your data model, then MySQL wins. If you care about preserving your data, in my professional opinion a good dbm used properly is as good or even a better choice than MySQL. If you want both, then you can use Postgres, Oracle, etc. But while I admit that the tied Perl interface is not the snappiest (though as some have discovered, it is pretty good), I do not believe that for straight lookups you are going to find any relational database that will beat a C application using Berkeley DB. It ain't going to happen, and it ain't going to happen for a lot of reasons.

So unless you know something about scalability that I, Sendmail, Cisco, Ask Jeeves et al don't, please don't fall into the trap of thinking that dbms are just lightweight toys.

For more information I suggest you visit Sleepycat Software.