Whether connection pooling can be done "easily" depends on how the underlying database client library handles sockets. AFAIK Oracle OCI doesn't support sharing connections between different clients, meaning that you'd actually have to write your own server daemon that sits between your web server and Oracle. This is seriously non-trivial to get right.

At a previous job we wrote such a beast - actually at the time we used FreeBSD for the front-end, and there were no Sybase libraries available for that platform. So we designed a simple communication protocol, with a server on the front-end listening on a Unix domain socket and then communicating with a multi-threaded server on the middle tier that itself talked to Sybase. It worked reasonably well, but there were always problems, and maintaining the two multi-threaded apps was a pain.

As perrin says - just use Apache::DBI, and maybe use a lightweight http proxy front-end to limit the number of processes that need to have a database connection.

Michael

P.S. - now if you were using Sybase instead you could use Apache::Sybase::ConPool :-)