Database Abstraction Layer

hi, i want to start a discussion, because i got confused. i'm developing database software with php. initially i used the common functions mysql_* or pg_*. then i moved to phplib to have more flexibility and to be able to port my applications. well, now i have to code a more complex application that needs transactions and has to be scalable enough for "heavy business". so i started to look out for solutions: i found: - adodb - metabase - pear i am very confused, because i thought that pear is meant to be the "code-central", but i read in the newsgroups that pear the pear group starts its own approach of coding a database abstraction layer. all of the coders are open-source-ppl but it seems to me that there wasn't a real try to combine alle that work to one big one into pear. i would like to know what other people think about that ? which solution would be the best considering flexibility and secure programming ? thx in advance i would appreciate a nice discussion kostas pyliouras