DB Abstraction Layers

Hi folks. Given the last response by Justin to my comment about some people's dislike for DB abstraction layers I'm electing to throw this out into the general discussion as a new topic. I have mixed feelings about abstraction layers. At times I love them (mostly) and at times I hate them. I posted a problem to a postgresql forum recently where I was having a query return different results from the command line and from within PHP, and it turned out to be a bug in ADOdb's debugging code which was the problem, incorrectly reporting the query which had just run ('%,123' is not the same as '%, 123' but it will drive you insane!!!!! ;o). I hated it for a brief time then ;o) The link below points to a page which contains a message from someone in response to my post looking for help solving the above problem, someone whom I consider to be quite, quite knowledgeable about databases and whose opinion I very much respect, so I think it deserves a read. It makes some interesting points: http://web.unbc.ca/~gossep/dbx.html Anybody care to throw their hat into the ring for a debate on the pros and cons of using DB abstraction layers such as ADOdb (which I really like) against Pear::DB and PHP's DBX functions, for example?? Cheers, Pablo