RevConference - Client Server vs Stateless

[HyperChris at aol.com]

In Jerry Daniels presentation on thin clients he stated a strong preference 
against using the client-server model, which I took to be Rev talking to the 
database directly, in favor of what he called a stateless model in which rev 
made http calls to the Apache/MySQL/PHP setup.

The obvious question I have is why ? Do the stateful operation of direct 
database calls create a reliability problem? (The little voice in my head says I 
am missing something here!)

Also, it occurs to me that the Rev engine could replace PHP and allow one to 
simplify that aspect of it a little bit. I say a little bit because one would 
still have to right scripts for the Rev engine that translate ones tags into 
database calls. Has anyone taken this route and care to share its effectiveness 
?

Thanks.