I used to not be a big fan of frameworks such as Rails (Ruby), Django (Python) or Cake (PHP) for two major reasons.  First, the frameworks did not seem all that mature requiring attention to patches and upgrades and raising at least the possibility of winding up on a “dead” branch.  Second, the frameworks seemed to add a lot of overhead (e.g., in the ORM parts) so that you would need more hardware and were less able to handle traffic spikes.  But I am slowly coming around to a different view. The major frameworks are clearly maturing and are in sufficiently wide use to assure ongoing support.  More importantly though, we have the potential of moving framework based apps into the cloud much more easily.  The reason is that with SQL abstracted away it is much easier to provide a cloud implementation of the data access.  For instance, the models in Rails do not have to run through SQL – they could be implemented in a way native to a cloud infrastructure without breaking the Rails application.  It is still early days on this latter point, but companies such as Engineyard and 10gen (USV portfolio co) are working on solutions that will allow framework based apps to scale easily.  I should hasten to add that this means you should try to do as much as possible inside the framework and resort to custom SQL only as a very last resort, as that custom code will most likely have to be rewritten to a particular cloud datastore.