Pro:

* More well-tested than ZODB and MiddleKit.

* Relational databases are based on solid mathematical foundations
  (relational algebra, set theory, etc.), but OODBMS have no such
  theoretical underpinning (Link to comments by Fabian Pascal).

* Full control over relations, transactions, etc.

* All queries are programmer-written and optimized, and by reducing
  the number of queries you give the RDBMS more opportunity to do
  optimizations

* You can leverage the database's abilities to cache and store
  information compactly

Con:

* Non-Pythonic: requires you to translate your objects and
  manipulations to/from SQL statements

* No layer of abstraction between the database schema and the Python
  interface -- if you change the schema you must change all the Python
  code unless you create your own abstraction layer or use views more,
  either of which is highly recommended

* Queries are programmer-written and optimized, which means more work

* It is difficult to cache results (though the database does caching
  of its own)

* Some operations are difficult to do in SQL

.. _Link: http://searchdatabase.techtarget.com/tip/1,289483,sid13_gci769023,00.html?FromTaxonomy=%2Fpr%2F284872

-- MikeOrr_ - 31 Dec 2001
-- EdmundLian_ -- 03 Jan 2002
-- IanBicking_ - 04 Jan 2002