[Zope] Choosing between ZODB and relational database/filesystem storage
Fri, 25 May 2001 18:17:27 +0200
> However, I think this is a weak argument for more complexity.
Actually we have even more complexity. We use Java servlets (over xml-rpc) as
a middle layer between Zope and the database (Oracle). Only transient
presentation-level data is stored in Zope. We chose this structure for
o It allows use to leverage all our existing competence, and we can work on
different layers in parallel without conflicts.
o The data structures and the inter-layer protocols are only changed through a
clear and defined process. They form important and 'automatic' documentation.
o Every layer is relatively easily replaced or even collapsed into another.
We can also have horizontal separation in each layer - for example using an
application written in C to provide some of the functions in the business
logic (Java) layer.
o Even though I love Python I feel more comfortable doing heavy computing in
Java. The JDBC interface to Oracle also works better for us than the Oracle
adapter(s) for Zope.
o We are buzzword-compliant.