Sunday, October 29, 2006

How to make progress by shoving problems under the carpet

I just commented out all of the loadbalancing code in the java port, on the grounds that it makes far more sense to develop loadbalancing at a JDBC driver level (the same place one normally does connection pooling). Once I get the application working with a single server, I'll come back to the problem of loadbalancing, which should, if done correctly, be almost completely transparent to the application anyway.

I'm certain that this problem will prove to be challenging, but I think it makes far more sense to reengineer it than to try to perpetuate an implementation which is designed for a totally different environment (specifically, one without a database abstraction layer).