The hash algorithm for the string class is not very efficient. It's fairly easy to drive things into a pathological state where many strings hash to the same bucket. The Java string hash algorithm is well documented, and also functions fairly well, so this would be a great replacement for the existing model. Code to implement this is in the 4.0 code base.
Anonymous
Logged In: YES
user_id=1125291
Originator: YES
Committed revision 1227.