The TSP applet is not intended to be anywhere near as comprehensive as  
your own work on the problem.  For example, my brute force implementation  
is very naive.  It does serve two purposes though.  Firstly as a  
demonstration of how to use the components in the framework to solve a  
problem and secondly as a demonstration of the effectiveness of  
evolutionary appoaches.  You are probably right though, with cross-over it  
could be an even more effective demonstration of the power of evolution.  
It was something that I initially intended to include but, for whatever  
reason, I didn't.  I may well add this in the next version and see what  
improvement I get.

Did you implement ordered cross-over as defined on this page  
(http://www.permutationcity.co.uk/projects/mutants/tsp.html)?

I think they are all in the Mitchell book.

Stochastic Universal Sampling is a fitness-proportionate strategy similar  
to Roulette Wheel Selection except that it ensures that observed selection  
frequencies are in line with expected frequencies.  It is most useful for  
small populations where statistical anomalies could skew the results  
considerably if Roulette Wheel Selection was used.

Rank Selection assigns selection probabilities as a function of the  
relative fitness scores rather than absolute scores.  This can increase  
selection pressure when there is little variation in the population or  
reign it in when the fitness scores vary wildly.  My default mapping  
function is very trivial (for a population of size n, the fittest  
candidate gets a score of n, the next n-1, etc. and selection is  
proportional to these scores).  This default linear function can be  
over-ridden with something else (maybe in some situations it would be  
better for it to be exponential or logarithmic).

Dan.

(I have removed comp.lang.java.programmer from the follow-ups as this  
isn't really on-topic)