Hello All,
This question popped into my head this evening, what is the right way to
handle multiple algorithms which purport to calculate the same thing? There
are, for example, a couple of ways to calculate the student t cdf. What is
the common's philosophy on deciding:
1. Whether to allow multiple algorithms.
2. How an algorithm is included.
a.) Does a 'bug' or shortcoming need to be shown in the current
implementation?
b.) Say that algorithm a works for a numerical range and b works best on
another. Are both included? Is a new 'meta' algorithm written which mixes
both a and b?
3. Does simplicity count?
4. Does speed matter?
A while back, Chris Nix reimplemented the SVD routine. I am not sure I
remember the old routine so I cannot say there was anything worth keeping
there. However was there a conscious decision to scrap it? Why not have it
live side by side with the new one? (Again, I am not saying the old
algorithm was better-Chris' contribution definitely was valuable). I think
we will run into these issues often.
Thoughts? If this has been discussed already, my apologies.
-Greg
