>>55
It actually isn't this clean cut, I need to reflect on this more. I might just read the Cormen or Knuth book mentioned. Luckily I have lots of time to decide.
>>44
With regards to this question the conclusion I came to was that it's basically write to only use local variables/functions when your sure no other procedure would benefit from their use. I also concluded with regards to macros that the same rule applies, except that some emulation of namespaces becomes exceptionally important. You could easily accidentally overwrite the memoization table so some even slower implementaiton or even worse one that doesn't work at all and have to deal with the joys of debugging that.
I just finished writing another question here when I realized the answer, I feel like just taking the time to write out my questions makes them clear. For those curious, in this case I was going to ask about adhoc abstractions, where an abstraction is created without any clear indication of what this thing is at a conceptual level just to pull out some repeated behavior. These seem to me to be useful heuristics, they are not in themselves beneficial but they point towards some more authentic hidden abstraction which can then be reified and clarify the process of the computation at hand along with make future procedures easier to define. The adhoc abstractions seem to typically be over specified.