> From: Christoph Rippel [mailto:crippel / primenet.com]
I have to follow up on this email:

> [...]
> As discussed in a private email your and my (optimisitic) algorithm are 
> identical(modulo details). My main objection about your current version
> is that it depends on the Hash-Hash realization of an ``Idbag'' and does
> not abstract this away as an implementation detail. 
> In principle a Hash - Hash (of Integer) encoding and the (Sym)Id_pair
> encoding are actually more or less identical and in an ideal version 
> you would write the whole ID-bag class in C. 

Soorry Ben, 

I was writing nonsense (I should have looked more carefully)
your new version does abstract away the implementation details 
- actually very nice!


[...]
 
> I also think that your objections towards symmetrization - i.e. swapping
> are not valid since they cost very little compared to the reminder division 
> invoked in any hashing scheme and running the interpreter it self of course
> (in my test it makes virtually non difference if you swap or don't swap - 
> even so swapping does not gain anything in the ``examples'') 
> - swapping IMO simply feels more robust.
Still true soo (but easy to change ...)


Christoph 
[...]