[SnapPea-planning] Finishing SnapPea 3.0

[Marc Culler]

On Thu, Nov 06, 2008 at 10:08:12AM -0600, Nathan Dunfield wrote:
> Nathaniel,
> 
> Great to hear that you are interested in working on SnapPea.   Let me  
> give some quick answers now, with perhaps more once I've thought about  
> things a bit.

While Nathan is thinking, here are a couple of things that we did, or
plan to do, in the python version.

(Done) It turns out that the Smith Normal Form algorithm is
notoriously nasty in that, during the course of the computation, it
tends to generate numbers that are much much bigger than either the
numbers in the input matrix or the output matrix.  This causes SnapPea
to overflow when trying to compute the homology of a not-so-big finite
cover.  The python version uses PARI to do this computation.  (And,
due to the design of PARI, this means embedding all of PARI into
SnapPy.)

(Planned) Walter Neumann's student, Christian Zickert, has an
algorithm that will compute Chern-Simons invariants quickly and
directly, without requiring initialization.  This should be
implemented in the SnapPea kernel.

- Marc