Constrained Nelder-Mead

All,

In looking at `optim', it doesn't appear that it is
possible to impose nonlinear constraints on Nelder-
Mead.  I am sufficiently motivated to try to code 
something in C from scratch and try to call it from
R....

Does anyone have some good references to barrier
and/or penalization methods for Nelder-Mead?  I would
ideally like some papers with pseudocode for method(s)
that are in some sense optimal for continuous 
functions.

Thanks,
   
     Greg