
Ryan's ideas:

use mpfr  (multi-prec) for tnorm.surv etc

use root finding too 
