The "make-sphere-generator" is basically a generator on the homogenous
space SO(n + 1)/SO(n). I am wondering how complicated it is to add
random generators for the elements of SO(n) itself (the probability
distribution will be given by the Haar measure, of course).

Random generators for the most important compact Lie groups will be
quite useful in applications and more than the
"make-ellipsoid-generator", which is already in SRFI 194.

Some theoretical and implementation background is in the introductory
article [1].

Marc

--

[1] http://www.ams.org/notices/200705/fea-mezzadri-web.pdf