The following proposal provides a simple, minimal interface to random number generation. It allows both controlled random seeding and also a single procedure that provides randomness. It is designed to be simple and fairly useful.
Randomness is a common feature to use in education and general programming. Therefore, it makes sense to have WG1 standardize randomness to some degree. The current proposals on the table are either too complex or do not provide the necessary features. Specifically, when teaching about randomness, it is important to be able to control the way that randomness falls out when grading assignments or testing them.
I do not see any reason to have two procedures for getting a random value. These can be combined into one without requiring much trouble. Additionally, this one procedure is more general than the two procedures suggested by RandomCowan. It is also more simple than SRFI-27. The possibility of returning a 0.0 value from random is intentional to make the procedure behave in a more consistent manner. This should not be a problem, as suggested by RandomCowan, with log if (log 0.0) is defined to return -inf.0.