The current specification for random deviates allows
parameters of the deviates to be random deviates themselves.
I am not sure whether this was intended.
Should the parameters be restricted to non-deviate expressions,
or should random deviates be specified to use the mean value of their arguments?
Since "expression parameter" can refer to deviate expressions indirectly,
non-deviate argument for deviate expressions cannot be enforced by the schema.
The current specification for random deviates allows parameters of the deviates to be random deviates themselves. I am not sure whether this was intended.
Should the parameters be restricted to non-deviate expressions, or should random deviates be specified to use the mean value of their arguments?