>=Arthur Flatau

 > The point is that the call to the time function is buried somewhere
 > inside the call to make-random-state (when given the argument t).
 > There is no way to XOR or otherwise modify the number that is used to
 > create a new random state.

Right.  But then there's no real need to, either.

 > This is unlike, C, for instance where you
 > can XOR the time with the machine's IP address (or similar schemes) to
 > initial the random number generator.  The only portable way to
 > initialize the random number generator is with a call to make-random-state.

The definition is arguably a "Good Thing" because it completely hides all 
the implementation details, such as whether it's based on time or 
radioactive decay (see http://www.fourmilab.ch/hotbits/), whether the 
"state" is only a fixnum or something more robust, etc.

Sanitary practice is to always consider all inscrutable system-supplied 
generators simply as somewhat suspect noise sources anyway, which you can 
use to implement suitable randomness for your application.

 > Wow, it is not often that I come up with advantages of C over Common 
Lisp.  :-)

(Philosophically, having to rely on ANY non-open-source generator isn't an 
advantage, no matter how many random features it advertises!  If you want 
real control, code your own...<;-)

It's all pretty academic.  The probability of two programs initializing 
within exactly the same window is probably less than lightning striking the 
user.