I want the option of efficient in storage, as well as speed.  If I can't
call pred->cs later, I have to deliver it as a precomputed bitmap.  Oink.

-----Original Message-----
>From: Brad Lucier [mailto:xxxxxx@math.purdue.edu]
> From xxxxxx@mazama.net  Wed Dec 20 13:34:58 2000
> I urge you to keep your original function, which was elegant: simply allow
> it, optionally, to be lazy.

Sounds no good.  I figure I should be able to implement such a thing
efficiently and I can't see how.