On 11/05/10 03:31, Jim Starkey wrote:

> If anyone is planning to screw around with encryption, may I recommend a
> formalism that I developed a awhile back that may help.

Looks like it's as designed here that transforms are objects of single
use - in constructor initial data in some form is passed to transform,
then we get() from transform result of it's job, and after it transform
can be only destroyed. Does it mean that on practice transform
constructors are so fast that we can ignore possible performance issues
with it?