Split of tr1 and std random facilities and completion of latter

[Paolo Carlini]

Benjamin Kosnik wrote:
> Yes.
>
> Your suggestions certainly seems reasonable. I didn't realize that the
> API changes post-N2798 were this level of significance, and that
> continued churn on this would be such an issue.
>   
Note that, irrespective of the new C++0x specifications, we really want
to reconsider the various choices for the algorithms, basing on the
feedback we got for the tr1 implementation and new literature to which
we have access.
> What is your preferred outcome for C++0x random support in 4.4.0?
>   
My preference, at this point, very close to the branching, is definitely
doing nothing, or at least not committing any large change. Simply, we
don't have the time to converge on a stable, well thought, set of
changes. Actually, I believe the same policy (essentially, fixes for
regressions and not-extensive - i.e., requiring invasive changes - bugs
only) should hold from now on for the rest of the library, certainly I
will adhere to it for my personal work.

Paolo.