Re: rcond, condest, and a block 1-norm estimator

[Jason Riedy]

And David Bateman writes:
>
> 1) Relicense under the GPL. I kept the original license as a comment,
> but perhaps that should be removed as well.

> + ##     * Neither the name of the University of California, Berkeley nor the
> + ##       names of its contributors may be used to endorse or promote 
> products
> + ##       derived from this software without specific prior written 
> permission.

I'm pretty sure the above chunk should remain...  It's an
additional restriction explicitly allowed under the GPLv3.
Obviously the combined work can be GPL because the licenses are
compatible, but that's not really a relicensing so much as an
implication of the individual licenses (inclusion in the build
and test process is more than mere aggregation).  But I'm only a
math/CS guy, not a lawyer (thankfully).

IIRC, the explicit LAPACK license has a similar non-endorsement
clause.  For a long, long time, LAPACK had a "fuzzy" license...
We finally made it explicit and simple.  But people still ask how
LAPACK is licensed.  sigh.  Even more fun explaining that parts
of the reference BLAS now are in the public domain (pre '78).

> 4) Change the name of block_onenorm_est to onenormest as something
> shorter. I hesitated to make this an internal function, but thought that
> a 1-norm estimator is useful in its own right.

I mostly use the 1-norm estimator for my work, so I'd hate for it
to be "internal".

> 5) Replace the block of code [...]
> [L, U, P, Pc] = splu (A);

Cool, didn't know about that one.  I mostly use a hacked-up
TAUCS, but I need out-of-core (and software extended precision).

> Are you happy with the proposed changes? Can you suggest any further
> changes?

Sounds great.  Thanks for all the work!  Now we just wait for
users to find bugs. ;)

Jason