Re: sqrtm

[Dirk Laurie]

John W. Eaton wrote:
> 
> One way around the problem would be to ask the original author to
> release a version of his code under the GPL too.  That might work
> unless he has already signed an exclusive copyright assignment to the
> MathWorks.  In that case, we might be out of luck unless someone can
> implement the algorithm independently.
> 
The postscript version of the paper is on the author's home page
and the code is therefore accessible to anyone with accurate typing
skills or good OCR software.

> But that looks like a sticky situation to me.  The paper includes the
> old and the new sqrtm, both with MathWorks copyright notices.  The
> algorithm is described in the paper, but could anyone actually
> implement it without generating essentially the same code?  Would they
> sue for copyright infringement?  Who knows?
> 
The part of the routine that actually calcutes sqrtm(A) is very short
and very straightforward.  It implements an algorithm that was published
in 1987.  Certainly anyone programming that part will be generating
essentially the same code, but I can't see that this part of the code
is any more copyrightable than is
  k=i+i:n; A(k,k) = A(k,k) - (A(k,i)/A(i,i))*A(i,k)
as a step in Gaussian elimination.

The rest of the routine adds various MathWorksy frills and the mere
decision to have such code in the routine might be proprietary.

Dirk