Even if we have to go through the incubator, I'm convinced that adding
the JAMA codebase into the math library is the best option. IMHO, I'm
convinced that while the JAMA folks were very generous and open to
providing the codebase to the public domain, that further enhancing its
capabilities and providing any user support is not really in their
interest. It would be far more in our interest if we forked the codebase
and supported it. Any suggestion that the "JAMA folks" would have to
"agree" to this is not the nature of public domain, IMO reuse of public
domain doesn't require any such acknowledgment, though we should
liberally acknowledge their contribution wherever possible.
http://www.google.com/search?hl=en&lr=&client=firefox-a&rls=org.mozilla:en-US:official&oi=defmore&q=define:public+domain
http://math.nist.gov/javanumerics/jama/
*Copyright Notice* /This software is a cooperative product of The
MathWorks and the National Institute of Standards and Technology
(NIST) which has been released to the public domain. Neither The
MathWorks nor NIST assumes any responsibility whatsoever for its use
by other parties, and makes no guarantees, expressed or implied, about
its quality, reliability, or any other characteristic./
Note, JAMA is not a large codebase, and is in the public domain. As
such, does this really require the need for an "Incubator project"?
Thanks Phil for taking such an initiative.
My $0.02,
-Mark
Phil Steitz wrote:
I have gone back and forth a few times with one of the JAMA developers
at NIST (Bruce Miller) and legal-discuss and have finally asked the
ASF board for a ruling on whether or not we can start pulling in some
JAMA code. I wanted to start with the QR class, so we can use that to
get multiple regression implemented with decent numerics.
Before we start down this path, though I want to present some logical
alternatives and get others' reactions to them. Rather than pulling
in bits of implementation code, we could
1) Introduce a jar dependency on JAMA and wrap / extend / directly use
the stuff we need. For the QR use case, there will be no performance
/ efficiency hit from this. For some of the other RealMatrix uses,
this might be tricky.
2) Bring in the full code base. This would probably result in a side
trip through the incubator and might provide the occasion to get us
kicked out of commons (I am sure there are some who will be happy to
see us go :-) but that might be a good thing. I don't know yet if the
JAMA guys would go for this, am interested in opinions of others here.
If I get the go ahead from the board, we will need to decide between
0) fork selected stuff into [math] and pass bug fixes back and forth
with JAMA
and 1) or 2). WDYT?
Note that the same considerations apply to RngPack.
Phil
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
