[Coin-discuss] OsiCbcSolverInterface turn off log messages

[Lou Hafer]

Folks,

	A few additional comments re. OsiCbc and other OsiXXX solvers.
	
	I'm in agreement that OSI is not the best interface for MIP.  As others
have pointed out, if you want to do any serious tuning of the algorithm, OSI
simply lacks the hooks.  That said, if you don't care about tuning, and just
want to call osi.branchAndBound() and get an answer, OsiCbc has some attraction.
In theory, any OsiXXX could implement the osi.branchAndBound() method by calling
libCbc, but ...

	As far as getting OsiCbc (or libCbc, for that matter) to work with
solvers other than clp, well, it's an ongoing challenge.  Two factors come into
play:  How different is the underlying solver from clp, and how well has a
particular OsiXXX maintainer done in trmes of keeping up with the evolution of
OSI.

	One of the things that I work on within COIN is trying to maintain the
notion that you can change from one OsiXXX to another.  To get a quick
assessment of the state of an OsiXXX, you can try two tests:  first, does it
pass the OSI unitTest (use the debug build!).  Second, does an OsiCbc(OsiXXX)
object pass the OsiCbc unitTest.  If both of these work, then the OsiXXX solver
*might* work with libCbc.  If anyone is seriously interested in this, let me
know.  I've got a generic version of cbc that's at the alpha test stage, but
I've been letting other work distract me from pushing it up to the repository.
And we really do need folks to step forward and work on several of the OsiXXX
implementations.

	There are similar issues in terms of getting an OsiXXX to work with cut
generators that require OsiSimplex methods.

							Lou