[Coin-discuss] setColSolution

[Lou Hafer]

Folks,

	Suppose I was trying to construct an application and wanted the
ability to use either an exterior point or interior point solver. It strikes
me that they might want significantly different starting solutions, and in
fact an application might need separate chunks of code to generate them. So
why not provide this capability in OSI? Create a new method,
inventInitialSolution, which derives an initial solution appropriate for the
solver, by whatever means the implementor feels is best. It still has to
satisfy our `reasonableness' criteria --- within bounds, whatever else we
impose. This might be the best way to maintain solver independence, and it
would avoid the problems associated with forcing an exterior point solver to
accept an interior solution, and vice versa.

							Lou