`solve` should also accept real `A` and `b` about intervallinearalgebra.jl HOT 5 CLOSED

Based on today's discussion, this is not a good idea, as it's quite likely that also other packages define solve(A, b), even using CommonSolve.jl this could be a problem as it would lead to methods redefinition.

A better way would prob be to define IntervalLinearProblem and call solve on that.

from intervallinearalgebra.jl.

That would actually be nice because one could have

DeterminedIntervalProblem <: AbstractIntervalLinearProblem
OverdeterminedIntervalProblem <: AbstractIntervalLinearProblem

and could then use solve to dispatch on solvers for determined and overdetermined problems (should the latter be added to the package at some point).

from intervallinearalgebra.jl.

Maybe LinearSolver.jl could be a good interface to start

from intervallinearalgebra.jl.

even using CommonSolve.jl this could be a problem as it would lead to methods redefinition.

note that LinearSolver.jl uses SciMLBase which uses CommonSolve, so there should be no conflict.

from intervallinearalgebra.jl.

epsilon_inflation is meant exactly for this

from intervallinearalgebra.jl.