The theory of buckling for thin-walled open-profile bars is criticized. Its several derivations are faulted for violating statics, using a variational theorem approximately, using an incorrect variational statement, and/or using an inconsistent filament representation of the bar. Significantly, the theory yields buckling loads that contradict engineering expectations. A theory to replace it with general equations for computing buckling loads is presented. A problem solved under the old and new theories shows how torsional buckling is viewed under the new.