Instead of calling XInternAtom once per atom, call XInternAtoms once
instead. This is both simpler code and should reduce network
traffic. Although the latter is probably not measurable.