Currently PR 4576 combines a bug fix and a new feature.

1. Removes wrong check and adds a verbose clarifying comment. This part should be backported to maintained versions.

2. Adds multiple explicit "return None". This is a minor improvement the quality of the code. Too minor for separate PR.

3. Makes methods for getting the address of the real network card preferring universally administered addresses and falling back to locally  administered address only if universally administered addresses are not found. This is a new feature. But are there computers that have both universally and locally administered addresses (and with locally administered addresses enumerated first)?