WoW lots of discussion on this issue.
So it was an oversight. If someone would have reminded us about 3580 then we would have just adopted that.
Since we didnt. It seems to me that a binary representation would be a more appropriate treatment for this value. The human presentation of the MAC address is just that for humans and RADIUS couldnt give care. Besides, the presentation of the MAC address has more then one format - one using ":" and one using "-". So it is best to leave the presentation to a presentation layer and not the RADIUS layer.
So 3580 IMO got it wrong. But as I said if one of you brought that up to our attention we would have fixed it to match 3580.
Which brings me to another point. It isn't busted. And changing it now will break backwards compatibility. And breaking backwards compatibility is being viewed as a very bad thing.
Actually if we want to open this up again, the I propose getting rid of the MAC address or encrypting it since it violates one of the WiMAX and of the IETF tenents which is identity hidding. We have the means to hide the user's identity but yet the MAC address is exposed.