there have been several documented problems over the years with using these addresses with certain OSes, and it's fairly sure that they will continue to have problems. One can also make a reasonable case that it's worth the administrative benefits to not use these addresses.

The only subset of documented problems I'd agree with is trying to supernet class C's into larger blocks. But that wasn't what I was talking about. It's a shame that we went through so much trouble to remove classfulness and artificial byte boundaries in address space only to have them be reinserted and carried forth forever with ad hoc policy like this. I understand your point about tradeoffs of administrative benefits, but in this case it seems particularly ill-advised to promulgate brain- damage and justify it by saying "well, broken hosts didn't get fixed fast enough, so we may as well continue the tradition." /cvk