after talking some more with greg oster, i've gone with a suggestion
from him to not bump the version, since the new field will be zero
in all existing labels already, so it can be safely always included
for now. the only problem we envision is an old kernel seeing a new
large set and cutting it off. i can't think of data loss situations
for this, just access errors (which is what i've seen happen.)
updated patches (i haven't merged any block/sector macros in yet):
http://www.netbsd.org/~mrg/rf-2tb.diff
http://www.netbsd.org/~mrg/rf-2tb-netbsd-5.diff
.mrg.
