Why not implement slices (as in FreeBSD), as a first step towards wegdes?

AFAIR there's never been any real implementation of wedges; discussion
on public mailing lists of how they should work gets hopelessly bogged
down in feature-creep and the downsides of naming via a userspace daemon.

Do slices first, get some experience. Then we could gather data to
compare wedge implementations: e.g., in-kernel, numbering of
extended-partition wedges, giving the same mapping Linux/i386 kernels
use), versus a fully configurable userspace-daemon approach.