> >> Modified Files:
> >> 	ufs_vnops.c 
> >> Log Message:
> >> Move some permission-checking code for file owner/group changes up to
> >> vfs_syscalls.c::change_owner().  Also, always update the inode's change time
> >> if the operation succeeds.
> >
> >The first half of this change (moving the permissions check up to
> >vfs_syscalls.c) make things somewhat inconvenient for third-party
> >filesystems like AFS which have their own permission model.
> 
> Ouch, I didn't realize the implications of this.  I use AFS every day
> on NetBSD; doing this would really screw me (and others).
> 
> Can we please back this change out?

"Ignore my question."  I just carefully looked at the diffs, and now
understand how it can cause problems.

This does seem like a bad idea.  "Not all file systems are UFS."

Does this potentially break the NFS case, as well?  I mean, what
semantics do file systems mounted via NFS have w.r.t. changing
ownerhip and group?  Is that something that should be (or has to be)
set by the client, or is it something that's controlled by the server?
If the latter, why are we limiting it on the client?


cgd