At 3:37 PM +0100 3/15/07, Giuliano Gavazzi wrote:
You could try filing a radar on this, but I suspect that the
behavior's correct.
done, and I agree with you, although I there is some basis in
saying that users should be identified by name alone, as it is
done at the ordinary unix permission level.
If you agree with me that this is the expected behavior why on
earth would you file a radar report it it's doing what we just said
we expect it should???
Users are *not* identified by names. Ever. Anywhere.
But I consider the ACL issue with rsync to have significant
philosophical complications no matter how you slice it. Yet these
are the most understandable and sane of rysnc's issues in Tiger.
I this this issue stems from:
uuid_to_name in file_cmds-116.9/ls/print.c
and in turn from mbr_uuid_to_id (see man page).
Also, nicl . -read /users|groups/whatever will show (for some
users|groups) something called generateduid, which might have to
do with the above.
I did another test: I used an ACL that referred to a group (admin
for instance). It transferred fine. In this case the generateduid
is the same on both (presumably all) systems. I also created a new
group (this one without generateduid) and it also worked.
No, not at all.
Don't pretend to "guess", either read the Internals book and/or the
code. It's quite clear how users are identified. And nicl deals
with netinfo.
So this is not an issue as long as:
1) one restores on the original system (that might not be
possible, but one should also backup all the relevant system files)
or
2) one uses only group based ACLs (which is a good idea).
Unfortunately there seem to be no way to specify a group acl when
there is also a user by the same name. If somebody knows how to
work around that please let me know.
It's not an issue at all, it's just you have a lot of woolly
thinking going on.
--
