CVS update: samba/source/nsswitch

Volker.Lendecke at SerNet.DE Volker.Lendecke at SerNet.DE
Tue Mar 16 20:57:57 GMT 2004


On Sun, Mar 14, 2004 at 11:25:10PM +0000, Andrew Bartlett wrote:
> The main argument I have for it is convention.  Now, this isn't
> partiuclarly strong, but I would suggest that users expect that if
> they modify 'idmap uid', that it only affects users (be that domain
> users, or some similar concept like the free RIDs being used here).
> Likewise, if would be good if the same applied to idmap gid - it only
> affects groups, and group rids.

My arguments for just incrementing the RID HWM is simplicity and the fact that
NT does it that way. When doing this right some time in the future we will end
up with some kind of distributed RID allocation algorithm, probably simply
handing out slices of free RIDs to masters. Having to cope with different uid
and gid spaces only complicates that unnecessarily IMO.

This only applies to the situation where winbind is running. There is exactly
one reason for running winbind: We have to cope with non-algorithmic id
mapping, in the simple non-vampired DC with only one domain we can perfectly
fine do purely algorithmic mapping and do not have to care about RID
allocation.

Volker
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: not available
Url : http://lists.samba.org/archive/samba-technical/attachments/20040316/7f9bab17/attachment.bin


More information about the samba-technical mailing list