jmcd at us.ibm.com
Thu Jun 15 20:08:32 GMT 2006
> First, is that really you who wrote code? ;-))
Yep, hehe, been a long time...and I'm required to at least attribute what
didn't come from me, even if I have to review and copyright it.
> Second, if you want to build a real infrastructure around
> the BUILD_FARM_HACK I added why not give Jerry the late
> triumph that the winbind local users/groups were the right
> thing in the end? I'm really not sure about this, but as
> winbind is doing something in that direction anyway, we
> could implement a short-cut in smbd to circumvent nss and go
> directly to winbind which then could do the whole nss thing
> for us.
Well, part of the idea was to have something that was reusable in
samba4...code that is fairly independent. I'm not sure I'm 100% up on the
idea of making winbind critical for testing smbd. I'll have to think
about that...depends on what the goal of testing it is. Perhaps there is
more than one nss wrapper that we'd want to do.
>Then third, how would you represent groups in your
Perhaps we would prefix each file with 'user' or 'group'.
>Or store the whole passwd info like primary
We can store stuff inside the files, I just wanted to be able to search on
both users and uids without opening each file.
> A TDB to me would look a lot more logical.
I thought about a TDB, but metze convinced me that it should be manageable
from the setup scripts, and unless we write another utility for this part
(I don't think the existing tdbtools are a decent fit for this), it
wouldn't be setup from the script.
Any solution that is not winbind-based will be fairly trivial to
implement; this is just one potential idea (this was my interpretation of
the discussion with metze).
IBM Linux Technology Center
6 Minuteman Drive
Scarborough, ME 04074
jmcd at us dot ibm dot com
jmcd at samba dot org
IBM tie-line: 349-5335
More information about the samba-technical