Code to replace the unexpected.tdb file

[Andrew Bartlett]

On Thu, 2011-01-06 at 01:36 +0100, Volker Lendecke wrote:
> On Wed, Jan 05, 2011 at 03:56:37PM -0700, idra at samba.org wrote:
> > Thanks Volker,
> > this code looks really interesting, although I have to say that /tmp is really
> > not a good place for something like these pipes.
> > 
> > In Fedora/RHEL we moved the winbindd pipes to /var since a long time already, and
> > at least Fedora is trying to get into a state where /tmp can be mounted into name
> > spaces that are private to each user.
> > 
> > So in the long term using /tmp to share data between processes run by different
> > users is not a good idea.
> > 
> > /var/lib/samba or similar is a better choice for this kind of thing.
> 
> So we end up with winbind with its pipe in /tmp and nmbd
> with its (VERY similar in concept) pipe hidden somewhere
> else? 

Yes, just like the NTP signd pipe (which Samba4 provides).  

> Is this a consistent view?

The difference here comes down to difference between services provided
to un-configured libraries (because nss_winbindd needs to
self-bootstrap, and we only support one user list per system), and those
provided to other parts of Samba. 

Also, while X11 (and to some extent winbindd) can claim to be
'grandfathered' because they are precursors to the FHS or it's general
enforcement, I just don't want to create more problems for others in
future.  For example:
https://bugzilla.redhat.com/show_bug.cgi?id=503181

Given this and other problems I mentioned in my other mail, such as a
need to manually configure this to be elsewhere for 'make test', do you
really think it is that important that this pipe be placed in /tmp?

Andrew Bartlett

-- 
Andrew Bartlett                                http://samba.org/~abartlet/
Authentication Developer, Samba Team           http://samba.org
Samba Developer, Cisco Inc.