Plans for pdb_ads and auth_netlogond?

[Andrew Bartlett]

On Tue, 2012-06-19 at 17:13 +0200, Kai Blin wrote:
> On 2012-06-19 15:08, Andrew Bartlett wrote:
> 
> Hi Andrew,
> 
> > - At best, they duplicate the supported, working and tested solution. 
> > 
> > - We should not release, even as a developer feature, code which is duplicate, 
> >   untested and which we do not wish to support. 
> 
> I don't recall a conclusive discussion on this. Arguably we do have a
> lot of duplicate components that we're currently shipping. They might
> not fulfil your criteria as well as pdb_ads, but probably many things
> get close.

I'm happy to talk about any of these you can nominate.  I'm serious:  We
shouldn't be shipping code that meets the full set of the above.  A good
example (only receives partial testing) is anything that is left of the
Samba4 AD-on-external-LDAP backend, which we agreed to remove when
inconvenient.  We only test some of the provision start-up code at this
point, and it clutters the provision command line.

> > Given that in the last year when we have largely solidified the
> > design of Samba 4.0, and that in that year there hasn't been any changes
> > (beyond those imposed by broader changes), or a way to automatically
> > test the code, it seemed reasonable to ask, and to act on the
> > affirmative response from the maintainer.
> 
> "defeatist" would be what I'd call that response, not affirmative. But
> that's a separate issue.
> 
> Anyway, I'd argue that if people still want that code, why not keep it
> in the tree? We can still axe it from the release branch if we decide we
> don't want to release with it.

The main issue I have with that approach is that we get to have this
whole debate again at release-branch time, and not because this code is
particularly good, or solves major user issues, but because it stands in
memorial to a different world that in difference circumstances might
have been.  

But in truth it is an implementation approach that didn't stand up to
the full demands of the 4.0 release.  Should we keep very other
incomplete, incorrect, untested and duplicate implementation for some
part of Samba, just because the code exists?

I should note that I wouldn't have removed the code if this debate had
sprung up in this form before I autobuilt the changes.  In perfect
hindsight of course I would have avoided the urgency of the debate by
delaying my action.  But as we are having this discussion, I would like
someone to stand up and say how these issues should be addressed (eg by
adding a testsuite, as has been possible now for 12 months). 

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