To release Samba 4.0 'as is'

Matthieu Patou mat at
Tue Nov 29 05:11:34 MST 2011

Hi Tridge,
On 23/11/2011 04:32, Andrew Tridgell wrote:
>> Without proper design and code to implement the source3 fileserver,
>> winbind and nmbd integration we simply don't have a finished
>> product. The integrated fileserver is a MUST HAVE for a 4.0 release.
> Integrating the file server portion when you are an AD DC isn't the
> biggest task in this list. We have had demonstrations of that for a
> while, although the methods Andrew and I chose to use to implement it
> weren't liked by some people so they never made it into master.
Is there still concerns about this ?
If so why not working at solving them together right now instead of waiting.

In a couple of other emails it's stated that the S4's file server is 
good enough and so we can afford to ship it like this.
So it's true that it works not so bad, and I have not so much concerns 
about the fileserver than the winbind.
But the horrible truth is that more than once I had application using a 
server share for the storage and those application were working with 
samba4's fileserver but not so good (and sometime very slow) and when I 
moved them on a small samba 3.5 server they started to perform much better.

I didn't investigate much as more than 2 years ago at SDC it was stated 
that we will use the s3 fileserver as the default fileserver for the 
Samba AD.

Also I find the logic of saying "well we need to release a stable for 
Samba AD soon so let's postpone the integration of fileserver for 4.1" 
very dangerous.
Because it seems that I understood that we will do the integration for 
4.0 and now that 4.0 is coming close we say "oh next time I promised" 
but we can understand that some won't buy it as we didn't honor our 
previous promise, at least it would be the case for me.
I have the impression that for 4.1 a discussion on the mutualisation of 
the epmd code will arise and maybe also some other components, following 
your talk it seems that important amount of work will be needed on one 
of the winbind implementation so if we also add smbd we are pretty sure 
that the integration won't happen.

> The biggest and hardest task is winbind, and the basic problem is that
> the s3 winbind is just not well suited for a AD DC, especially in a
> multi-domain forest.
But for the 4.0 release we don't target multi-domain forest so why 
bother ? Maybe in the long term the current s3's winbind is not the best 
solution but what about the short term ?


Matthieu Patou
Samba Team

More information about the samba-technical mailing list