I totally agree as its mixed feelings.
I can totally understand the need to simplify and reduce the support complexity of Zentyal.
I have been the source of much controversy and a big pusher to drop S4sync.
The thing is I don't think you need to drop modules to reduce the complexity of Zentyal.
The main problem is the focus of the "Box" and the idea of the box is a little long in the tooth with the current market.
It shouldn't be box it should be container or VM or several boxes. Hardware is cheap, support is not.
I am a big fan of LXC because of its low resources but it could be a full hypervisor such as KVM.
If zentyal had a base network layer that supported virtualisation where a box is partitioned into containers then all these problems with complexity would vanish.
I have been playing with the Samba4 binaries because I was very aware of the loading and employment of the Zentyal design caused many problems.
Personally I think it is design and much can be done with the Base networking and Samba4.
Still we are stuck with a full Bind9 implementation and the purpose built internal DNS of Samba is left out in the cold that only provides disadvantages.
Going back to our Eric
Perfection (in design) is achieved not when there is nothing more to add, but rather when there is nothing more to take away. (Attributed to Antoine de Saint-Exupéry)
We already have a DNS server installed, the one we use is purely a legacy consideration as Zentyal has reused what they had.
The replication of the internal Samba4 DNS works 100%, tried tested and done on my test bed sourceforge project
http://sourceforge.net/projects/samba4all/The only thing that I am finding is that with 4.1.6 and server side copies is that disk permission problems start to occur.
I am waiting a couple of days for the new 4.1.7 binaries and I will be pushing another rake of Virtualbox OVA's for Debian and Ubuntu (14.04) being quite important.
They are only for demonstration as they are not automated in the way Zentyal is, for production use you will have to get down to the dirty CLI.
What they do demonstrate is how light Samba4 is and how it makes User/Group and DNS replication childs play.
Samba4 makes it very easy to federate servers of specific function whilst keeping to the structure of domain security.
With virtualisation LXC or KVM I can partition complexity into containers so that simplicity is gained without any loss of function. Or simply split the functionality over a couple of low cost pieces of hardware.
The arguments with OpenLdap, S4sync, Split DNS and various others have saddened me slightly as they became very polarised into yes and no camps and it got very personal and full of individual agenda's.
What saddened me was I always thought it was very possible to support all. It still is and could be if there was some form of open discussion, but apparently that is not part of opensource.
I still feel a standard LDAP can be coupled to samba4 as a proxy and provide everything the 2.0 series had.
I believe Asterisk can be brought back online and various modules could remain and be supported.
I have said it before, that a modular product such as Zentyal is losing revenue by being forced into a monolithic support option.
There is no feedback to what is of worth to the community with a monolithic offering. Its an insurance farce where for many of us smaller more tailored systems would return more value.
The current commercial support subscriptions forces me to use Zentyal as an All-In-One which is something I will never do.
The result of this is that I am starting to use Zentyal less and less.
My servers don't have the simple common web admin console which is a real shame. I have them separated into area's of common functionality.
Gateway, Web and Email, File server is my basic three server setup nowadays. It makes the install and configuration much easier.
They are still on the same box but by being in containers they are isolated and much stronger for it.
Again I have been saddened at this polarised argument that seems to be short sighted greed that I am trying to get things for less. Whilst my argument has been the opposite and about increasing adoption and about more.
Both in infrastructure and subscriptions making things modular would only increase stability and revenue.
Maybe one day