[Matahari] Matahari Project status

Zane Bitter zbitter at redhat.com
Mon Sep 3 12:42:19 UTC 2012


As will be evident to those of you who have been subscribed to this list 
for a while, I am not working on Matahari and nor is anybody else at Red 
Hat. There are no plans to put any more work into the project.

The primary reason for this is the lack of end-to-end authentication in 
the Matahari architecture. Conversations with colleagues and potential 
users alike have repeatedly shown that the access control and auditing 
capabilities that this would allow are essential to adoption. However, 
the effort required to retrofit it into QMF is prohibitive.

As I understand it (this was well before my time), the project began as 
a proof-of-concept for an RPC messaging system, presumably with the idea 
that a security model could be bolted on later. Based on lessons learned 
from Matahari, I would recommend that anybody undertaking a similar 
endeavour begin with a proof-of-concept of a security model that is easy 
to configure and bolt on RPC later. Security is the _core_ problem in a 
remote management system.

Sadly, I am not aware of any remote management frameworks that replicate 
the bus architecture of Matahari. Thanks largely to the Web, we live in 
very much a client-server world, and it may be some years before the 
pendulum begins to swing back. In the meantime, I expect that system 
configuration tools (such as Puppet and Chef) that rely on pull rather 
than push will continue to be prominent.

Fedora users may wish to checkout cura: https://fedorahosted.org/cura/

If anybody wishes to continue development on Matahari they are, of 
course, quite welcome to do so (subject to the existing GPLv2+ licence), 
and I will be happy to assist with the transition. The mailing list will 
remain available until I get sick of dealing with the daily spam 
filtering ;)

regards,
Zane.


More information about the Matahari mailing list