We discussed this trend some days ago with Torben, and came up with several points:

• Midgard 1.x is starting to show age. The framework was designed back in 1998
• Code quality in 1.6.0 alphas and in some major modules like Aegir is really bad
• The competition is getting tougher
• Installation is still horribly difficult
  

However, even despite of these, we decided there is no reason to get too alarmed. In experienced hands, Midgard 1.x is still fairly competitive, and the work being done on Midgard 2.x and MidCOM 2 is promising, especially with the EU support. The recent focus on usability should also help.

Also, if we get 1.6.0 stabilized in this weekend's Poznan meeting, things should look a lot brighter, with bunch of new things:

• Apache 2.x support
  
• Integrated PAM authentication support
• UTF-8 as default encoding
• Multilingual content support by default
• MidCOM and Aegir bundled as default
• Datagard, the new database installer

Of course, it should be noted that SecuritySpace survey isn't the whole truth on CMS market shares. (thanks for reminder, Alexander!)

Updated 13:30: Tarjei gave his list of issues:

• There are too many administrative user interfaces, but they all share the same weaknesses. Solutions could be:
  
• Unify the UIs to Aegir 2 and MidCOM 2
• Add WebDAV support to make development more file-based
• Aegir* and Nemein* libraries are too hard to access, making it difficult to build on top of them. Solution:
• Start using PHPdoc for documenting the libraries and get an unified backend OO framework
• Demand PEAR coding standards
  
• Midgard does not support read-level access controls
• Midgard must get closer to the PEAR and PHP communities
• Midgard lacks a package format that would enable deleting packages