To be truly effective, a CMS being used to store files for a DITA project must be aware of the tree-structued ("web") nature of the project, the content contained in the source files, and the relationships among the files and their content. The CMS must also be able to report on this kind of information: both meta information about the files, and syntactic and semantic information about the content. Ask your CMS vendor about the product's content analytical capabilities before deciding to buy.
It can also be useful for the CMS to include DITA Open Toolkit processing, and debugging and reporting aids that operate along with file processing. CMSs should also provide basic library (source control) functionality.