Hi Frederic, thanks for your valuable feedback. I really appreciate this input compared to walking away silently.
I think your comments resonate with me, and looking at the discusions we have on the dev-list and in the JCR expert group I am convinced though that Jackrabbits (and JCR's) roadmap is well aligned with your enhancement requests.
- lack of administration possibilities: It is currently impossible in JR to modify an existing node type, to add/modify/remove properties. Refactoring is important for us, and impossible in Jackrabbit.
This is important to us as well. I think it is very important to mention that JSR-170 did not even specify a registration of nodetypes. JSR-283 proposes those exact administration features. [1]
- Strong constraints on the repository structure : we saw in the different mailing lists that Jackrabbit works (quite) well with a specific architecture, and that not following it induces very important and unacceptable performance loss (both for writing and searching);
While I would argue that you would have to follow certain constraints with respect to the datamodel in any data container to make it scale (be it an fs or an rdbms) I would agree with you that we still have a lot of room for performance improvements in the JCR space as a whole. Performance and scalability is always a tuning excercise though. I guess the Jackrabbit community is very interested in learning about your usecases and I am sure we will benefit from the understanding on how you would like to use the repository and learn how we can embrace those usecases.
- Search is also a (very) important feature for us, and currently Jackrabbit is much too limited in this area. SQL is not complete (well we don't need full SQL, but at least....joins...); and xpath is limited also, dereferencing is impossible or must be developed as an upper layer above Jackrabbit.
I am happy to report that the current draft of JSR-283 also addresses joins in the query language.
Some details also maybe, but most important problems are the three listed above (admin, repository structure and search). I think Jackrabbit is good to create a blogging system, a forum or any article-based simple website, but it is definitely not suitable for professional, generic CMS.
Since Day Software is shipping a JCR based WCMS that is in production for some of the worlds most high profile public corporate websites, I could not disagree more. Generally I think we are very aligned in our vision on how Jackrabbit should evolve, so I think that filing Jira issues for the above would be great. I am convinced that improving Jackrabbit (keep in mind, this is an Opensource project) could cover your needs much quicker than, building your own repository from scratch. I am certainly looking forward to your "Public Review" feedback on JSR-283 which is about to start within the next weeks. I will certainly keep this list posted as soon as we go into "Public Review". regards, david [1] http://jcp.org/en/jsr/detail?id=283
