I'm going to try to finish off the Cursor implementation for the JdbmPartition in my private branch in the couple weeks before ApacheCon. Incidentally this work will not be included in the 1.5.2 release we're planning. I hope to incorporate these changes into 1.5.3.
I had started this work a couple weeks ago, and found enough time this weekend to document what will hopefully be the end result. The documentation shows exactly how the JdbmPartition is designed. It also shows how optimized search operations are conducted on Partitions implementing the indexing scheme used by any BTree based Partition. Here's the doco:
I'd love feedback from others if they have the time or interest to read over this material. My technical writing skills are pretty shabby so I hope I don't confuse people. If so, I will correct this ASAP as people find parts of the document difficult to understand. This part of the server is probably the most critical and complex part, and it's perhaps the last area where we had a complete lack of documentation. For these reasons I really want to make this documentation as easy to read and understand as possible.