Hi, On Wed, Nov 6, 2013 at 5:31 AM, Tommaso Teofili <[email protected]> wrote: > out of curiosity, I was wondering if there's any guideline other than > following the API on writing a new MK implementation. > What one should start with, design concepts, experience on known dos, > donts, caveats, etc.
The node state documentation in [1] should give you an idea of how higher levels of Oak look at and access the underlying data structures. Looking at just the MicroKernel API can easily lead you to design something that doesn't work too well in practice, like we experienced with the original MongoMK codebase. Also, the SegmentMK [2] and MongoMK [3] design documents describe two existing approaches. They highlight most of the key issues to be considered. [1] http://jackrabbit.apache.org/oak/docs/nodestate.html [2] http://jackrabbit.apache.org/oak/docs/segmentmk.html [3] http://jackrabbit.apache.org/oak/docs/mongomk.html BR, Jukka Zitting
