On Wed, 2008-03-26 at 20:33 +0100, Leslie P. Polzer wrote: > I suppose this is a good opportunity for me to chime in with > a few thoughts. Aren't B+trees a choice that is too conservative > for a modern storage backend? > There seem to be more modern data structures (S(b) trees or > fractal trees) that are especially well suited for storing > variable-length keys.
Perhaps. I would certainly hope that our design would allow such structures to be swapped out as a "strategy" pattern. Personally, I had never heard of those structures until you mentioned them, and a quick search does not yield a concise description of their advantages---or of how to implement them. If you can briefly describe the differences that would be great. On the other other hand, a LISP-native back end using B+-trees would be a nice leap forward for us; using something better might be even better but would not lift us into a new valence band of quality. The fact that one can find example implementations, possibly even in LISP, of the B+-tree, is an advantage. My personal philosophy is gradualism --- crawling is the best way to learn to walk, and having a B+-tree implementation gets us half the way to the latest data structure. > > Leslie > > _______________________________________________ > elephant-devel site list > elephant-devel@common-lisp.net > http://common-lisp.net/mailman/listinfo/elephant-devel _______________________________________________ elephant-devel site list elephant-devel@common-lisp.net http://common-lisp.net/mailman/listinfo/elephant-devel
