here are recent thoughts I posted to this list
- slicing a chunk: easier if an interface for the store is specified, to retrieve written data - optimizing a sequence that includes sliced chunks by reorganising their metadata. i think i included a number of optimizations in one of the attempts that might be copyable out. - selecting which nodes to merge when writing a new chunk so as to make balanced trees. i really simplified code for this in the flat_tree repo.
1. selecting which nodes to merge (xloem/flat_tree append structure, can be improved to work from middle) 2. slicing chunks (means the library must have access to old internodes to pull chunks out of them) 3. optimizing out redundant information when reusing old nodes
I'm pretty sure I memorized a next step. The next step I memorized is to add an interface for data access to xloem/flat_tree . I have some confusions associated with xloem/flat_tree. It makes sense to take it a little slowly.