Today I got the library working, and can insert a batch of (path, byte[]) pairs in the most IO efficient way, by building out a tree of byte[] indexed by paths, and then recursively reducing the tree. This means each byte[] is only written and hashed once, and only the updated nametables are update–also only being hashed/written once. Now “co library import” works, and I have stack.co being imported into the library. This is a huge milestone.