Something which might be good to add: order documents based on name before serializing them. So when reading a binary we can assume a binary search is valid and speed up access to document without requiring an explicit index.
Except, for smaller archive sizes this is hardly going to improve performance (actually binary search is possibly slower).
Still, ordering documents is "for free" and any decision on best indexing strategy can be just postponed really.
Originally posted by @KaruroChori in #13
Originally posted by @KaruroChori in #13