I think this could potentially really reduce the amount of memory required - especially in cases where there is a lot of repetitive prefixes.

Would be interesting to try this out