Lisp Signatures...

But the protocol allowed me to debug the working code for get-hash, (setf get-hash), delete-hash, map-hash, against an in-memory version in just about 2 hours of expended labor from start to finish. Then once I was sure the in-memory version worked correctly, it took about 1 hour to get the external persistent version working. 

Component protocols, as in B-Trees were for components known as HashTable and HashTable-Bucket. The higher level code just assumed things about these two kinds of storage receptacles, like the ability to store keys in a bucket, and to find buckets from hashed key indexing into hash-table bins -- whatever those two things really are... They only have to behave externally in a specified manner. What they actually do, and how they do it, is of no concern to the get-hash, (setf get-hash), etc., code.

The only change to the code for external persistent hash-tables was an augmentation resulting from cooking up the implementation of the component protocol for persistent storage, to parallel that of the in-memory version. The actual get-hash, (setf get-hash), delete-hash, and map-hash code remained completely common, and fully debugged, to both versions.

This is a really nice way to work, without having to stand on your head attempting to force subclass relationships between wildly different beasts.