Content addressing

Content-addressable storage means the bytes determine the key: an object is identified by what it contains, not by where it lives.

encoded bytes -> hash -> digest -> object identity

A store can therefore answer:

• is this object already present? (Exists)
• what is the canonical byte sequence for this value?
• can identical data dedupe automatically? (PutDedup)

Why it matters

• identical payloads dedupe naturally — same encoded bytes, same digest
• objects are immutable in practice: changing content changes the digest, there is no in-place update
• object identity stays stable across renames and storage moves
• the data model stays simple and easy to audit

In go-cask

Store.Put encodes the value with a Codec[T], wraps it in a small self-describing envelope (see object format), and hashes the whole envelope with the caller's Hasher. The digest — not a database row or a filename — is the object's identity. Different applications can layer JSON, binary, or custom representations on the same underlying model without changing how identity works.

This is the same conceptual model as Git's object store: content determines address, not a mutable pointer. go-cask keeps the same idea but stays generic about the object types on top.