I'm not sure you read what I wrote correctly. The technique is called "string interning" regardless of what exactly you're interning. In our case, we have a table assigning IDs to 32-byte bytea values.
(Also, to be pedantic, a Postgres BYTEA is a string; it's just what a programming language would call a "raw string" — i.e. a string of bytes. Postgres TEXT, meanwhile, is a string of characters, required to be valid in a given character encoding. The PG TEXT and BYTEA types are the same ADT with the same set of applicable operations. Of course, this puts the lie to the name "BYTEA" [i.e. "byte array"] — it's definitely not an array ADT of any kind. If it were, then you'd be able to use PG's array operations on it!)
(Also, to be pedantic, a Postgres BYTEA is a string; it's just what a programming language would call a "raw string" — i.e. a string of bytes. Postgres TEXT, meanwhile, is a string of characters, required to be valid in a given character encoding. The PG TEXT and BYTEA types are the same ADT with the same set of applicable operations. Of course, this puts the lie to the name "BYTEA" [i.e. "byte array"] — it's definitely not an array ADT of any kind. If it were, then you'd be able to use PG's array operations on it!)