I found out this paper while I was checking out Postgres source code [1]. This is used for generating hash table of postgres keywords, where the hash function is produced on compile time (?) and used in that hash table - if I got everything right.
I discovered this paper in the late 90s. It changed the nature of my work for the next decade or so. An O(n) algorithm for generating MPH functions seemed like a miracle, at the time. Being able to solve "impossible" problems (at the time) gave me access to many opportunities.
minimal perfect hash functions are really underrated