The main answer is because we try to keep things consistent across
database types and there are some databases that have max key sizes
larger than the author + id + filename values can be.
Originally there was a PK but we were running into problems where
people needed a longer path and/or author field and there was no way
to increase that due to the PK constraints.
Since we really only do a full table scan anyway, it was better to
drop the PK and increase the column sizes than to keep the PK that is
rarely to never used and be either stuck with smaller columns or have
different table definitions per database type.
You can always manually add a primary key if you like, Liquibase
will not care.
Nathan