Remote data sources look as if they can be treated as tables, and in fact, they are represented as such by Foreign Data Wrappers. Unfortunately, in practice, this doesn't work in all the ways you might hope and expect.
However, by writing your queries and code in the standard way, you give the database usable context information about what you are trying to achieve; future PostgreSQL versions might achieve better optimization on the same SQL code. This is a general advantage over custom solutions, which are usually opaque to the server and thus cannot be optimized further.
Ideally, we would like to use foreign tables interchangeably with local tables, with minimum possible performance penalty and maintenance cost, so it is ...
