I understand the differences between cluster and database and pg_dumpall =
and pg_dump.
In my opinion a pg_dump of a database should pack all informations of =
the application (the database) in the dumpfile in one container, to be =
able to restore it full functional at a different place.
Because the search_path is a crucial information for the application to =
work correctly (like any other object inside the database) it should be =
packed into this container called pg_dump dumpfile.
This should be independent of the current implementation, where we store =
the search_path in a cluster record: The informatation belongs =
semantically to the content of the database, even if it is stored =
elsewhere.
My concern with promoting this suggestion is to avoid trouble in =
emergency cases, logical consistency, intuitive usage of pg_dump and =
smooth experience for non-expert people.
Hans B.
