On Sun, 2013-07-07 at 22:50 +0200, Dimitri Fontaine wrote:
> Would it make sense to review the local_preload_libraries semantics
> instead? It looks like session_preload_libraries only adds
> flexibility,
> and that current usages of local_preload_libraries would be served as
> well by the new setting.
>
> Did you figure out a case where a local_preload_libraries setting
> would
> stop working if made into a session_preload_libraries setting?
local_preload_libraries allows unprivileged users to preload whitelisted
libraries. session_preload_libraries allows superusers to preload any
library. It's hard to see how to consolidate those, at least without
adding another setting that whitelists the libraries somehow, at which
point you haven't really gained anything in terms of complexity.
I don't know of any actual legitimate uses of local_preload_libraries.
I recall that the plpgsql debugger was meant to use it, but doesn't
anymore. So it's hard to judge what to do about this, without any
actual use cases.