Re: Open 7.3 issues - Mailing list pgsql-hackers

From Tom Lane
Subject Re: Open 7.3 issues
Date
Msg-id 24445.1029691157@sss.pgh.pa.us
Whole thread Raw
In response to Re: Open 7.3 issues  ("Nigel J. Andrews" <nandrews@investsystems.co.uk>)
List pgsql-hackers
"Nigel J. Andrews" <nandrews@investsystems.co.uk> writes:
> Daft question but isn't this an administrator's issue?

The feature wasn't going to change; the argument was just about whether
to change the factory-default permissions mask for the socket.  An admin
could override the default in any case (and probably should, in many
cases).

The argument for 0700 was that it is "secure", the argument against
was that it's useless.  No one sane runs their apps as the postgres
user, therefore you'd always be having to open it up anyway.

We thought a little about 0770 but the problem is that the actual
behavior is extremely context-dependent.  On some Unixes this is
just what you want, on others it's little different from 0777.
In any case depending on group membership to control database access
means that the DBA *must* have root privileges, which is not an
assumption we like to make.

So unless someone comes up with a new idea, we're sticking with 0777
for backwards-compatibility reasons.  But certainly socket permissions
ought to be on the short list of things to think about during every
database install.
        regards, tom lane


pgsql-hackers by date:

Previous
From: Vince Vielhaber
Date:
Subject: Re: Open 7.3 items
Next
From: Joe Conway
Date:
Subject: Re: Remove implicit unique index creation on SERIAL columns?