[mythtv] Renewing PostgreSQL support

[f-myth-users at media.mit.edu]

    > Date: Tue, 22 Jun 2010 09:42:50 -0400
    > From: Daniel Kristjansson <danielk at cuymedia.net>

    > It should be fairly easy to patch in SQL access for development and
    > for research projects. There will still be a mysql instance or ten
    > in there after all, it just won't be exposed to the world.

Oh, I assumed that would be the case, but one of the issues there
might be requiring anyone who might need SQL access for any reason to
be building their own myths, hence losing some of the advantages of
package builds, at least for people who aren't making more-extensive
modifications.  Or we might see a PPA advertised (similar to JYA's at
the beginning of the VDPAU support) that's "Myth but with SQL access".

The -goal- of keeping the DB better-controlled (not only random writes
to it, but all the charset horribleness that, e.g., the Gentoo people
ran into, etc) is laudable; I wonder if there's a way to compromise.

I believe that some of the pushback we're seeing on this idea (besides
feelings of paternalism or TiVo-ization or whatever) may be analogous
to the issue that the linux kernel went through with whether or not to
allow closed-source kernel modules.  They had two choices:  disallow
them entirely (hence forcing people to build their own kernels and/or
fork entirely), or allow them, but mark the kernel tainted, so kernel
developers could (if they wished) say, "You've got stuff in your
kernel we don't know about, so we won't support your bug reports."

I think Myth could do likewise.  Make it relatively easy to allow raw
SQL access to the database (through the backend itself, so the single-
thread semantics are still true), but set a flag if it's used that
says the DB is now tainted in unpredictable ways.  Presumably the
listener wouldn't be enabled by default, but having a command-line
option (as opposed to recompiling) might enable it.  [Even fancier
might be to taint only if it's then -written to-, but that might be a
giant PITA to do right, given SQL, or to make the listener not have
write privs by default but taint if they're granted.]

The alternative might be to have people go around Myth anyway, but
just not tell you.  That's probably not much better...