gSoC Implement a better database schema

[Jeff Mitchell]

To start off with:  Ben is right, and Mark is right in that Ben is right. 
But a few specific comments below:

On Wed, 25 Mar 2009 19:45:58 -0400, Caleb Cushing <xenoterracide at gmail.com>
wrote:
> On Wed, Mar 25, 2009 at 7:03 PM, Ben K <mrmrpotatohead at gmail.com> wrote:
>> Fixing that bug doesn't require a schema change; you just write nulls
>> instead of zeros and change the handling of the two values.
> 
> as I recall on a previous occasion I was told that it would require
> more than that because there are lots of things expecting 0's instead
> of nulls.

Yes.  But pretty much all database handling takes place in a few files, and
given things like column names to grep for, this would really not take that
long.
 
>> The Amarok devs spent quite a lot of time coming up with new schema used
>> in
>> Amarok 2, and I suspect they'll be reluctant to change it.
> 
> I must have missed this one too?

It doesn't mean it's a good/great/perfect schema...just that it's better
than what we had in 1.4.  We didn't have an expert on hand.
 
>> There were also
>> quite specific reasons for moving to MySQLe and only MySQL backend - it
>> means far lower support costs than when you have potentially 3 different
>> backends (you don't have to manage three sets of interfaces).
> 
> I'm aware, but there are some of us who want to see (better) support
> for pg. I think we can live without sqlite (although all this FPIC
> stuff is more annoying to distro's)

So, of course, if you implement such a feature, you will support postgres
from here to eternity?

Amarok 1 saw a lot of releases with very buggy postfix support because none
of the devs were running it so no one knew if they had broken anything. 
We'd rather not support postgres than go back to that.
 
> well you might be right, it's entirely possible that the complexity of
> this issue has been previously exaggerated to me. tbh, I've no idea
> what it would take, I've looked at the schema, it'd be a couple of
> days to rewrite that, but none of the C++. 

You're assuming the schema needs to be rewritten.  I've not heard a
compelling argument why, and have heard compelling reasons why not.

> it's kind of a priority
> thing, and then I thought, maybe I could wrap all this stuff into one
> gsoc project.

No.  You'd be done in a week tops, not eight.  Sorry, but while fixes like
this are appreciated, this is far too simple/easy for GSoC material.

--Jeff