bug 225221 "longdescs table needs a primary key"

[Christopher Hicks]

On Tue, 9 Mar 2004, Gervase Markham wrote:
> I think calling that an "obvious one-line patch" is rather far from the
> truth. Jouni, bbaetz and I all had comments - which means that the right
> thing to do wasn't actually obvious.

Obviousness is obviously in the eyes of the beholder.

> When told that there was need for more than the one-liner, you said: 
> "Pushing all of this onto somebody trying to get their first simple 
> patch through seems a very effective way to keep the database structure 
> dorked." Who would you have do the necessary additional work if not you?

The point was that the additional work was needed for extending this to
fix other bugs.  The fix for this bug I believe is the one I've suggested.  
If people want to extend it to help fix other bugs, that's fine.  What I 
put forward is a necessary prerequisite for any of those extensions.  You 
need some way to uniquely refer to rows to update them which implies a 
primary key which leads to my straight forward patch.

Given that all of this is brought up in a conversation about code quality,
who ever reviewed things to allow a table to be created without a primary
key in the first place?  This violates such a fundamental rule of database
design and maintenance it's difficult to believe it occurred in the first
place and it's even more difficult to believe that it has persisted for so
long with noone objecting.

> On the other hand, I should have paid the bug more attention and given 
> more help. Mea culpa - I just didn't have time. Apologies for that.

Life happens.

-- 
</chris>

No, no, you're not thinking, you're just being logical.
-Niels Bohr, physicist (1885-1962)