On Thu, Jun 13, 2002 at 04:51:54PM -0400, Alan Langford wrote:
> At 2002/06/13 15:16 -0500, cmpilato@collab.net wrote:
> >It seems kinda Band-aid(tm)-ish to me, but only because of the
> >redundancy aspect (we *can* derive that information, it just takes
> >time proportional to the size of the changes made to any given
> >revision). But if a Band-aid(tm) is what we need, then it's what we
> >need.
Warning: twisted world-view rears its ugly head.
I think this is better than a band-aid.
I think this is better than carefully reasoned denormalization.
I think this is a glimpse into the real schema:
It's the rest of the data-model that is the denormalization.
Think about a database for a bank account. You usually have a base record,
with things like customer name (or id) and address (or address id), etc.
One of the things in there, for example, might be "account balance".
Now, you also have the transaction history. This is critical. You CANNOT
lose the transaction history, or you're the worst bank ever. It is the
real data, not the silly "account balance" field in the customer record.
These other things that we store, like the undeltified contents of the
HEAD, *those* are the denormalization!
The only thing that is real is the transaction history.
> >Wah-lah, constant-time answers to "What changed in this revision?"
>
> That's voilą. ;)
>
I presumed that he was kidding.
Maybe you're kidding; maybe you knew that he was kidding!
--ben
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri Jun 14 01:45:46 2002