On Thu, Jan 13, 2005 at 03:57:04PM -0600, Scott Marlowe wrote:
> What cases are you thinking of? I've seen some very limited ones, like
Off the top of my head
- legacy application, closed, where you can't fix the source and can't
have larger than 32bit datatype, but you have another way to ensure
no dups.
- the queue-management item you mentioned.
- optimistic cases where a short search range is more important than
that a transaction doesn't fail on insert
- circular number spaces (xid uses this, after all)
> the time someone is thinking of doing so it's because a boss who doesn't
> get it wants a pretty list with no holes in the sequence or something
> equally silly.
Like they have some business problem they need solved, and doing it
this way is ugly but relatively cheap, and doing it the other way
means replacing 4 software systems and retraining 100 people. Is it
a pretty design? Probably not. Is it something that is, of all the
compromises available, the best one under the circumstances? I
dunno; I'd have to look at the circumstances. I think it's probably
usually a good idea to avoid this, sure, but I'm not willing to make
it a blanket statement.
A
--
Andrew Sullivan | ajs@crankycanuck.ca
The fact that technology doesn't work is no bar to success in the marketplace. --Philip Greenspun
