Re: Best way to index IP data?

On Fri, Jan 11, 2008 at 04:32:05PM -0500, Michael Stone wrote:
>On Fri, Jan 11, 2008 at 03:19:35PM -0500, D'Arcy J.M. Cain wrote:
>>Besides, there are many cases where you want to track both ipv4 and
>>ipv6 for the same purpose and requiring two different fields would be
>>less than ideal.
>
>And, there are many cases where you don't. I've got two kinds of db's
>that have IPs in them. In some, the IP is a small part of a table which
>is focused on something else. For those I use inet, which provides a
>nice bit of future-proofing. In other db's the IPs are the primary
>focus. There are lots and lots of IPs, and the space used by IPs may be
>the largest chunk of a particular table. For those tables, I don't use
>inet because the overhead really is a significant fraction of the space.

Oh, yeah, the latter type also has seperate IPv4 and IPv6 tables,
because there's no point in bloating 99% of the data for the 1% that's
IPv6. Is that a niche requirement? Maybe--but I think that storing
netmasks is even *more* of a niche...

I'm not arguing for the removal of inet, but I do think there's room for
more than one type--and I certainly think its nuts to pretend that inet
can meet every requirement well.

Mike Stone