Re: Patch: add timing of buffer I/O requests

On Sat, Apr 14, 2012 at 3:27 AM, Greg Smith <greg@2ndquadrant.com> wrote:
> On 04/13/2012 06:22 PM, Tom Lane wrote:
>>
>> But (a) I *don't* want to seriously break things, and don't see a need
>> to; (b) interval is expensive and has got its own problems, notably an
>> internal limitation to usec resolution that we would not be able to get
>> rid of easily.
>
> A straight float seems pretty future proof compared to a usec resolution
> interval.  Jim was commenting in the same direction I already did, that ns
> resolution is not impossible to see coming.
>
> I also expect to compute plenty of derived statistics from these numbers.
>  Interval math is good enough that I'm sure such things could be done, but
> it seems odd to start with those units.  I appreciate that the interval type
> has a nice purist feel to it.  My pragmatic side says we're going to pay
> overhead to create in that type, only to find people end up converting it
> right back to other types for easier math tricks.

I'm still rooting for numeric.  As somebody said upthread, performance
ain't critical here; and that lets us whack around the internal
representation however we like without worrying about it.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company