On Mon, Jul 18, 2011 at 3:28 PM, ktm@rice.edu <ktm@rice.edu> wrote:
> On Mon, Jul 18, 2011 at 03:12:20PM -0400, Tom Lane wrote:
>> Heikki Linnakangas <heikki.linnakangas@enterprisedb.com> writes:
>> > On 18.07.2011 18:32, Tom Lane wrote:
>> >> Hmm. Well, it's not too late to rethink the WaitLatch API, if we think
>> >> that that might be a significant limitation.
>>
>> > Right, we can easily change the timeout argument to be in milliseconds
>> > instead of microseconds.
>>
>> On the whole I'd be more worried about giving up the shorter waits than
>> the longer ones --- it's not too hard to imagine using submillisecond
>> timeouts in the future, as machines get faster. If we really wanted to
>> fix this, I think we need to move to a wider datatype.
>>
>> regards, tom lane
>>
>
> You could also tag the high bit to allow you to encode larger ranges
> by having microseconds for the values with the high bit = 0 and use
> milliseconds for the values with the high bit = 1. Then you could
> have the best of both without punching up the width of the datatype.
Considering that we're just talking about a function call here, and
not something that ever goes out to disk, that seems like entirely too
much notational complexity.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company