Stephan Szabo <sszabo@megazone.bigpanda.com> writes:
> On Thu, 22 Apr 2004, Development - multi.art.studio wrote:
>> so in 7.4 seq-scan is about '1' (ms or what?) slower than 7.1
> The cost numbers from explain have no direct connection to real time.
... or at least, if your concern is poor choice of query plan, it's
folly to treat them as being equivalent to real time. Actual runtimes
are the things to compare.
Also, I wouldn't care to bet that estimated costs from 7.1 are very
comparable to 7.4, anyway. We rejiggered the cost models enough in
between that it's quite likely to be a meaningless comparison.
regards, tom lane