At 11:32 PM 4/11/00 +1000, Jan Mikkelsen wrote:

>
>
>More overflow pages than necessary and unused space are pretty much the same
>thing.
>

Not really. Overflow space is ignored during an exhaustive scan.
Unused space is wasted disk i/o.

>An alternative might be using conventional data pages as overflow pages,
>along with a better way for the engine to know how much free space is
>available on data pages.
>

IDMS (and it's PDP-11 derivative) kept space allocation pages that
reflected the percentage fill of the data pages. Almost every data
page update required an update to the space management page. In
a careful write environment, this would be a disaster.

Most records really aren't that big given run length encoding to
eliminate trailing blanks and unused fields and blobs to partition
out the really big stuff. Given the wholesome trend towards larger
page sizes, lost sleep can better be applied to higher payback
problems like increasing the granularity of threads.

Jim Starkey