Hannu Krosing wrote:
> ...is storing all tuple visibility info in a separate file.
>
> At first it seems to just add overhead, but for lots (most ? ) usecases
> the separately stored visibility should be highly compressible, so for
> example for bulk-loaded tables you could end up with one bit per page
> saying that all tuples on this page are visible.
Seems you could do "one bit per page" compression even with
visibility data stored in the pages themselves.
I could imagine a table re-writing "vacuum freeze" that finds
pages with all data visible to everyone and packs them with
a single bit saying "everything here's frozen".
The penalty would be an expensive splitting of the page (and
who knows what evil locks would be needed) if an update is
ever done on those wholly frozen pages -- but we're talking
about read-mostly tables here so that tradeoff might not be bad.