mcinp(at)o2(dot)pl writes:
> If you create a GiST index on a geometry(Polygon,4267) column and do updates
> on the table, the index size grows enormously even if the indexed column
> contains only nulls.
I don't see any comparable misbehavior when testing a GIST index on a
plain polygon column full of nulls, so I would say this is specific to the
geometry opclass. Which means you need to tell the PostGIS folk about it,
not us.
regards, tom lane