Simon Riggs <simon@2ndQuadrant.com> wrote:
> Do we need table-level predicate locks at all? What would they give
> us? Why not just go straight for fine-grained page-level locks?
I don't want to get too far into implementation discussions at this
phase (see Tom's slides ;-)), but suffice it to say that a table scan
can cover more pages than we'd want to track individually....
The coursest possible resolution allows proof of concept. Tests can
be written that work at that level which should not break as
finer-grained locks are implemented. (See how I'm drawing from
another presentation? ;-))
-Kevin