Chapman Flack <c...@anastigmatix.net> writes:
> But let me return to the earlier idea for a moment: are you saying
> that it might *not* be sufficient to find an applicable snapshot at
> the time of constructing the object, and register that snapshot
> on TopTransactionResourceOwner?
The problem is to know which snapshot is applicable; if the transaction
has more than one, you don't know which was used to read the row of
interest. I suppose you could be conservative and use the oldest one,
if snapmgr lets you find that.
regards, tom lane
