> Yes, yes, and yes ... but aside from the problem that you use the very

> ambiguous word "timestamp" (which somehow suggests using a "clock" of 
> some sort), isn't the "begin" timestamp of a long running transaction
imho a begin timestamp is near useless

> worse than the "commit" timestamp, when all its work got visible to
> outside world instantaneously?

This is one of the areas I am still worried about. Is one commit lamport
timestamp enough ? 
I think for some conflict resolutions we need to look at the
row level, and resolve conflicts per row and not per transaction 
(yes, this means that a tx might get partially replicated).

What I am trying to lead at is: maybe an infrastructure to produce
wieck lamport timestamps, that can be used in different places like
commit hooks and column defaults, would be of more general use. Maybe
a column could be a system column that is not visible with "select *"
for those cases where commit is not enough. And a commit hook could
insert it into clog like storage.


---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
       choose an index scan if your joining column's datatypes do not

Reply via email to