Hi all,
I have been scanning the thread from Jan Vicherek who is having heavy
browsing activity blocking updates.
Basically there should be no need for read locks in postgresql,
since the before images are in the table until the next vacuum.
Therefore a scheme could be implemented, where readers
always read data as it was at the time of their begin work statement.
(The Oracle way)
An other alternative would be to implement a:
set isolation to {dirty|uncommited} read;
This enables readers willing to take the risc of reading uncommitted
data
to read without a read lock. This is often very useful. (The Informix
and DB/2 way)
In the long run I think one of the above should be implemented (I prefer
the second,
as it is less resource consuming, and easier to handle (e.g. with
indexes)).
Andreas