I'll look at this in the weekend an report later...
Craig Ringer wrote:
> sent/received from client ing PGStream.
>
> Can't you extend or wrap the pg_input / pg_output streams to do the
> accounting in their read(...) / write(...) method instead? The way
> you've done it is rather fragile and is very prone to bitrot as other
> parts of the driver are changed.
--
Best regards,
Alexander Pyhalov,
system administrator of Computer Center of South Federal University