Fetch Sizes and Blocks

Hi,

I have been trying to read the code and understand how the results
fetching works in ODBC. It looks like by default, all tuples are
pre-fetched before they are returned to the client. Unless if you
specify

UseDeclareFetch=1
Fetch=1000

then it will use cursors to fetch 1000 rows at a time.

My question is at network layer, does the driver fetch individual row at
a time from server (for pre-fetch or cursor), or does it use some kind
"bulk" fetch so that it minimizes the network traffic? I also did not
see "Fetch" value coming across as connection string at server side,
when it is set.

PS: we wrote a PG wire protocol over http://teiid.org and want improve
the performance.

Thank you.

Ramesh..