Hi,

> We have a program which does a lot of processing and just over 400
> queries. Using Wireshark we looked at the number of network packets
> transmitted during this time between our application and FB, via the
> jaybird driver.
> With FB 1.5.4, the number of packets transmitted is 4535. With FB
> 2.1.1 the number of packets is 6600.

Strange... I do not remember modifying wire protocol significantly
between 1.5.x and 2.x versions. I suspect that driver fetches more
metadata, this causes more packets. But this should not happen in normal
use as well - we tried to keep changes minimal, only when user asks
explicitly.

So, it would be great to have few junit test cases which run with 1.5.x
and 2.1.x versions of driver and execute your standard queries/behavior.
If you could prepare such, I could check exact reasons for more packets
and see what can be done to reduce it.

> Is this right, or have we done something wrong? From a doc re. 'remote
> interface improvements'
> (http://tracker.firebirdsql.org/browse/CORE-971) it suggested there
> may be 50% less packets, but there is almost 50% more.

As to the remote protocol improvements, they are not yet implemented in
Jaybird's pure Java protocol version and also won't be implemented in
2.2 release either. You still have an option use NATIVE version (i.e.
fbclient.dll/so) to see whether it changes things...

> We thought
> maybe we could try a jaybird 2.2 driver, but we cannot find one
> (alpha, pre-release, or whatever... the most recent one around appears
> to have been compiled in 2006??).

Jaybird 2.2 is almost ready for beta testing, but I need to give some
nail polish before release. So, if you wish, you could try version from
CVS directly. But again, it does not contain wire protocol improvements.

Roman