In the short-duration/dialer market, you would be surprised at the
proportion of answered calls that were disconnected by the originating
party, not the terminating party. Often these are working-number or fax
machine probes, or failed "direct to voicemail" call forks, or Wangiri/ring
once schemes.
This is precisely due to the proportion of short-duration calls. But I agree
that it is a very stupid way to approach the problem, particularly when you
consider where most of the hangups in short duration calls actually come from.
Some of them do you come from the originator, in the case of AMD
I'm assuming you have a lot of short ALOC calls in your mix to your
vendor? You can do a delayed BYE with OpenSIPS, but it's a bit tricky,
as once you do the dialog match on the BYE, the dialog gets torn down
automatically by the OpenSIPs logic and the accounting is finalized.
In 2.2 I had to write
Eek. This sounds like a recipe for future problems with this carrier. If
you really want to try, you could do a sleep() or usleep() on BYE requests
in the section of your script that processes sequential requeststhe
has_totag() and loose_route() part (assuming no topology_hiding). That's
goi