To be honest I think the rationale is a bit flawed. As it was explained
to me by Sprint's reps, this is how they insure that load on their
server farm remains balanced. If 1 server is getting hammered, when we
reconnect we'll be connecting to a different server (LB behind same IP).
It would of course have been much easier for them to simply start
issuing unbind's in these cases, but try explaining logic to a company
that size. In any case this is a requirement for Sprint (CDMA), and I
(and anyone else wishing to connect to them) have to oblige. 

On another note, this work could lead to a fully non-persistent mode
with a bit more tweaking, however that's beyond the scope of my
requirements, and unfortunately I don't have time to pursue it right
now. 

Josh McAllister


Alexander Malysh wrote:
> Hi,
> 
> it would be interresting to know the rationale of this patch 
> better to say requirements for XXX SMSC?
> 
> Josh McAllister wrote:
> 
> > This patch adds the ability to set smpp connection idle limits, and 
> > max submits per connection. This is a requirement for Sprint's SMSC.
> > 
> > Adds 2 smsc group config directives:
> > sms-max-submits
> > and
> > sms-max-idle
> > 
> > The patch does not change existing behavior if these are 
> not set, or 
> > set to 0. If sms-max-submits is set, it will shutdown the 
> connection 
> > (unbind / wait for unbind_resp / etc) once we've submitted 
> > sms-max-submits messages, and immediately re-connect. Likewise, 
> > sms-max-idle will shutdown and immediately reconnect if we have 
> > submitted no messages in > sms-max-idle seconds.
> > 
> > Note: the test for max-idle will not be executed while 
> we're sitting 
> > around doing nothing, so make sure that 
> enquire-link-interval is set 
> > to some number < max seconds you can be late disconnecting. 
> In other 
> > words, if sms-max-idle is 30 seconds, and 
> enquire-link-interval is 30 
> > seconds, it could be 1 minute before we disconnect.
> > 
> > This is my first patch submission so please be gentle... 
> but of course 
> > feedback is appreciated.
> > 
> > Thanks,
> > 
> > Josh McAllister
> 
> --
> Thanks,
> Alex
> 
> 
> 
> 

Reply via email to