The triple key is not valid. From personal experience I was getting the SMS 
parts from all SMSC's not the one that transmitted the first part only. All the 
operators I worked with do the same thing. So the key should remain msisdn + 
refnum only.

Dimitris Evmorfopoulos

-----Original Message-----
From: Alexander Malysh [mailto:[EMAIL PROTECTED] 
Sent: Monday, January 08, 2007 4:01 PM
To: [email protected]
Subject: Re: [PATCH] Re: MO Concatenation

Hi Paul,

+1 for the at2 part.

-1 for the reassemble part. Your patch will never work reliable because  
not only msisdn + refnum should be considered.
Reassemble should have triple as key: SMSC, msisdn, refnum.
IMO bb_boxc.c is the wrong place for this, bb_smscconn suites better  
because bb_boxc just generic connection module for external boxes and  
should have nothing todo with SMS magic.


Am 08.01.2007, 05:28 Uhr, schrieb Paul Bagyenda <[EMAIL PROTECTED]>:

> Oops! Yes there was a small bug  once duplicate copies of the same
> message part were received (trying to debug-dump the sender address
> after free-ing the message structure). I've corrected this -- patch
> attached.
>
> I've also included a patch (also posted earlier) to the AT2 module
> that is still missing from CVS and is crucial: The way UDH is parsed/
> stored by the AT2 module is different from how other modules (at
> least SMPP) do it, which means that you see a different UDH byte
> sequence internally if you use AT2, from what you see with say SMPP.
> Without this patch too, the concat module would not work with AT2.
>
> Please test and put through some votes!
>
> P.
>



-- 
Thanks,
Alex








Reply via email to