[Asterisk-Dev] Possible bug in DUNDi

[William Waites]

There seems to be a bug with the way the Asterisk implementation
of DUNDi (from CVS-HEAD-05/12/05) deals with transaction IDs.

The internet draft specifies a transaction ID as being a 16 bit
value. It appears that Asterisk only pays attention to the first
15 bits.

In the dump below, 192.168.0.1 is running an new implementation
of DUNDi that is under development, 192.168.0.2 is running 
Asterisk.

Notice that the source transaction ID is a random number in the
first packet which happens to be 0xBC0A.

The reply contains 0x3C0A, that is 0xBC0A with the top bit
stripped off.

Looks like there is a cast to a signed short or something similar
happening somewhere in the Asterisk code.

00:20:01.048177 192.168.0.1.1894 > 192.168.0.2.4520: udp 8
0x0000   4500 0024 5a49 0000 4011 1bbd c0a8 0001        E..$ZI.. at .....{.
0x0010   c0a8 0002 0766 11a8 0010 1cf2 bc0a 0000        H....f..........
0x0020   0000 0900 0000 0000 0000 0000 0000             ..............
00:20:01.115952 192.168.0.2.4520 > 192.168.0.1.1894: udp 8 (DF)
0x0000   4500 0024 0045 4000 3111 44c1 c0a8 0002        E..$.E at .1.D.H...
0x0010   c0a8 7b06 11a8 0766 0010 0306 15ec 3c0a        ..{....f......<.
0x0020   0100 8c00 5555 5555 5555 5555 5555             ....UUUUUUUUUU

-- 
William Waites, Consulting Technologist
Consultants Ars Informatica S.A.R.F.
ww at magicphone.ca / +1 416 848 1527 x514
                   +1 514 963 4096 (Direct)