After the upgrade from 2.8 to 2.9 the IP packing no longer works. What is the reason for that?
check config, maybe there is something not fully converted after upgrade ?
there is new feature in 2.9, which enables commpression for wireless interfaces ‘interface wireless set wlan1 commpression=yes’
set compression=yes ← this is for hardware compression, with ip packing mt make software compression
Did you ever get a technical answer to this? I’ve been trying to test M3P on 2.9 but can’t see any difference between when it’s enabled or not 
try to do a bandwidth test with small UDP packets and compare the results with and without M3P.
Tried that (see http://forum.mikrotik.com//viewtopic.php?t=4969) If I use 100byte packets I see no difference if packing/unpacking is set to simple or none.
Interestingly enough I just roled back to 2.8.28 and it works fine.
may be i do the same thing 
(v3.22)
I just tested it on a Nstreme WDS link. I tried udp test, tcp test and udp test with 100 bytes packets, in one direction, in both directions. I saw no difference. I enabled ip packing compress all/compress all on the WDS Bridge interface and on the Wireless interface at the same time on the AP and on the wireless interface on the client.
p.s. according to the manual http://www.mikrotik.com/testdocs/ros/3.0/interface/m3p.php settings are advertised via MikroTik Neighbor Discovery Protocol, and those packets are sent out from time to time, and I tested too fast. This time I will enable packing, wait for 2 minutes for routers to see each others’ MNDP packets and test again lol.
p.s. 2 I see it re-advertises via MNDP as soon as I change packing settings, this time I enabled the packing on the bridge interface where I see the MNDP is recieved in Neighbors. Testing now…
and result:
Um isn’t NStreme already packing packets into larger ones?
It should be. But there is still change in the screenshot I show. I guess I could capture traffic with the internal sniffer for analysis. Have other things to do now so not soon. If at all.
Compress-all is pretty CPU hungry. What boards are you using?
RB433AH. CPU usage was even lower or did not change. Difference was so minimal. Maybe avg pkt size, when UDP 100 byte test was off, was pretty large.