Hi Michael,
On Wed, Sep 12, 2018 at 07:16:00PM +0200, Michael Andersen wrote:
My company is working on integrating some BTS (Nokia
Flexi ESMB) with Nokia
packet Abis with OsmoBSC.
this sounds very exciting! I'm really happy to read that somebody is working
on supporting some other BTS models / dialects.
The Nokia packet Abis seems to be Nokia OML/RSL over
SCTP/IUA (RFC 3057).
Interesting. Do you have protocol traces from a working setup with the Nokia
BSC?
My usual approach for implementing Abis protocol dialects in all of the Osmocom
history has been the following process:
* obtain as many protocol traces as possible (sometimes only 1 or 2) from a working
setup
* try to find vendor documentation on the Abis specifics (often not available)
* try to see if there are protocol analyzers that can decode the vendor specific
dialects (Nethawk, MA-10, Tektronix, ...)
* implement a wireshark dissector at least for those bits of the protocol that
are understood/known/standard
* try to re-implement the protocol from the Osmocom side
To get started, we have built a little stub that takes
the incoming SCTP/IUA
connection and carries out ASP_UP/ASP_ACTIVE handshake. After this handshake
is complete, we see OML coming in from the BTS.
Great! I hope you have seen libosmo-sigtran. It already implements SUA and M3UA,
and a lot of the infrastructure and the state machines would also workfor M2UA, M2PA,
and likely also for IUA.
The next step then would be to think about how to use this new Abis
within the context of libosmo-abis, which currently only deals with
T1/E1 and IPA/Osmo-stylae Abis/IP.
It seems that this OML is in a form that would be
understood by
"bts_nokia_site.c". So now we would like to continue the experiment by
feeding the OML/RSL into a BTS configured in OsmoBSC as type "nokia_site".
great!
1. Is it possible to configure a BTS of type
"nokia_site" to run Abis over a
unix domain socket ?
The unixsocket.c code in libosmoabis should be rather generic. I'm not
sure anyone has tried to use it with Nokia BTS model so far, but it *should* work.
2. Is there a better way to prototype this integration
?
See my above thoughts.
Whatever comes out of this and however far you get, I strongly
suggest that you publish all findings for the benefit of the wider
community. Thanks!
Regards,
Harald
--
- Harald Welte <laforge(a)gnumonks.org>
http://laforge.gnumonks.org/
============================================================================
"Privacy in residential applications is a desirable marketing option."
(ETSI EN 300 175-7 Ch. A6)