Hi Neels, On Mon, Jul 17, 2017 at 10:09:27PM +0200, Neels Hofmeyr wrote: what you are observing is the de-facto situation for the NITB for many years. Unfortunately nobody has investigated this thoroughly and implemented a solution. In classic/commercial GSM networks a call is always transcoded, so basically at the level between MSCs there is PCM audio, and somewhere (typically already at the BSC) the voice is transcoded for each of the two call legs. There's a convoluted mechanism called "tandem free operation" (TFO) in which the two transcoders can use the LSB of the PCM audio samples to implement a "covert channel" by which they can detect their presence and simultaneously switch both transcoders off. This is to avoid the speech quality degradation that happens by transcoding twice, particularly with HRv1. None of the above is implemented, as we take a "we don't transcode in OsmoNITB" attitude. The problem (or rather difficulty) with that is that in a mobile-to-mobile call, we don't yet know the least common denominator of the codecs supported by both the MO call leg and the MT call leg until very late (after having paged the subscriber, authenticated him and performed the first couple of CC message exchange). Please note that we actually need to take the codec restrictions of all elements into consideration, i.e. * capabilities of the phone(s) * capabilities of the BTS(s) * capabilities + policy of the BSC(s) * capabilities + policy of the MSC(s) Where it becomes even more interesting in the new post-NITB world is once we have calls between two subscribers at different MSCs, where the codec capabilities / negotiation will have to be forwarded between them e.g. by means of SIP :) "It just needs to be implemented" There should be open tickets about this for about a decade now. Sadly it seems nobody has had a significant enough interest or need to actually get it done. I'm not sure why the BSC would ever say something like that. The BSC normally simply picks a codec from the codecs that the MSC permits for the given call. The MSC should be in charge of requesting a specific codec if it wants a specific one. -- - 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)

On Tue, Jul 18, 2017 at 11:58:58AM +0200, Neels Hofmeyr wrote: I completely understand that there is a point to having a policy that can be configured at the BSC to restrict its supported codecs for whatever local policy/configuration reason. This makes complete sense. I however don't understand why we should enforce that this policy does not include both a restriction for which codecs to use on TCH/H and also which codecs to use on TCH/F. Holger, do you remember? -- - 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)