do we actually need libosmo-mgcp as shared library?
I'm wondering why we are actually creating + installed a libosmo-mgcp shared library? Isn't it only used by osmo-mgw? In that case, it would be best to simply have it as an internal, non-installed static library (like libbsc.a in osmo-bsc/openbsc) instead of installing it into the system. This means there's no API/ABI compatibility to maintain.
For the client library (libosmo-mgcp-client), this is of course completely different, and it must be installed and have a stable API/ABI.
If there are some smaller utility functions used by e.g. osmo-bsc_nat (which uses -legacy at the moment anyway), then we should try to abstract those out and have only those in the shared library, leaving most mgw code private to the osmo-mgw repository.
Please share your thoughts/feedback, particularly neels + dexter.
On Sun, Nov 19, 2017 at 11:24:19AM +0000, neels [REDMINE] wrote:
one other user is the osmo-bsc_nat, which I presumed would also move to libosmo-mgcp at some point. If
osmo-bsc_nat will stay at libosmo-legacy-mgcp, we could indeed move libosmo-mgcp firmly into osmo-mgw.
I think it is not yet clear what will happen to osmo-bsc_nat at this point. I would argue that once that is decided, we can introduce a libosmo-mgcp shared library, but we should drop it until we introduce any application (beyond osmo-mgw) actually using it