## The problem I was trying to solve During my time at Karnaphuli Online Limited, I worked with SIP-based voice services on networks where bandwidth was an operational constraint. Replacing all the SIP endpoints and downstream gateways was not the objective. The useful question was whether the constrained portion of the path could carry calls more efficiently while the existing SIP-facing equipment stayed in place. ## The arrangement I developed The approach accepted calls at a SIP-facing gateway, transported them over an IAX2 connection across the bandwidth-limited segment and returned them to SIP at the receiving side. This kept the local SIP interfaces familiar to the surrounding infrastructure while changing how calls crossed the restricted link. I investigated gateway configuration, codec choices, call connectivity and the bandwidth associated with the alternative paths. ## What the work established My historical account records approximately 29–32 kbps per call in the original arrangement and 9.6–10 kbps in the optimised test arrangement. Those figures suggest roughly three times better bandwidth efficiency under the conditions tested. They should not be read as a universal SIP-versus-IAX2 result: codec, packetisation and transport overhead all matter. The original monitoring records are no longer available for a fresh audit of the exact method or test conditions. ## What I took from the project The lasting lesson was to identify precisely which section of a system is constrained before changing its architecture. An improvement on a narrow network link can be valuable even when everything outside that link continues to use the same interfaces. I see the result as an applied engineering optimisation, not an invention of a new protocol.