In a heterogeneous architecture of diameter stack where the actual stack runs on a network front-end machine, with diameter application (eg., HSS) running on the back-end machines. So far this front-end network stack was stateful i.e. if it had sent an outgoing request, it knew to which back-end machine the response has to be routed to.
This state-full property brings-in some issues and thus we're planning on to make the network front-end stack stateless. For this, we need the ability to have some information to be put in the request with the guarantee that the same will be returned in the response - basically the state information. For stateless agents, RFC already defines the Proxy-Info AVP for storing and retrieval for their state information. There has been suggestion to use the same for a new outgoing - not forwarded - requests. We know that AVP name itself implies that only agents can utilize them. But, can this be possible to use this AVP for new outgoing request to store the stateless information? Is this practiced anywhere? How do the popular implementation of stack/diameter application deal with a 'Proxy-Info' AVP without let's say a Route-Record AVP - i.e. Only if Route-Record AVP is present, it means it has passed through an agent?