As mentioned previously, it maybe be an improvement for usability or performance or both to offer different request and response types in the client vs server.

The justification for this is that there are some operations that are unique to these types in different scenarios. As one example: the server's copy of a request would include information about the peer (client), whereas the client's view of its request would not - instead that kind of information would be available from the response.

Note that interceptor support does require some operations that would not otherwise be expected. E.g. you might expect a request in the client to only need to set headers, but an interceptor on the client may need to read the headers as well. So the number of things we are able to do differently on the two sides may not be very significant.