L402: streaming response gRPC calls don't support creating L402 token

[guggero]

For some reason, when the server sends back the 402 payment required response code if the user doesn't already have an L402 token, the client interceptor code doesn't detect that properly and doesn't attempt to create a token, resulting in a terminal error.

It is yet unclear where exactly the problem lies, the following possibilities come to mind:

• The server's response status fields aren't set correctly for the client interceptor to recognize:

  aperture/proxy/proxy.go

  Line 466 in b1b0dc5

  func sendDirectResponse(w http.ResponseWriter, r *http.Request,

  . Perhaps the gRPC/HTTP/2 specific trailer fields need to be used?
• The client library just doesn't look for the response code in the right place (it interprets the error from the stream interceptor, but perhaps it needs to look at response header/trailer fields?)
• This is a limitation of the gRPC Golang library itself and cannot be fixed easily.

In any case, it should be pretty straightforward to set up a new unit test (taking

aperture/proxy/proxy_test.go

Line 322 in b1b0dc5

func TestProxyGRPC(t *testing.T) {

as a template) that executes the streaming case and allows for local debugging of the issue.

Workarounds

The current workaround in most Lightning Labs application is to make sure we call a non-streaming (unary) gRPC method first, which then triggers the L402 token creation and allows subsequent streaming gRPC methods to work.