My idea is to make it possible for web applications working on slower connections to fallback on alternative smaller asset versions(like images, videos, etc.). HTTP Live Streaming has similar behavior but for video streams.
At first I thought, this can be achieved by using HTTP/2 server push. The web server would send a response, measure the goodput and push appropriate asset versions before closing the connection. But this would eliminate the advantage of the server push because the client can start downloading the assets only after receiving the page body. Since assets are not served like streams, the server should detect goodput before sending the HTTP response to the client.
I wanted also to check if I can get throughput from TCP stack, for example values from the slow-start. But I guess it would be available only upon sending the response.
Does anybody know how it is possible for a web server to measure goodput before sending a response?
