Think about extreme forms of server-side rendered applications where you need to keep the client device as simple, cheap and locked-down as possible.
All you need are 2 streams: 1 from client-to-server w/ all touch/keyboard events, and another from server-to-client with JPEG frames. One could easily build client devices using FPGAs or other integrated technologies with this sort of architecture.
The only caveats are latency & bandwidth requirements, but there are strategies for managing this, as well as problem domains where sufficient constraints exist on both counts (i.e. secure local networks).
All you need are 2 streams: 1 from client-to-server w/ all touch/keyboard events, and another from server-to-client with JPEG frames. One could easily build client devices using FPGAs or other integrated technologies with this sort of architecture.
The only caveats are latency & bandwidth requirements, but there are strategies for managing this, as well as problem domains where sufficient constraints exist on both counts (i.e. secure local networks).