questions about the future of websocket

[mmmray]

I am trying to understand what the path forward is with regard to CDN setups, so I am summarizing what I know here before I dive into questions.

1. WebSocket connections (not just tunnels, also legitimate ones) are detectable via ALPN, because 1) cloudflare does not advertise support for websockets over http2 2) therefore chrome always initiates with http/1.1 alpn, even if there is an existing connection. This is a known issue, can't find the link right now.
2. there is speculation that up/down flow patterns are observed, similar to TLS-in-TLS but rather "websocket-in-tls" windows httpupgrade panic #3128 (comment)
3. because of this, there are successor protocols for ws, called grpc and httpupgrade

there are still some reasons why i want to keep using websocket for now:

1. client support is good
2. there are very few CDNs that support grpc. there are some CDNs that support websocket but try to parse/reencode the websocket messages (therefore httpupgrade transport fails) -- websocket/httpupgrade has the advantage over grpc of being proxyable through many shared webhosts via htaccess. sometimes even domain fronting works. (not all versions of Apache, needs enabled mod_proxy with ws protocol support)
3. particularly in iran, during sensitive times is when websocket and CDN-related setups become relevant, but iran has shown that it is willing to just terminate all TCP sessions after N seconds and block all UDP, making websocket perform the same as httpupgrade/grpc
4. [entirely theoretical and perhaps not practically relevant] websocket is a browser protocol, so there is more legitimate traffic of it. even if websocket connections are perfectly detected (via websocket-in-tls or alpn), it is more expensive for the censor to block all websocket connections than it is to block all grpc.
5. [again theoretical] websocket is a browser protocol therefore perfect parrotting via browser dialer is possible

some questions:

1. does this make sense?
2. there are some up/down traffic patterns mentioned in tickets that might make websocket detectable, and particularly distinguish xray websockets from legitimate websockets. is this considered inherent to websocket connections, and that's why websocket is considered a dead end, or is there potential to improve camouflaging here?
3. is browser dialer something that was developed in a rush in response to censorship events (because it was suspected that utls is faulty), or will it be further developed?
4. is anybody using browser dialer to get around blocks today?

[mmmray]

previous train of thought: net4people/bbs#345