Milan Miladinovic

Removed `enableWebSocketCompression` `bool` from both client and server: https://github.com/capnproto/capnproto/pull/1505/commits/b66dac90e501d401d822e5359a984a65f0aa2267 Set `AUTOMATIC_COMPRESSION` in tests: https://github.com/capnproto/capnproto/pull/1505/commits/fd9b618719fa05a870c0935313f0e3b7bd0d9b76 Going to rebase, then reorder commits so the compression modes is one of the earlier commits.

>Note that we need the same enum on the client side. In automatic mode, KJ should automatically offer the extensions it supports. In manual mode, the application needs to pass...

> > Automatic: take the Sec-WebSocket-Extensions header given to us by the caller and run it through the checks from the previous PR before sending off the validated extensions to...

> Same on the server side: In automatic mode, `acceptWebSocket()` should ignore the `Sec-WebSocket-Extensions` header passed from the application and always send back a header based on what the client...

> The answer to this might be to set manual mode, but also have KJ provide a separate function that can be called to ask it to choose the appropriate...

> Is there anything else you planned to do before merging? Not on the KJ side, no. > Did you want to test the BFINAL handling in some way? We...

> > I think we can leave the fetch() implementation as is, since you could easily just set the extension header yourself: > > Naive question, but I figure most...

> > The scariest thing to me is that we'd be going from a world where we have no compression to a world where (potentially) all websocket communication is compressed....

Since compression is still behind a compat flag, could we merge this PR as is? If the (existing) compat flag isn't provided, we strip the header right after anyways. We...

> Exciting! Let me know if/when you'd like me to take a look at this. There's still 1 failing test (`ActorCache list stream cancellation`) but definitely feel free to read...