deprecate ignoreUncleanSSLShutdown: Bool parameter

[weissi]

We should probably deprecate the ignoreUncleanSSLShutdown parameter. AsyncHTTPClient knows when it can and when it cannot ignore an unclean SSL shutdown.

Basically if an .end was received and we have either a content-length OR a transfer-encoding: chunked header from the server, we can safely ignore. For HTTP/2 we can always ignore.

If we do have neither a content-length, nor a transfer-encoding: chunked header, then this should error.

[weissi]

@Lukasa does this sound right?

[Lukasa]

Probably, with the caveat that in principle we may want users to be able to tolerate even the last case if they're sure they should. But frankly I think deprecating sounds good.

[weissi]

I agree with that. My thinking was that EOF framing over TLS should be super rare (am I right here?), and even rarer without CloseNotify, so I don't think that's a thing we need to support. And if we do then we can add a new one with a more scary name :). ignoreTrulyInsecureUncleanTLSShutdowns: Bool :)

[Lukasa]

I believe it should be super rare, yes. As noted, I’m happy to deprecate.

[weissi]

For more context: #238 contains the full explanation why we don't need this parameter.