Socket Connection Factory

[caleblloyd]

• Fixes Communication with TLS bus behind proxy results in SslPolicyErrors.RemoteCertificateNameMismatch #802
• Builds on top of Fix proxy with TLS scenario with SocketConnectionFactory #826
• Introduces public INatsSocketConnectionFactory, which is implemented by TcpConnection and WebSocketConnection or custom in a SocketConnectionFactory opt
• Introduces public INatsTlsUpgradeableSocketConnection which is implemented by TcpConnection or custom in a SocketConnectionFactory opt

[wolfkor]

This is more in line with my understanding of a factory, but it changes the API ;-). I tried it with an implementation of INatsSocketConnectionFactory and INatsTlsUpgradeSocketConnection, and it seems to work. But shouldn't it be called INatsProxyUpgradeSocketConnection?

[caleblloyd]

This is more in line with my understanding of a factory, but it changes the API ;-

Yes, it seemed more natural to use an interface that the internal and external implementations could both implement

Shouldn't it be called INatsProxyUpgradeSocketConnection

I don't think that it is specific to proxies, you could implement other kinds of sockets like Unix sockets too

[wolfkor]

but also not specific to TLS ;-)

[caleblloyd]

INatsSocketConnection is the one that is not-specefic to TLS, the library will not attempt to do a TLS upgrade on that socket.

But INatsTlsUpgradeableSocketConnection (which extends INatsSocketConnection) will cause the library to attempt to do a TLS upgrade if one is warranted.

[caleblloyd]

Tests are now passing but I think we should explore 2 more areas:

1. Pull Connection Timeout logic up a level so that each implementation doesn't have to handle it
2. Evaluate if we can simplify the INatsSocketConnection interface to get SignalDisconnected + WaitForClosed out of it

[caleblloyd]

I reduced INatsSocketConnection to SendAsync, ReceiveAsync, and DisposeAsync. Closing the socket is now done with DisposeAsync

[mtmk]

Awesome feature! LGTM thanks @caleblloyd