ForwarderDevice and StreamerDevice not working as expected

[OJ]

I've been looking to use these two devices in an application and I believe the current implementations are incorrect.

The ForwarderDevice is part of the PUB-SUB pattern, which means that the front-end should be a SUB socket and the back-end a PUB socket. This is correct as it currently stands. However, when setting up the ForwarderDevice the constructor currently calls Bind() on the front-end socket when I believe it should instead be calling Connect(), since a SUB socket should be reaching out to the publish source (not expecting the publish source to connect to it). Here's what it looks like now:

    public ForwarderDevice(ZmqContext context, string frontendBindAddr, string backendBindAddr)
        : this(context)
    {
        FrontendSetup.Bind(frontendBindAddr);
        BackendSetup.Bind(backendBindAddr);
    }

    public ForwarderDevice(ZmqContext context, string frontendBindAddr, string backendBindAddr, DeviceMode mode)
        : this(context, mode)
    {
        FrontendSetup.Bind(frontendBindAddr);
        BackendSetup.Bind(backendBindAddr);
    }

And this is what I believe it should look like instead:

    public ForwarderDevice(ZmqContext context, string frontendBindAddr, string backendBindAddr)
        : this(context)
    {
        FrontendSetup.Connect(frontendBindAddr);
        BackendSetup.Bind(backendBindAddr);
    }

    public ForwarderDevice(ZmqContext context, string frontendBindAddr, string backendBindAddr, DeviceMode mode)
        : this(context, mode)
    {
        FrontendSetup.Connect(frontendBindAddr);
        BackendSetup.Bind(backendBindAddr);
    }

The StreamerDevice suffers the same issue, but in the other direction. That is, the back-end should instead be calling Connect() instead of Bind().

Is this a bug, or is this my lack of understanding that's causing issues? I can see in very early versions of clrzmq that the ForwarderDevice actually had the correct semantics so I believe that this is a regression.

Thanks.

PS. I'd have already fixed it and submitted a pull request, but the way the tests are set up for the devices is bound heavily to both ends calling Bind(), resulting in threads assuming a certain order of processing. It's a pain in the ass to unravel in such a way that it makes it generic for all tests. A bit of refactoring is needed here.