ovyalov added a comment.
In http://reviews.llvm.org/D14085#275442, @zturner wrote:
> I'm not sure I agree with this change. The interface to programming with
> socket is by definition a property of the Host operating system. It seems to
> me like Host was already the correct place for this
zturner added a comment.
I think most of the time they are used in a Connection class, but I don't think
it's necessarily guaranteed they will always be that way.
I can think of at least one use case in the future where we will need to open a
socket to a server we don't control and stream some
clayborg added a comment.
In http://reviews.llvm.org/D14085#275507, @zturner wrote:
> I think most of the time they are used in a Connection class, but I don't
> think it's necessarily guaranteed they will always be that way.
>
> I can think of at least one use case in the future where we will
zturner added a comment.
I still think implementation should be in Host as well. If the idea is to
simplify the creation scheme, then you could create an enum:
enum class SocketType {
Tcp,
Udp,
UnixDomain
};
and provide a static method on SocketBase like this:
class