[omniORB] tcp transport wrapper

[Renzo Tomaselli]

Duncan,

> I don't really want to change anything like that before 4.0.1, since
> it involves an addition to a semi-public interface, which it's best to
> avoid in a micro release. Also, I don't think it's a good idea to
> expose the idea of a socket descriptor number in the giopEndpoint,
> since it is meant to abstract away from such details.

Ok, in the mean time I will clone the tcp transport. In any case we got the
chance of dynamic transports, which is great for security.

> Having said that, it does sound like you have a good reason for
> getting hold of such details. I think it's too major a thing to change
> in the 4.0.x series, but it will be sensible to revisit for 4.1. If
> you can come up with a description of what you need to do, that will
> help guide the most sensible interface.

In short, I guess many transports are (or would be) based on tcp, when
acting as filters. Event ssl is a kind of it, but unfortunately it comes
from a different source so that it does its own socket handling.
The point is that tcp should expose the concepts of endPoint and connection
so that they can be used by others. A simple pattern could be achieved by
having endPoint acting as a factory of connection, thus hiding the socket
itself.
I guess we will talk again about this subject at the time of 4.1.
Thanks,

Renzo