Based on the discussions that have taken place so far, I believe that we're
trying to define something more complex than what's needed. Here's what I think
that we want:
At a minimum, we need an assigned service ID to identifies a TCP/IP connection
service. For simplicity of the implementation, I would use an ID similar to
that defined for SDP:
0x00 14 05 xx xx xx xx xx
I don't know that the SWG or IBTA needs to be involved defining the protocol
beyond assigning the service ID.
The connection service can define service IDs as:
0x00 14 05 00 00 00 dst port
And a private data format for the CM REQ:
version(8) | reserved(8) | src port (16)
src ip (16)
dst ip (16)
user private data (56) /* for version 1 */
Other private data would be left unchanged, though if we wanted to get more
sophisticated, we could define REJ codes to indicate bad addresses/version/etc.
Not surprisingly, this is exactly what's implemented in the CMA and working today.
On a related note, it would be convenient if SDP were changed to run over this
protocol.
- Sean
_______________________________________________
openib-general mailing list
openib-general@openib.org
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
