sean> At a minimum, we need an assigned service ID to identifies a sean> TCP/IP connection service. For simplicity of the sean> implementation, I would use an ID similar to that defined for sean> SDP: sean> sean> 0x00 14 05 xx xx xx xx xx sean> sean> I don't know that the SWG or IBTA needs to be involved defining sean> the protocol beyond assigning the service ID.
Standardizing the protocol will ensure interroperability. sean> The connection service can define service IDs as: sean> sean> 0x00 14 05 00 00 00 dst port sean> sean> And a private data format for the CM REQ: sean> sean> version(8) | reserved(8) | src port (16) sean> src ip (16) sean> dst ip (16) sean> user private data (56) /* for version 1 */ Are the numbers in parens in bytes or bits? It looks like a mixture to me. sean> Other private data would be left unchanged, though if we wanted sean> to get more sophisticated, we could define REJ codes to indicate sean> bad addresses/version/etc. Not surprisingly, this is exactly sean> what's implemented in the CMA and working today. I agree. We should keep it simple. sean> On a related note, it would be convenient if SDP were changed to sean> run over this protocol. Agreed. _______________________________________________ openib-general mailing list [email protected] http://openib.org/mailman/listinfo/openib-general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
