Well, before you decide this is crazy, a UDP endpoint has a "socket", so
an unconnected cm_id is not really all that crazy. A cm_id at the layer
we've defined it encapsulates primarily application layer information
like callback function pointers, contexts, and addresses. These values
are every bit as important for an "unconnected" QP. 

The fact that iWARP doesn't currently have a notion of multicast, etc...
doesn't mean that a) other transports besides IB won't, b) that iWARP
companies haven't ever thought about how to do it and will never do it,
or ...

Ok...maybe it really is crazy....

On Fri, 2006-01-27 at 16:43 -0800, Roland Dreier wrote:
>     Roland> I'm not sure this makes sense.  Only IB has the notion of
>     Roland> multicast on QPs and only UD QPs can be used for
>     Roland> multicast.  So why would we want a transport-neutral API
>     Roland> for an IB-specific feature?
> 
> Oh yeah... also why would you want a CM ID for an unconnected QP?
> 
>  - R.
> _______________________________________________
> openib-general mailing list
> [email protected]
> http://openib.org/mailman/listinfo/openib-general
> 
> To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general

_______________________________________________
openib-general mailing list
[email protected]
http://openib.org/mailman/listinfo/openib-general

To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general

Reply via email to