[openib-general][PATCH][RFC]: CMA header

[Sean Hefty]

Roland Dreier wrote:
>     Sean>  From an implementation viewpoint, I'm not sure we can
>     Sean> distinguish between rejected and peer rejected.  How about
>     Sean> just rejected with some additional reject information in the
>     Sean> case that the user cares?
> 
> I think the right way to think of this API is as implementing an iWARP
> emulation layer for IB.  For a TCP connection, there's only one
> "rejected" status, ie ECONNREFUSED.  So I don't think we can expose
> multiple reject reasons or additional reject data.

How does iWarp fit in with this API?  I was assuming that it was beneath it, 
with this API being more like the connection portion of kDAPL.

- Sean