[openib-general] [PATCH 0/6] Tranport Neutral Verbs Proposal.

[Caitlin Bestler]

openib-general-bounces at openib.org wrote:
> On Mon, Jul 31, 2006 at 09:52:48AM -0500, Steve Wise wrote:
> 
>> rdma_* is more descriptive than something like ofv_* or of_* in my
>> opinion.  I would think the prefix should help describe the
>> functionality being implemented:  Transport Neutral RDMA.
> 
> Some functions are RDMA. Others are not. If all are called
> RDMA, that's misleading.
> 
> For example, in IB, there is send/receive as well as RDMA.
> ULPs often use send/receive for short messages.
> 
> I wouldn't know anything about the non-IB parts of Open
> Fabrics, but I would bet that there is non-RDMA functionality in them.
> 
> The common concept is messaging, not RDMA.
> 

That would imply that the purpose of the openfabrics stack
is to replace netdev. I don't think we want to go there.

The broadest scope I can imagine is that openfabrics is
for networking where the immediate using layer is aware
of memory registration.

RDMA-associated send/recv is message based, but then so
is SCTP. And SCTP is handled through netdev. The more
important distinction is that even when using anonymous
buffers via openfabrics they are *registered* buffers.
Further, they are supplied by the application layer.

That contrasts with non-QP-based networking where buffers
are supplied without pre-registration and some degree of
system buffering is available to hold received content
before the application asks for it.

Now you could argue that "RDMA" is not the best label for
this class of service. You might be tempted to say it is
"qp based" networking, but then you realize that a "Queue
Pair" is an RDMA capable network endpoinit that is implemented
using three or more queues. In other words, no short memorable
label is ever perfect. Just go with the label that people 
recognize whether it is the best theoretical description
or not.

Aside from *what* the label is, the key question is whether 
a given function is "generic RDMA" versus IB or iWARP specific.
I believe that is the intent of the "rdma_" prefix, and I haven't
spotted any misapplications yet.