[openib-general] putting in dead wood for DAPL and similar abomination

[Talpey, Thomas]

At 05:05 PM 3/1/2005, Christoph Hellwig wrote:
>Similar hint to the NFS over RDMA folks at CITI -
>if you want your stuff to go in use the openib helper directly below
>the transport switch - differnet RDMA transports are too diverse to
>be sanely abstracted out and DAPL does a horrible job at that.  If
>we need to consolidate code for differnt transports we can put it
>into a library later on.

Ok, I'll speak for the NFS over RDMA implementation. (I've brought
your hint to the attention of the CITI folks - we are working together
this week here at the NFS Connectathon).

The NFS/RDMA client, and soon the server, use kDAPL for a simple
reason - we need an RDMA API which allows us to plug in RDMA
NICs without also having to modify NFS client, server and RPC code.

You're trying to sentence us to coding NFS to individual hardware.
It's unacceptable to have to modify NFS and RPC just because a
new adapter has been attached. It's the same NFS/RDMA protocol
over IB, iWARP, and even VI.  Offering "consolidation" "later on" is
an enormous step backward from what we're using (successfully)
today.

Tom.