Re: DDP draft specification

At 07:15 PM 8/21/2005, Caitlin Bestler wrote:

>On Aug 19, 2005, at 10:28 AM, Sanjay Goyal wrote:
>
>>Hi,
>>
>>The below 2 sentences are from draft-ietf-rddp-ddp-05.txt
>>
>>1. DDP provides enough information in each DDP Segment to allow the
>>ULP Payload in each inbound
>>     DDP Segment payloads to be directly Placed into the correct ULP
>>Buffer, even when the DDP Segments
>>     arrive out-of-order.
>>2. This specification requires reliable, in order Delivery LLPs.
>>
>>First sentence says that DDP segments which are MPA FPDUs in TCP
>>case can arrive out-of-order and
>>second sentence says that LLP (MPA layer) needs to provide ordered
>>delivery. Isn't it contradictory?
>
>The same distinction between placement and delivery that applies for
>RDMAP and DDP
>MAY apply to the DDP/LLP interface.
>
>The DDP/LLP interface is not standardized, but the essence is that:
>
>1) the LLP MAY supply DDP Segments in the order received (some might
>even say SHOULD).
>     The adaptation layers for TCP and SCTP are specifically designed
>to enable this by providing
>     sufficient data to enable placement even when LLP re-ordering
>has occurred.
>2) The LLP MUST still indicate in-order delivery to DDP. DDP needs
>this because it MUST NOT
>      deliver anything to RDMAP that was not fully delivered by the LLP.
>
>One of the review criteria that the IETF RDDP work group insisted on
>was that DDP must
>be implementable over a standards socket interface. DDP in fact
>allows this, but that interface
>does not provide optimization for out-of-order placement.
>
>An implementation can define its own DDP/LLP interface that
>distinguishes between
>placement and delivery. Such an interface can even be described as
>"cleanly layered",
>it just isn't using the classic sockets interface to define the clean
>layering.

Side note: There is also a new Sockets Extension API that adds memory 
management, event management, async comms, etc. that people should take a 
look at since it is a Unix standard.  It will work well for RDMA / SDP.

Mike

_______________________________________________
rddp mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/rddp