Re: DDP 7.3 - Data Source MUST transmit DDP Messages in order

At 01:57 PM 11/5/2003 -0600, Caitlin Bestler wrote:
>Section 7.3 of the DDP spec states that:
>
>    At the Data Source, DDP:
>
>      * MUST transmit DDP Messages in the order they were submitted to
>         the DDP layer,
>
>
>There are two possible interpretations of that requirement, depending
>on what the meaning of "transmitting" a DDP Message is.
>
>The liberal interpretation is that the DDP Layer "transmits" a DDP
>Message when it presents the last segment in the message to the LLP.
>
>A more complete requirement is that ALL DDP Segments must be presented
>for transmission to the LLP *before* the first segment of the next
>DDP Message may be submitted.
>
>
>For most DDP receivers there is no difference between those two
>requirements. However, there can be a significant difference for
>a DDP receiver that accepts DDP Segments in order from the LLP
>and is using a buffer sharing technique such as a Shared Receive
>Queue where receive buffers are allocated "on demand".
>
>The latter more restrictive interpretation (which matches what
>virtually all implementations would do anyway) would allow a
>receiver working with in-order deliveries from the LLP to know
>that it would never need to have more than one pending receive
>buffer. By the time it needed to allocate the next buffer, the
>prior one would be eligible for delivery.
>
>Receivers which process out-of-order would never detect the
>difference between the interpretations. Receivers working with
>connection specific receive resources (such as a classic
>receive queue) would also be indifferent.
>
>However, I believe there is real value in being allowed to
>bound the number of pending allocations from a shared receive
>queue when DDP Segments are delivered in order by the LLP.
>Value that exceeds any imposition the tighter restriction
>would place on the Data Source.
>
>
>I would request the following:
>
>- Does the WG agree that the current wording is compatible
>   with both interpretations? Or is the latter interpretation
>   locked in by some other language the at I overlooked?
>
>- If there is agreement that the current language is at
>   least potentially ambiguous, is there any objection to
>   adding the clarification that all DDP Segments for a
>   DDP Message must be submitted for transmission to the
>   LLP before any Segment for a following DDP Message.

Implementations should be able to deal with OOO packet delivery and being 
able to target multiple receive buffers if they are posted (shared or 
exclusive).  Given there is no delta in the wire protocol or the expected 
interoperability in this respect, I do not favor trying to tighten how the 
transmitter is implemented as that is potentially delving into the specific 
design options in the hardware.

Mike