Re: silent dropping of TCP segment

On Mar 25, 2004, at 6:21 PM, Barry Reinhold wrote:

>
> It is not clear to me why there is a "MUST" in clause 5.1.2. It would
> seem desirable to allow MPA some control over when the ACK is to be
> generated.
>
> I would like to see the "MUST" removed from clause 5.1.2.
>

As noted previously, clarifying that the MPA layer MAY be able to
reverse a prior dtermination by the TCP layer that a segment was
valid might be worthwhile.

But simply dropping the language is the wrong way to do it.
There are two important constraints that I believe the same
language is part of:

a) the TCP layer MUST NOT pass duplicate segments to the
MPA layer multiple times (shifting the responsiblity for detecting
duplicates up a layer).

b) a rototilled implementation MUST NOT attempt to do "cut
through" placing. It cannot be a DDP Segment until *after*
it is a valid TCP Segment or SCTP Data Chunk.