[tcpm] Re: Some feedback on draft-ietf-tcpm-rfc2581bis-01.txt

> Page 8. The draft says:
> "    6.  When the next ACK arrives that acknowledges new data, a TCP
>          MUST set cwnd to ssthresh (the value set in step 1).  This is
>          termed "deflating" the window.
> 
>          This ACK should be the acknowledgment elicited by the
>          retransmission from step 1, one RTT after the retransmission
>          (though it may arrive sooner in the presence of significant out-
>          of-order delivery of data segments at the
>          receiver). "
> 
> Can the ACK really arrive sooner than one RTT? Unless the RTT for
> the connection decreases, I cannot see why this would happen.

As noted, out-of-order delivery can make this happen ... i.e., the ACK
*should* be trigged by the retransmission.  But, if there was just
reordering and no loss then it might not be.  So, the ACK covering the
retransmit could be triggered by the original transmission (and, so
could arrive earlier than one RTT after the rexmt was sent).

> The draft goes on with:
> 
> "       Additionally, this ACK should acknowledge all the
>          intermediate segments sent between the lost segment and the
>          receipt of the third duplicate ACK, if none of these were lost."
> 
> What if it doesn't? Should we stay in loss-recovery until this
> happens, or should we end loss recovery, anyway? (The former, right?)

Loss recovery ends with the receipt of the first new ACK as the document
says.  Right after the above quoted text is this:

    Note: This algorithm is known to generally not recover efficiently
    from multiple losses in a single flight of packets [FF96].  Section
    4.3 below addresses such cases.

RFC 2581 does not try to specify NewReno or SACK-based loss recovery
schemes - even though it recommends implementing such schemes (which are
specified elsewhere).

allman