[tcpm] Comments on draft-blanton-tcpm-3517bis-01

[Matt Mathis <[e-mail]>]

I would like to suggest an alternate wording for Section 5.1 paragraph 1:

To make SACK robust in the presence of unknown implementation bugs and
reneging, RFC 2018 indicates that the sender MUST ignore prior SACK
information a on a timeout.   However, since preserving SACK
information on timeouts greatly reduces wasted network capacity under
overload conditions, this requirement is probably too strong.
Errata1610 amends the MUST to a SHOULD and suggests that as long as
there are robust tests for inconsistent SACK information, an
implementation might retain and use SACK  information across a
timeout.  The SACK consistency checks might include snd.una advancing
to just before a byte that has been reported in a SACK block, and
repeated timeouts or DSACKS.   This document does not formally update
the specification in [RFC2018], however implementers should consult
any updates to [RFC2018] on this subject.  Furthermore, a SACK TCP
sender SHOULD utilize all SACK information made available during the
slow start phase of loss recovery following an  RTO.

(Feel free to make additional adjustments)

Thanks,
--MM--
The best way to predict the future is to create it.  - Alan Kay