[tcpm] updated reaction to spurious timeouts

Folks-

Based on some feedback from this group and some further thinking we have
revised our i-d on tweaking the RTO in response to spurious RTOs.  The
new version is:

        Title           : Using Spurious Retransmissions to Adapt the
                          Retransmission Timeout
        Author(s)       : M. Allman, et al.
        Filename        : draft-allman-rto-backoff-01.txt
        Pages           : 7
        Date            : 2005-10-4
        http://www.ietf.org/internet-drafts/draft-allman-rto-backoff-01.txt

The biggest technical change is that we recommend using the standard RTO
and not using the adapted RTO when the cwnd is <= 4 segments.  The
motivation for this is that (a) this is the regime where RTOs are
"expected" when loss occurs and so delaying RTOs in this area may be
more likely to delay needed retransmits and (b) RTOs that happen in this
region have little downside, even if they are spurious (i.e., the cwnd
is not reduced all that much and there isn't all that much needless
go-back-n going on since the window is small).

We'd be interested in hearing what people think about this i-d.  And, in
particular, we'd like to make it a WG item and see about publishing it
as an RFC.  (My co-chair hat is, of course, off in asserting this.)

Thanks!

allman