Does this surprise you as much as it does me?
I don't know whether I should care, or whether this is bad for the
Internet, but it is certainly something.
No, I knew that Linux used TCP-BIC by default. I think it is bad for
the Internet.
Both Linux and Microsoft, I believe, are implementing non-default TCP
congestion control mechanisms that have never been through any form of
community feedback or IETF review. Their main characteristic, I believe
(that is, in my view), is that they are both more aggressive, in terms
of the
response function, that the response function for TCP as standardized in
the IETF.
I would hope that this is not the beginning of a "war" of each TCP
implementor
trying to use a response function more aggressive than that of its
competitors,
in an effort to give their users a higher fraction of the link
bandwidth.
This is a competition that would just lead to higher and higher
steady-state
packet loss rates for a given available per-flow bandwidth.
Not too charming.
I have expressed this privately to Injong Rhee, and he told me that
there
has been speculation about whether Microsoft's TCP or Linux's TCP
currently has the most aggressive response function, but that the answer
was not yet know.
(Writing from vacation in the French countryside, with very
limited Internet access. So I haven't read all of the email on all of
the
mailing lists on these issues..)
- Sally
http://www.icir.org/floyd/
