[e2e] Google seeks to tweak TCP
Craig Partridge
craig at aland.bbn.com
Mon Feb 6 09:26:36 PST 2012
> On 3 February 2012 04:06, Detlef Bosau <detlef.bosau at web.de> wrote:
>
> > Everyone is free to do research in this field and to discuss his results
> > within the community - and when changes make sense, everyone is free to
> > contact the IETF and to convince them by compelling work.
>
> When the IETF started, the founders were told that everything should
> go through the ISO. The IETF community found the ISO too slow, and
> just experimented for themselves. The same applied to the ITU-T and
> the ATM forum.
We're deviating from E2E to history but I think this point matters, so
a brief comment from one of the IETF founders. We were never told to go
through ISO. But the broad issue of speed is entirely right. IETF was
a split off from INARCH (the Internet Architecture Task Force) and was
motivated by a bunch of folks, largely ISPs and campus network operators,
who desperately needed a bunch of nitty-gritty operational problems solved
and documented -- problems that seemed to be falling through the cracks
between the various Internet Task Forces of the time (there were several).
As a result, the mode of operation was "find a working
solution, document it in an RFC, and move to the next problem."
Note that most TCP work was NOT done in IETF but was done in the End-to-End
Task Force (which created this mailing list).
> Is it possible that the "TCP innovators" are now just finding the IETF
> too slow, and doing exactly as the founders of the IETF did? Some
> argue that the Internet is more vital infrastructure now, and so this
> isn't a fair comparison. However the telephone infrastructure was
> already vital when the ATM forum took matters into their own hands.
>
> To remain relevant for congestion control, the IETF needs to balance
> responsibility with responsiveness.
Always a killer problem -- indeed, within a few years of the IETF getting
going, Steve Crocker (an IESG member) noted we were beginning to develop the
processes that tend to stymie standards organizations such as the ability
to appeal decisions and copious mechansisms to ensure fairness over
effectiveness. These, incidentally, were in response to the arrival of
corporate representatives at IETF -- in the early years, folks showed up
to solve problems and didn't worry if the solution was optimal vis-a-vis
their corporate market strategy.
Thanks!
Craig
More information about the end2end-interest
mailing list