[e2e] the evolution of deployability
David P. Reed
dpreed at reed.com
Tue Dec 17 06:15:07 PST 2002
At 09:14 AM 12/16/2002 -0800, Dave Crocker wrote:
>This suggests looking at a number of human-related research areas, both with
>respect to barriers within individuals and barriers within groups.
Dave - your insights seem to me to be "dead on". I would add one
additional factor from my experience.
Large vendors tend to interact with the human system in non-positive
ways. In particular, product ideas that serve their own interests rather
than the networks' or the users' are often marketed heavily to the IETF
constituency. NAT is a great example. Vendors M and C (who shall remain
nameless) got very aggressive in selling NAT, despite its severe design
limitations and application impact. They marketed heavily to users and
IETF a set of solutions that clearly broke end user applications already
deployed. At the same time, these two vendors refused to consider an
incremental migration path to IPv6, arguing that it "wasn't ready
yet". The internal politics (including the weak protocol designers who
have flooded the IETF) could not be organized to do wise things, so they
responded to these pressures by creating more barriers to deployment.
These players exploited the "human systems" to block deployment of what
could have been very simple and surgical evolutionary improvement paths.
A collection of children given the world's best scientific instruments are
not going to invent Quantum Theory. We shouldn't expect a democratic,
market-driven mob to invent and deploy major design improvements in the
Internet by consensus and votes.
RED+ECN is one of the "surgical" improvements that should just happen - no
flag day is required and every actor can act independently. It's possible
to do this change incrementally, and the benefits are dramatic - especially
because they are compatible with a wide range of protocols beyond the
TCP-based protocols. Yet the mob is incapable of even this much
More information about the end2end-interest