ietf
[Top] [All Lists]

Re: "redesign[ing] the architecture of the Internet"

2001-02-03 17:10:03
From: Keith Moore <moore(_at_)cs(_dot_)utk(_dot_)edu>

Even so, we as an organization have NO claim on the future of the "Internet
Architecture", whatever that is.  

I strongly disagree.  IETF essentially "owns" the Internet Protocol
specification and has change control over it.

If you have "change control" over something and I change it, then things
happen.  First my changes are backed-out.  Second, I'll be yelled at,
removed from the access list, demoted, and/or fired.  The legal wrangling
between Sun and Microsoft is a demonstration of what happens when one real
world outfit violates the change control of another.

If a rogue IETF WG or cabal of commercial developers violates the "change
control" of the IETF on IP, what happens?  Does Fred Baker remove names
from an access list at venera.isi.edu?


A small group individuals with a cute idea can have dramatic impact, 
no matter what the IETF thinks.  Witness WWW and NAT.

no argument that such a group can have "dramatic impact" (for good 
or ill) but that's not the same thing as changing the architecture.

a bomb can have a dramatic impact on a building also.  but when a 
building is bombed, we don't use the word "architecture" to refer
to the rubble.

When I filter the evocative words from that, it seems to support Mr. Lear's
point.  A minority inside or outside the IETF group can affect things.
As Mr. Lear said, the only "change control" of the IETF over the supposed
"architecture of the Internet" comes from virtues of the IETF's work, not
copyrights, contracts, laws, CVS access control lists, or mission statements.

If official "Network Architecture" were such a good thing, the various
GOSIP's would have been followed, and we wouldn't have this Internet mess.

Among my pet peeves is inflating self- or group-importance with the word
"architecture."  Talk about "the architecture of the Internet" is as hollow
as "implementing TCP/IP" among those who unpack and install things.  The
Internet has a shape.  Much of that shape was intended, but at least as
much just happened.  Calling the shape of the Internet "architecture" as
opposed to the results of a bomb is at least optimistic and arguably
obviously wrong.  If the Internet has an architecture, then what are those
NAT boxes and redirecting proxies?  Where is the ECO that authorized the
world wide web?  The HTTP RFCs are archeology and triage, not engineering
change orders to change the architecture.

Appealing to the change control that the IAB has over the architecture of
the Internet is a bad way to deal with rogue groups, whether inside or
outside IETF.  If it were good, we'd have TP1 instead of IPv6, since that
was the Architectural Decision of the IAB.  Far better is to ensure that
the consequences of the rogue (or any) group's plan are understood widely
and early, and then let the intellectual and commercial markets decide.
(but that does not mean that the IETF gets to review all documents and
monitor all discussions.)

In other words and contrary to some recent complaints, it's good to
publicize the problems in the notion of NAT as it has evolved.  Thanks in
part to Keith Moore's efforts, the part of the public that prattles
"implement web a server" to mean "configure Apache" instead of "write
Apache" is beginning to get a glimmer that NAT boxes might not be wonderful
cornucopias of IP addresses that also provide perfect security.  The best
way to deal with the next error is not to outlaw discussion of it or even
to require that its advocates meet in public (because you can't), but to
think and talk about it.


Vernon Schryver    vjs(_at_)rhyolite(_dot_)com