[9432] in Commercialization & Privatization of the Internet
Re: AUPs and Connectivity
daemon@ATHENA.MIT.EDU (Stephen Wolff)
Wed Jan 5 09:31:51 1994
Date: Wed, 5 Jan 1994 07:59:23 -0600 (EST)
From: Stephen Wolff <steve@nsf.gov>
To: Lars Poulsen <lars@eskimo.cph.cmc.com>
Cc: com-priv@psi.com, mstrange@fonorola.net
In-Reply-To: <9401051008.AA22384@eskimo.CPH.CMC.COM>
> It would seem to me, that if NSF shifts funding from direct payments to
> ANS to the regionals so that the regionals buy their own interregional
> connectivity from ANS (or competitively from ANS, AlterNET, Sprint and
> PSI) then there will no longer be an NSF backbone on which to apply the
> AUP.
>
> It would also seem to me that the vBNS has been defined to be
> functionally different from the current NSFnet and to explicitly NOT be
> meant as an inter-regional default backbone.
Exactly right on both points.
> 1) Why is NSF planning to extend the ANS agreement rather than shift
> the funding ?
> 2) Why would ANS need a year's guaranteed funding to shift to a
> competitive environment ? They have had plenty of warning, and
> with the head start they have enjoyed, they would appear to be
> positioned to win most of the customers anyway.
It's not ANS alone that needs the year. It's the regionals + ANS/Merit.
We (NSF) need to get money to the regionals, they need to conduct their
competitive procurements from the transit service providers and, when they
turn up, start transitioning the routes off the current Backbone Service
to their individual new arrangements.
You remember the last "flag day", right? When we switched to TCP in 1983?
The net was a LOT smaller then, but even so the "instantaneous" transition
took the better part of eight months and was accompanied by massive
service disruptions.
By contrast, the last NSFNET transition (T1->T3) took also several months
and (modulo the T3 technology being not quite ready for prime time) went
moderately smoothly (for some value of moderately) as routes were
transitioned a few at a time. And that was in a network at least two
orders of magnitude larger than in 1983 by almost any measure (nets,
traffic, hosts, bitrate,...).
In a net that's grown even bigger since then, I cannot imagine the
forthcoming transition being done any other way. We've told both Merit
and the regionals that stability of service to the networked community is
NSF's first priority during the transition.
> To my naive ears, the talk of funding a graceful shutdown of NSFnet
> sounds like bridging until vBNS can take over. Letting vBNS assume
> the same old role as the default backbone, seems like a prescription
> for endless grief.
Nope. The vBNS is NOT be a general-purpose transit backbone; that would
indeed engender enduring, if not endless, grief.
-s
