[1362] in Commercialization & Privatization of the Internet
Will ANS connect to CIX? Why is this a difficult question?
daemon@ATHENA.MIT.EDU (Christopher Davis)
Wed Sep 18 10:30:36 1991
Date: Wed, 18 Sep 1991 10:17:42 -0400
From: Christopher Davis <ckd@eff.org>
To: steve@ncri.cise.nsf.gov (Stephen Wolff)
Cc: com-priv@uu.psi.com
In-Reply-To: steve@ncri.cise.nsf.gov's message of 18 Sep 91 12:20:08 GM
SW> == Stephen Wolff <steve@ncri.cise.nsf.gov>
SW> [...] By your use of the loaded term `hedge' I think you
SW> misapprehend ANS' natural and quite proper concern, as the operator
SW> of the biggest, fastest, and best-managed IP pipes around, to be
SW> fairly compensated for others' use of its facilities. Of course
SW> what's `fair' is negotiable, but at a minimum I should think it
SW> includes a return to ANS adequate for it to expand its facilities
SW> enough at the margin that the CIX traffic carried does not degrade
SW> ANS' delivered performance to its other customers - especially the
SW> one I care about the most, the NSF community via Merit. -s
I find it hard to believe that a T1 connection (or two) to the CIX can
somehow pass enough traffic to choke a T3 network. After all, the
bottleneck will be the T1! Even if, for the sake of argument, both
CIX-East and CIX-West are somehow completely being used to funnel
traffic between {AlterNet,PSINet,CERFnet} and ANSnet, the bottleneck
will keep the total traffic to 2xT1. (Now, it may get backed up at the
CIX, but that's not ANS's problem...).
As for the value of the T3 network, it still seems to be in the teething
stage. Being in Boston, and seeing both the NEARnet and AlterNet sides
of things, I see T3 NSFNET problem reports at a very steady rate. I see
CIX/AlterNet/T1 NSFNET problem reports much more rarely. For example,
the Georgia Tech and Pittsburgh ENSSes have been (according to my mail)
down since about 3 am this morning (I haven't seen any indication that
that's changed; traceroutes to psc-gw3.psc.edu are routing over the San
Diego interconnect and across the T1s).
It seems to me that, historically, bandwidth has been the prime limiter
of "overuse" of IP links (hence the move to T3s to open up more
bandwidth). I believe (and I do not speak for the EFF, nor for ANS ;-)
that the T1 link will act as an "equalizer" at the CIX, such that ANS
cannot be "flooded" from the far side in any way.
Christopher Davis <ckd@eff.org> | ELECTRONIC MAIL WORDS OF WISDOM #5:
System Manager & Postmaster | "Internet mail headers are
Electronic Frontier Foundation | not unlike giblets."
+1 617 864 0665 NIC: [CKD1] | -- Brian Reid <reid@decwrl.dec.com>
