Interesting People mailing list archives

a bit more on ANS, CIX -- a good argument and set of observations


From: David Farber <farber () central cis upenn edu>
Date: Sat, 20 Nov 1993 20:09:15 -0500

Posted-Date: Sat, 20 Nov 1993 19:34:41 -0500
From: cook () path net (Gordon Cook)
Date: Sun, 21 Nov 1993 00:34:38 GMT
X-Mailer: Mail User's Shell (7.1.1 5/02/90)
To: farber () central cis upenn edu
Subject:  for int peop


This is a revised and expanded version of what I posted to com-priv last night.
It may be cross posted to other lists as long as it it posted in full with no
edits.


ANS joins the CIX!  Good news indeed.  What does this do to level the commercial
market place?  And does it mean that there is still an ANSnet AUP?


1.  Let's say Toys R Us wants to join the internet and that universal
connectivity is VERY important for Toys R Us.  It can connect as a PSI or
Alternet or other CIX customer and now be assured of getting to every mid-level
via the ansnet backbone?  It sends commercial traffic wherever it wishes
including anywhere in the CIX.  Right?


2.  Lets say it wants to join ANS - which now also is the CIX.  If it connects
at a CNSS backbone site, it pays whatever rate it can negotiate with ANS and
sends commercial traffic wherever it wishes.  If on the other hand it connects
at an ENSS site or via a mid-level it must pay the 2,000 or 4,000 dollar charge
to the infrastructure pool depending on whether the link is 56k or t-1?  And
once it does this it sends commercial traffic wherever it wishes including
anywhere in the CIX?  Right?
'
3.  Or lets say it connects to a mid-level. If it does this it is free to use
ANS backbone routing as soon as it ponies up the 2 or 4 thousand dollar
surcharge?  It can send commercial traffic anywhere it wishes except to other
CIX networks not also attached to ANSnet?


But if Toys R Us is connected to a mid-level that is a cix member, then it can
use its and ANS's CIX membership to traverse the ANSnet backbone if it needs to?
(WITHOUT paying the 2 or 4 k dollar surcharge?)  And it sends commercial traffic
wherever it wishes?


What about the mid-level (regionals) who are NOT members of the CIX?  They may
provide connectivity for their commercial customers to all the other commercial
customers of all other ANSnet connected regionals by paying the 2 to 4 thousand
dollar a year surcharge to ANS.  However their commercial customers would still
be blocked from getting to customers of CIX member networks not also attached to
the NSFnet backbone?  Why? Because they are customers of a non CIX member
network.


Under such circumstances it seems that it will be to the advantage of the
regionals to join the CIX rather than pay ANS the commercial surcharges (aka
transit fees)?  A regional would have to have *very* few commercial customers
and be located in the middle of the US before the line charge to cix east or cix
west and cix membership would wind up being more than the cost of buying transit
from ANS. Besides if a regional (mid-level) joined CIX rather than paying ANS
the transit charges, the regional would get greater connectivity for its
commercial clients.


Net effect:  the internet is far less balkanized.  The NSFnet AUP is de-facto
dead -- unless that is ANS can convince a customer to sign up in a situation
where it is liable for having to pay the infrastructure pool commercial
surcharge.


I conclude that ANS is *unlikely to keep the surcharges* in effect because they
will now place it at a marketplace disadvantage and that the NSF AUP is once
more at long last dead - de facto if not de-jure.


By the way what happened to the transit fees that NEARnet and BARRnet customers
are paying for ANSnet transit?  Seems like they should not have to pay them
anymore since ANS by virtue of being a CIX member now has to carry their
commercial traffic?


What are the implications for Russian routing?  If every CIX member must route
to every other, can NSFnet routing to Russia still be refused?  I hope not.


Are my conclusions incorrect?  If so would people please point out where and
why?


PS.  Seems like some of the most serious aberrations of the last 3 years are on
their way to being fixed.  With the kinks almost out of the current
architecture, isn't it a shame that the NSF is being allowed to impose the new
architecture on us all.  Are we really powerless to STOP this imposition?


_______________________________________________________________
Gordon Cook, Editor Publisher:  COOK Report on Internet -> NREN
431 Greenway Ave, Ewing, NJ 08618
cook () path net                                   (609) 882-2572
_______________________________________________________________


Current thread: