Ian & Stuart's Australian Mac: Not for Sale/Another.not.for.sale (Australia).iso/hold me in your arms/Nexus/nexus.providers/DGIX

Internet Message Format | 1994-04-07 | 5KB

From: Adam Feuer <[e-mail]>
Date: Tue, 5 Apr 1994 02:07:44 -0500
Reply-To: [e-mail]
To: Multiple recipients of list <[e-mail]>
Subject: Re: The whole CIX concept is flawed (fwd)

an interesting tidbit from a simpatico at sprint.

BTW, i don't know who he is or anything about him, just that these are
interesting ideas. this came off a list discussing the privitazation of
the internet, a done deal now of course, but flavored toward the iNet's
evolution. . .

the description of Russia's internet is particularly interesting, i am
interested if anyone knows more about this. . .

adam

ps. quick glossary:
IX == Internet Exchange, a network interconnection point, usually not for
profit and neutral; eg, CIX [commercial IX], FIX [federal IX], etc

NSP == Network Service Provider

---------- Forwarded message ----------
Date: Sun, 3 Apr 1994 12:15:11 -0400
From: Vadim Antonov <[e-mail]>
To: [e-mail], [e-mail]
Subject: Re: The whole CIX concept is flawed

Distributed Global Internet Exchange (D-GIX) is a pet project of
Peter Lothberg; the idea is to extend level-2 (Ethernet/FDDI)
interconnection "point" to Europa and across the continental USA
using level-2 bridging instead of level-3 routing to allow national
and large regional NSPs to exchange traffic w/o any policy restrictions.

Since D-GIX does not have CIX-like restrictions small NSPs can simply
purchase connectivity from national NSPs with no addititonal fees.
Sprint currently investigates feasibility of building the "center" of
D-GIX in Pennsauken, NY.

I personally do not see D-GIX as the best model; something like
a set of collocated (i.e. not distributed) IX-es and a system of
transit agreements between regionals and national (long-distance)
NSPs can do better.  The idea is to allocate a room; run FDDI
and allow people to bring in their own boxes attached to their own
wires.  The host party can esentially provide room, infrastructure
and maintenance for free to compensate others (since it is getting
"free" connectivity).  This means that effectively shared costs are
zero; i.e. no membership fees etc are necessary.  Some entry barreer
may be necessary to prevent degradation of the structure which
inherently has limited capacity.

This way, there are three categories of NSPs can be accomodated:

1) national NSPs, present at all or majority of IX-es; they do not
   need transit agreements with other NSPs
2) large regionals, present at one IX; in order to maintaing global
   connectivity they need to enter transit agreement(s) with one of
   national NSPs
3) local NSPs, purchasing service from 1 or 2; they do not need any
   transit agreements since their global connectivity is provided
   by their provider's presense on all IX-es or by it's transit
   agreements.

What most national NSPs do not understand is that local NSPs are
providing extremely labor-intensive services like customer support,
consulting and initial set-up.  They also attract beginner customers
who don't want to spend big money on purchasing high-bandwidth
connectivity from national NSPs.  Essentially, they create market
for national NSPs, and migration of customers to national NSPs
when their demands grow is quite natural.  In order to maintain
the business small NSPs should be aggressive and innovative, finding
new niches and, being small companies they're positioned to do
exactly that.  Also, when some kind of Internet billing system will
be created there will be a perfect opportunity for small *information
providers* to gain world-wide reach by utilizing large NSP's capacity
(and probably their collection system, too, for an agreed percentage).
It means that an individual with an idea will be able to build a
nation-wide business in *days* and with minimal expenses.

American domestic Internet market is so huge that one needs to be
extremely short-sighted to *not* cooperate with small NSPs.

This model is tested and works remarkably well (see the example of
RELCOM, the Russian national TCP/IP and UUCP network which currently
has 2 large national NSPs (Joint Stock Venture Relcom and DEMOS+),
over 200 regionals of different sizes in practically *every* city
and about 15000 customer organizations (the prices are still too high
for most individuals).  Also, there is a lot of information providers,
including information agencies, trading companies, electronic publishers
and USENET-based "marketplace".  4 years ago RELCOM started from a
single 386 PC with Unix; the whole thing developed without a ruble from
the government and in economy which can hardly sustain the customer
base.  Two years later the eneterprise gained sufficient weight to break
down the legal monopoly of Ministry of Communications.

--vadim