[1331] in Commercialization & Privatization of the Internet
Re: Questions: ANS Plan for Commercial Services
daemon@ATHENA.MIT.EDU (Ittai Hershman)
Mon Sep 16 10:17:20 1991
Date: Mon, 16 Sep 91 10:14:32 EDT
From: Ittai Hershman <ittai@shemesh.ans.net>
To: proll@casbah.acns.nwu.edu (Peter Roll)
Cc: com-priv@uu.psi.com
In-Reply-To: Your message of Thu, 12 Sep 91 08:18:55 CDT
Peter Roll asks:
Why do RE and CO communities have to be considered as and treated
differently? Why not just one user community? Is differential
pricing between CO and RE legal or sustainable? Is it acceptable
or desirable public policy? Will changes to statutes be required?
Has a legal opinion been rendered on this matter?
We (the Internet community) are in the early stages of privatizing and
commercializing a government sponsored and funded network. It remains
clear that at least some portions of the Internet in the US will
continue to be federally funded (e.g. the federal agency networks).
Any part of the Internet which is directly funded by the federal
government is obliged to restrict itself to traffic in support of
research and education. As long as that is the case, RE and CO
traffic (note that this issue is "traffic" and not whether the
companies involved are commercial entities) will need to be identified
as such.
It should also be noted that there is a history of offering special
treatment to support research and education in the US (e.g. hefty
discounts from most computer vendors to RE institutions).
Why is there any limitation on acceptable use of the ANS or ANS
CO+RE backbone? What is ANS's Acceptable Use Policy?
What is the ANS CO+RE Acceptable Use Policy?
Why is a commercial (ANS CO+RE) facility not willing to carry any and
all traffic, leaving the users/originators responsible for it?
Are there legal ramifications here that are not clear?
The AUP issue was discussed on com-priv in the spring. I refer you to
Al Weis's posting of 23 July 1991 (which I will send to you offline).
Why is it of interest to gather application type identification
information (page A-4)? This seems to be unwarranted and unnecessary.
The section you reference concludes "it is not practical to gather
application identification information for OSI traffic".
There seems to be no representations of the commercial customers
that pay for the infrastructure pool on the allocation committee
(pg 6). Will this form of taxation without representation work,
and is it good policy?
This is an interesting point. We are not hardbound to a particular
definition of the composition of the Resource Allocation Committee.
Thanks for the feedback.
Service and Technical Implications
==================================
Is discrimination in favor of large packet sizes (pg. A-6) going to
hike the imputed costs of casual e-mail traffic?? What are the
implications of this for various kinds of NREN uses? Will it
discourage the kind of mass low-performance communication among
scholars and students that we are trying to stimulate? If in
fact the imputed costs of casual e-mail traffic do NOT flow
back in any way to RE user organizations, with the large
packet size discrimination nevertheless inhibit e-mail and
other small-packet communications by CO organizations with which
we would like to encourage greater interaction?
To the contrary, promoting efficient use of the backbone will help
make the backbone available and affordable to low-bandwidth users. I
should also point out, that e-mail per se does not necessarily use
small packets. A particular implementation may. Our point is merely
to promote efficient use of the network by rewarding (with a discount)
such usage.
What burdens will the COMBit technique place on midlevels to
charge back to their CO+RE users appropriate amounts to cover
the CO overpayments they make (pg. A-11)? Will the technology
to do this be available and affordable for midlevels? Will the
technique tend to force midlevels out of the CO business to avoid
this problem? Will it induce them to segregate CO and RE users
on separately-gatewayed subnets? What are these implications for
midlevels? Have these been thoroughly understood and debated by
the FARNet membership?
As the paper states at its start and finish, this plan was developed
in conjuction with representatives from NSF, FARnet, and several
mid-levels.
We don't think it will stifle mid-levels at all. To the contrary, we
think that ANS can cooperatively work with them to commercialize the
Internet.
Definitions
===========
The definitions in Attachment C do not make it clear who is
included in the RE and CO categories. Schools, colleges, and
universities clearly qualify as RE. Westinghouse Research Labs
is a research organization, for instance. Does this mean that
Westingouse Electric Corp. is RE? That Westinghouse Research
must be separately incorporated to qualify as an RE org? Or
something else? Are Commercial Service Providers and Commercial
organizations both CO orgs? If so, why bother to differentiate
between them?
CO and RE are not sites, but usage units known as COMBits. A network
(e.g. midlevel) representing a single autonomous system may advertize
multiple network numbers into the ANS network. Each of these network
numbers may or may not represent a site. However each network number
is labelled as CO or RE. COMBit units are measured for network
numbers whose traffic traverses a single ANS gateway. The COMBit
units may be summarized by percentages of CO COMBits or RE COMBits for
any organization that represents a collection of these network numbers
which are adverstized through the gateway. An organization may be a
midlevel, commercial service provider, or anyone else who represents a
group of network numbers which are advertized into the ANS network.
Early in the document the term "network service provider" is used
to include commercial orgs such as ANS, ANS CO+RE, PSI, CIX, etc.;
and midlevels which may accomodate RE and CO traffic. What kind
of "institution" (pg. C-1) is each of these network service
providers, or are they a separate class of org becuase they do
not, themselves, originate traffic?
PSI, Alternet, and CERFnet are treated as network service providers,
just as the mid-levels are. ANS, by the way, is a non-profit company;
CO+RE is its taxable (for-profit) subsidiary.
Other questions
===============
On Pg. 9, is it ANS or ANS CO+RE that funds "....the difference
between the CO and the RE price of its attachement fee to the
infrastructure fund....?"
ANS CO+RE.
It is my understanding that NSF grants do not always cover the
total costs of LXC loops. Am I wrong about this?
I'm not sure I understand the issue here. The NSF grant is credited
to the Gateway attachment invoice as is iluustrated in Attachment B,
Example B.
-Ittai
