[2034] in Commercialization & Privatization of the Internet
Re: Understanding COMBITs
daemon@ATHENA.MIT.EDU (Einar Stefferud)
Sun Jan 19 04:35:51 1992
To: Marvin Sirbu <ms6b+@andrew.cmu.edu>
Cc: com-priv@psi.com, Steve Simmons <scs@lokkur.dexter.mi.us>
In-Reply-To: Your message of Tue, 14 Jan 92 10:40:35 -0500.
Reply-To: Stef@nma.com
From: Einar Stefferud <Stef@nma.com>
Date: Sun, 19 Jan 92 00:46:54 PST
Marvin Sirbu wrote:
> My reading is that this is NOT correct. Commercial surcharges would
> NOT be incurred by traffic between a CO+RE designated customer of ANS
> and an RE customer of a midlevel.
I agree entirely with this interpretation of current reality under the
NSF RE-AUP. But, lets generalize it a bit while we are here.
Think of the NSFnet and the NSF sponsored mid-levels as "private"
networks (which is why they must obey an RE-AUP, just because they are
funded for this purpose by the Govt which provides them with a
non-public network for the purposes of RE, and not for any other
purpose). Any other private enterprise or government network must
have the same kind of AUP (e.g., APPLE does not allow traffic on its
corporate IP network that is not related to some APPLE company
purpose, and neither does IBM, or DEC, or NMA for that matter).
I realize that many of us do not want to think of the NSFnet as
private, but I assure you that it fits the model just fine, and I
strongly believe that if we sort out out terminology here, we will all
come to a sudden and startling understanding of this muddle.
So, we can now understand that the basis of any AUP is that the owner
of any "private" network facility has the right (and the US Govt has
an obligation to the taxpayers) to protect its facility from
inappropriate (unauthorized) use of any kind.
So, now lets look at the Plain Old Telephone Service (POTS) model and
see what we can learn. When Joe Blow (in his commercial office calls
a Govt employee using Joe's commercial business telephone line, the
call is switched at some point onto the Govt FTS2000 system (which is
a private Govt Telephone network and subject to a very real AUP) and
when the govt employee answers, hey engage in a conversation because
the govt employee decides that this call fits within the AUP rules of
FTS2000. If not the Govt emploee should hang up. Indeed, and hey
discuss commercial aspects of a consulting engagement that Joe is
selling. We all understand that there is nothing wrong with this
picture. One end of the connection determines that this is an AUP
complaint use of the FTS, and so the call is just as AUP compliant as
though the call was between two FTS2000 users who are compliant.
However, we all understand that someone operating as a bookie from a
govt office is out of compliance with the AUP, as are many other
possible ways that people can misuse the govt FTS2000 network (or the
NSFnet) if they try. The number of ways to violate are legion, and I
am not going to enumerate them here. The essence of the issue here is
that someone at the PRIVATE end makes a determination of PRIVATE AUP
compliance and stands to face the consequences if that determination
is in error.
(Yes, I know that incoming calls from the PUBLIC network might
severely test the judgement or the patience of the PRIVATE network
callee, but that is also not the essence of the issue we are
addressing here.)
So, we now have three general kinds of connections that we can
examine, so let me try to use some helpful labels for each end point
of each possible pair.
PUBLIC <===> PUBLIC (Only AUP is "must pay the bill")
PRIVATE <==> PUBLIC (Must comply with PRIVATE AUP)
PRIVATE <=> PRIVATE (Must comply with PRIVATE AUP)
If it will help you, consider that PUBLIC here is simply defined to
mean NOT-PRIVATE in the terms of this message.
Now, back to the ANS COMBITs rule where the relevant AUP is the NSF
RE-AUP.
PUBLIC <===> PUBLIC (Only AUP is "must pay the bill") -- COMBITS = YES
PRIVATE <==> PUBLIC (Must comply with RE-AUP) ---------- COMBITS = NO
PRIVATE <=> PRIVATE (Must comply with RE-AUP) ---------- COMBITS = NO
To say this in a sentence, if one end is RE, then the whole connection
is RE, because RE means that it is compliant with the NSF RE-AUP, and
there is no rational way for any connection to be RE from one end to
the middle, and not-RE from the middle to the other end.
Any other conclusion in the COMBITS case is to argue that FTS2000 is
doing it wrong, and that FTS should change what it is doing. It is
also to argue that APPLE and IBM and DEC and NMA and many others are
doing it wrong for their IP networks.
So, my conclusion is that ANS is simply doing it wrong.
Now then, do not misunderstand that counting COMBITs has nothing to do
with NSF requirements that any PUBLIC service provider must block
(filter-out) traffic from deliberately declcared non-RE-AUP
"commercial" customer networks, or must get permission from any NSF
supported midlevel network. This is just the case that is covered by
the ANS "connectivity agreement".
COMBITS should only be counted when both ends are from "gateway
agreement" customers.
Cheers...\Stef
