To State Dept: On IP-Based Networks and the CNRI Definition of Internet
———- Forwarded message ———-
From: Seth Johnson
Date: Wed, May 15, 2013 at 1:19 PM
Subject: On IP-Based Networks and the CNRI Definition of Internet
Okay, I promised to post something bringing together and streamlining my verbose comments regarding the CNRI comments on IP-Based Networks and the FNC Definition of the Internet, which Patrice Lyons posted.
At the WTPF, there are several information documents posted by Richard Hill on distinguishing the Internet, and since participants might thereby bring up this topic at the WTPF this is especially relevant today. (SEE: http://www.itu.int/md/S13-WTPF13-INF/en)
Patrice noted these three documents:
The FNC Definition of the Internet from October 1995:
CNRI’s Comments to the Working Group on Internet Governance in 2005:
Some Myths on the Internet:
In these documents, CNRI presents the FNC definition and reaches two conclusions based on it:
First, since the FNC definition essentially defines the Internet as the set of networks that use a common set of universal identifiers, it follows that 1) the Internet is not packet-based.
Second, since the FNC definition identifies the universal identifiers for the Internet as those used in the Internet Protocol, 2) the idea that some IP-based networks can be said to be separate from the Internet is false.
In the 2005 document CNRI recommends that the WGIG use the FNC definition with a few words added.
In a later statement to the FCC, CNRI disavows even defining the Internet in terms of IP addresses (https://prodnet.www.neca.org/publicationsdocs/wwpdf/0114cnri.pdf), suggesting that other kinds of identifiers might also be a part of the Internet — so the problem that I will describe below has broader implications beyond the WSIS. But for the purposes of the WSIS, it will suffice to address the CNRI’s contributions to the early WGIG discussion.
I am in complete agreement that these conclusions follow from the FNC definition. The problem is that the FNC definition does not adequately address the basic problem the Internet solves.
The FNC definition does not recognize how IP enables interoperability of diverse applications across independent networks, and the use of universal identifiers is not sufficient to account for how IP does this. The packets that happen to use universal identifiers are specifically what makes the Internet platform capable of supporting the broad variety of communications patterns that application developers may wish to employ.
In addition, there are specific types of applications that are not readily supported across independent networks unless those networks treat packets for certain applications specially. QOS is a notable example. Services like this can be supported by IP-based networks — if the routers in the network can be made to support a unified policy by some core authority. But they cannot be readily supported across independent, autonomous networks using IP to interoperate — i.e., using IP to internetwork.
Networks that interoperate in this way are using IP to support the broadest diversity of applications possible across autonomous networks, thereby enabling users and providers to develop innovative applications on the basis of a premise that networks throughout the world will treat IP packets uniformly. They develop applications for the Internet by relying on uniform treatment of IP packets making possible a general purpose platform across independent networks.
CNRI simply has overlooked this characteristic in its analysis of internetworking when it comes to defining the Internet. They have done this even though they frequently characterize the Internet in terms of its supporting interoperability across independent networks in its various commentaries, including those Patrice posted.
The use of IP for general purpose interoperability across autonomous networks is distinct from the use of IP within individual networks where policies can be applied treating packets in specialized ways across routers that are subject to a core authority. Internetworking is thus a type of IP-based network (of networks) that is distinct from IP-based networks that use IP in ways that treat packets specially.
There are two more key points to recognize regarding the FNC definition. First, its specifically citing TCP/IP was a significant advance over the circuit-based orientation of the traditional telecoms. However, its use of the IP addresses within the IP RFC alone as the central characteristic to define the Internet is not adequate to allow us to distinguish between internetworking — interoperation across independent networks — and networks that implement specialized types of services that are not readily supported except by the application of a unified policy of specialized treatment of packets across routers by a core authority. It was adequate to cite the TCP/IP protocol to distinguish internetworking from the circuit-based orientation of the traditional telecoms, but today we are moving past that problem.
And second, when the FNC issued its definition, the telecommunications environment in the United States was characterized by a highly competitive market among thousands of independent ISPs, because at that time telecommunications providers were required to lease their lines under Title II. It is understandable and perhaps to be expected that that element of the underlying context was not specifically recognized in a definition of the Internet issued in 1995.
However, the FNC’s 1995 definition cannot serve as a basis for distinguishing the Internet today, when the foundation in interoperating across independent providers needs to be specifically understood and recognized as a key characteristic that makes the Internet so powerful and dynamic. The use of a common set of identifiers does not adequately serve that purpose.
In this letter from Bob Kahn to Sally Shipman-Wentworth when she was at the State Department, Bob Kahn notes the failure of the WGIG to use the FNC definition, but he also notes the problem of recognizing NGNs: http://2001-2009.state.gov/e/eeb/cip/wsis2005/50918.htm. This problem of recognizing when we’re talking about Internet and when we’re talking about other types of networks remains critically important, and is far more important when contemplating intergovernmental frameworks for Internet stewardship.
The CNRI recommendation was to adapt the FNC definition to encompass more types of networks — adding a qualifier allowing for higher layer services that integrate with lower layers. In his letter to Sally Shipman-Wentworth as well, the important thing to note is that the concern Bob Kahn expresses regarding NGNs is not only for how to distinguish them, but also for providing for “internetting” with them. That is, he approaches them in a way that might allow NGNs to be subsumed under the governance regime being developed by the WGIG. This may account for the way CNRI has attempted to keep the term Internet broad, though it works against actually distinguishing it by some of its most important characteristics.
One can understand that a broader definition of the term Internet would create the opportunity to apply Internet governance to a broader variety of networks, but when the question is how to distinguish the Internet by its important characteristics so that policy making does not undermine it, it does not help to overlook the way in which internetworking is accomplished between independent networks by means of IP packets. While CNRI’s conclusions that the Internet is not a packet network and cannot be distinguished from other types of IP-based networks do follow from the FNC definition of the Internet, the FNC definition cannot serve as a basis to distinguish the Internet and to allow us to recognize impacts on it, specifically because its focus solely on universal identifiers within the IP RFC is inadequate for understanding how IP packets make internetworking possible.
Okay: only somewhat redundant. But much more direct and clear.
On Thu, Jan 31, 2013 at 5:28 PM, Seth Johnson
> Okay, I really addressed the FNC definition along the way in the last
> post, though I did not provide references to the definition itself. I
> just need to add a couple of notes on that, and I think I have
> incidentally addressed the E2E point sufficiently, though if prodded I
> could read the treatment of the Myth#2 Patrice gives us regarding the
> assertion that the E2E principle is “essential” to the Internet. That
> treatment has some interesting features, that would be interesting to
> note, but I think the key points I wanted to draw are here. I have
> always found it notable, and interesting in terms of the kind of
> posture CNRI consistently strikes, how much Bob Kahn’s views are
> consistent with what many people see coming from the Internet, but
> maybe they don’t really see how those aspects arise the same way.
> This includes his view on NN, which actually is fully consistent with
> real NN, the NN that naturally results when you have competing
> autonomous networks that must nevertheless interoperate, and that was
> there from the start.
> What follows is just a few short comments on the FNC definition. If I
> post one more, that post will be a synthesis, reducing the verbiage
> and redundancy of my first commentary on the IP-based network position
> of the CNRI, at least as presented in that document, and integrating
> this on the FNC definition with it.
> The FNC Definition, from http://www.nitrd.gov/fnc/Internet_res.aspx,
> is as follows:
> “The Federal Networking Council (FNC) agrees that the following
> language reflects our definition of the term “Internet”.
> “Internet” refers to the global information system that –
> (i) is logically linked together by a globally unique address
> space based on the Internet Protocol (IP) or its subsequent
> (ii) is able to support communications using the Transmission
> Control Protocol/Internet Protocol (TCP/IP) suite or its subsequent
> extensions/follow-ons, and/or other IP-compatible protocols; and
> (iii) provides, uses or makes accessible, either publicly or
> privately, high level services layered on the communications and
> related infrastructure described herein.”
> It defines the Internet in terms of the globally unique addresses,
> providing for the TCP/IP and other protocols that may use IP with a
> provision that says that global information system is able to support
> them. In her comments to the WGIG she posted here, she proposes to
> add a clause that says this global information system not only
> supports “high level services” not only layered on, but also
> “integrated with” the infrastructure
> The only part of the FNC definition that tells us about the role the
> Internet Protocol plays in making networks and other elements
> interoperable, is the universality of the IP addresses that are
> incorporated in the header of the packets defined in the Internet
> The important things to remember about this definition are 1) that it
> was significant in 1995 to declare a definition that recognized the
> TCP/IP protocols and the IP addresses in the definition, simply to
> clearly delineate the Internet from the oft-noted services or circuits
> orientation of the traditional telecom providers; and 2) at the time,
> the fact that there were many independent providers was an underlying
> and unacknowledged premise. At that time, you could readily get a
> block of addresses and become an ISP on land lines, because the
> telecoms were required to lease their lines to you. So the multitude
> of autonomous networks that the Internet Protocol enabled to
> interoperate was a part of the context.
> That is not the context at present, and without identifying the
> function of enabling autonomous networks to interoperate as a key
> characteristic of the Internet, this emphasis on the addresses as what
> collects the system together overlooks the function of
> interoperability and the distinction between autonomous networks that
> raises the problem of interoperability. It therefore lets us lose
> that ability as dominant providers get to treat their intrAnets as if
> they are Internets.
> This definition can’t give us a picture of when the ability to
> interoperate is being affected by specialized functions, particularly
> functions that really aren’t compatible with general purpose
> interoperability, functions which are viable only across routers that
> you control within your own network.