Community Content Networking

TCP/IP vs. Content-Centric Networking

Our current internet structure is built on transmission control protocol/internet protocol, or TCP/IP[i]. TCP/IP is a layered stack of protocols for communication between computers and connected devices. The TCP/IP stack has four layers, from the bottom up: link, internet, transport, and application. To pass data through these layers is the fundamental design and purpose of the internet. The Open Systems Interconnection, or OSI, protocol stack has seven layers. The layer most users are familiar with is the application layer, where the world wide web (WWW) is located, as it’s an application built on the internet layering protocol stack. ?

Data carried across the internet is organized into packets. Each packet has delivery instructions included in the header. The basic unit of transaction in a connectionless service in a packet-switching network such as the internet is the datagram, which is composed of a header and a payload section[ii].

In this stack, layer 2 is segmented for just the internet protocol (IP) layer. The internet protocol specifies how packet headers are formatted and resulting instructions for routers. Internet protocol delivers packets from the source host to the destination host using the IP addresses contained in the packet headers. The dominant version of the internet protocol in use today is IPv4, although IPv6 has been undergoing extensive rollout for some time.

You’re likely familiar with your IP address, or at least intuitively the concept of them. These addresses define important location information for how to find the content you’re looking for. When you visit Netflix, for instance, your computer, the client, sends a data packet through the TCP/IP stack and addresses Netflix’s server at its location via the IP address and routing information contained in the packet header. If the message is valid, the Netflix server will respond with the content you’re looking for.

In our current internet paradigm, location, or IP address, is everything. Packets are routed based on their IP addresses; a crowd of networked computers attempting to access the same resource on a server can cause latency and bandwidth issues by overloading the server with too many requests at once. Data, it is said, is a second-class citizen in this paradigm—second to content location.

Information-centric networking (or content-centric, named data networking (NDN), etc.) fundamentally changes our internet paradigm by replacing the internet protocol naming and addressing scheme with a direct naming scheme. Rather than a packet being addressed with its IP address for finding host locations, data becomes a first-class citizen of the net, routed by the name of the content itself in the header, which can be anything desired, depending on the naming scheme chosen and its accompanying semantics.

Common content-centric networking (CCN) setups include:

·????? Pending interest table (PIT)

·????? Forwarding Information Base (FIB)

·????? Content Store (CS)

A user sends an Interest packet to a router. The router checks its Content Store for data that matches—if it finds it, it returns the requested content to the requestor. If it doesn’t, it records the requestor in the Pending Interest Table’s entry. The router forwards the Interest packet on based on information in the Forwarding Information Base. Once the content is located that satisfies the entry in the PIT, it sends the content back and satisfies all the PIT entries for that content.

The important thing to note in this paradigm is that packet information, both Interest and Data, do not contain any host or location-identifying information. Routers forward interest packets based solely on the name in the packet header, and data packets are forwarded based solely on the PIT state information.

One can see that with names becoming a first-class identifier for a wealth of packets (and remember that every piece of data on the internet is exchanged via packets), a naming convention is essential, especially one that makes sense to both producers and consumers of content. Although CCN and NDN do not provide rules for allocating names or defining namespaces, this is our principle interest for the purpose of community content networking.

Community Content Networking

Part of my attraction to the new data-centric naming and routing paradigm of future internet proposals is the ability for communities and networks to define their own spaces and data on a deeper, maybe even more thoughtful level. In this proposal, not only are names for data and packets allowed in the application layer, but they are, or can be, contingent on feedback and input from users with a variety of feedback mechanisms.

I’ve long dreamed of folksonomies, or user-generated taxonomies and controlled vocabularies, being used for actual internet naming and routing. For instance, numerous sites let you tag articles, pictures, blogs, etc., and these tags can then be used across that site or platform’s ecosystem for the discovery and categorization of content. In a community content networking paradigm, users can actually tag webpages themselves and have the packet names be stored in the Content Store or entered in the Pending Interest Table as these names, which can consider symbolic names that locate physical items. Call it folksonomic content routing, if you’d like.

Part of the attraction is the ability to scale a naming scheme exponentially using the power of human computation and decentralized naming consensus. Another attraction is the various mechanisms that can be used for consumers and users to participate in content networking naming, and how these shared nomenclatures and terms can come to define community content itself.

Semantically interlinked online communities, or SIOC (pronounced shock), are a semantic social web construction for interlinking user vocabulary and content stores and links. Community content networking might expand this concept, allowing diverse and connected online communities to set the mechanisms and vocabulary by which the content they use and consume is named and entered.

Of course, such a system is ripe for abuse. It would likely work best with self-certifying names based on cryptographic signatures, and with the removal of IP address and location information from packets, it might prove acceptable to have any contributions to data and packet naming only done by verified users, such as through self-sovereign identity (SSI) credentials or a web of trust system.

Just like with our internet now, words are immensely valuable. Websites are defined by their keywords, phrases, content, and countless other factors. SEO is a huge business; content writing is a massive one. Ecommerce and digital advertising, based on interests formulated through words and queries and clicks, remains one of the largest and most lucrative industries in the history of the planet (including lucrative keyword auctions on Google). So the words that will be used to define packets and their storage in routers will likely prove very important, and potentially very valuable, owing to their scarcity of naming and identifying.

While it’s an economically attractive proposition to imagine using combinatorial auctions, algorithmic bidding languages, or other market mechanisms for naming packet contents, there’s the (justified) concern that the rich will get richer based on their ability to game the system and purchase the most valuable names for content networking. Community consensus or voting mechanisms, secured through one-person-one-vote with SSI and registered on a blockchain, might prove more useful than pure market mechanisms. We might see splintered content communities and community routers that segment communities based on their words and naming nomenclature in packets, such as alternative DNS systems that often toy with the idea of creating different root zone files for the internet as a whole.

It would also be possible to incentivize user tags and semantic contributions with paid tokens or digital currencies on blockchain, if desired. A class of experimental methodologies called semantic acquisition games show promise for users to find terms based on the gamification of word choice and usage. In the future, it might be that word games are a profitable, useful mechanism for naming the data packets making up our internet.

Community content networking, to use the library parlance of Elaine Svenonius[iii], is based on user warrant rather than literary warrant. In a literary warrant, scholars or librarians set terms and define semantics based on the accepted nomenclature of their industry. In a use warrant, it’s based on user semantics and community feedback. Stevens and Alemu’s An Emergent Theory of Library Metadata encourages bottom-up, user-generated categorizing and library naming and referencing schemes based on user feedback mechanisms of enrich-then-filter: ask for community tags, and then refine them to the final set that defines your resources.

It's interesting to see the return of named content as an important resource in the internet scheme. Ted Nelson’s Xanadu project has been a long-running attempt at creating what we would consider an alternative to the web we use today. In Xanadu, content is paid for; it’s indexed based on Tumblers, or transfinite numbers (for its addressing scheme); and it’s based on user and community-defined notions of content and content sharing with defined and variable names.

As we look to the potential future of the new internet, perhaps we went so far as to come back again, a loop. But either way, it’s becoming increasingly clear that the future internet will have a large place for user-defined values and segmented community spaces that allow networking based on use warrants and shared language.?

?

[i] Van Schewick, Barbara. 2010. Internet Architecture and Innovation. The MIT Press

[ii] Clark, David D. 2018. Designing an Internet. The MIT Press.

[iii] Svenonius, Elaine. 2000. The Intellectual Foundation of Information Organization. The MIT Press.