NAMING SYSTEM LAYER

CROSS-REFERENCE TO RELATED APPLICATION(S)

[0001] This patent application claims the benefit of U. S. Provisional Patent Application Serial Number 60/656,733, which was filed on February 26, 2005, entitled "CoCo Naming System Layer," and is incorporated herein by reference in its entirety.

BACKGROUND

[0002] Computers have been networked to exchange data between them for decades. One network, the Internet, comprises a vast number of computers and computer networks interconnected through communication channels. In general, computers are made aware of other computers by referencing a domain name server or some other server that is aware of a number of interconnected devices. However, these servers require the interconnected computers to be fixed in space. As an example, when a portable computing device moves from one geographical location to another, its Internet Protocol address could change.

BRIEF DESCRIPTION OF THE DRAWINGS

[0003] Figure 1 is a block diagram illustrating layers of a CoCo protocol.

[0004] Figure 2 is a tree diagram illustrating a tree fragment associated with the naming system layer.

[0005] Figure 3 is a block diagram illustrating relationships between application program interfaces of layers associated with the naming system layer.

[0006] Figure 4 is a ring diagram illustrating rings corresponding to names associated with the naming layer.

[0007] Figure 5 is a ring diagram illustrating key-spaces associated with the naming layer.

DETAILED DESCRIPTION

[0008] A naming system facility is provided. In some embodiments, the naming system facility is the highest protocol layer in the CoCo Protocol hierarchy excluding the user application layer, as is illustrated in Figure 1. In broad terms, the naming system is a mechanism that enables nodes to locate each other in a CoCo hierarchical network. The naming system exposes an application program interface to provide access to the naming system's services. The naming system is located "above" the circuit layer because it uses circuits to maintain connections between nodes.

[0009] Figure 1 is a block diagram illustrating layers of a CoCo protocol. The protocol employs several layers 100, including a physical interface layer 102, link layer 104, routing layer 106, circuit layer 108, naming layer 110, and user application layer 112. The naming and circuit layers are described herein.

Addressing

[0010] The described protocol is the logical separation of a device's name from its location. The term "location" (rather than "address") is meant to suggest that it varies during a session as the network dynamically reconfigures. Each CoCo device is provisioned with a name that exists for the lifetime of the device. The name is a hierarchically-structured string resembling an Internet domain name — for example ¹¹ smith. police . Seattle . wa . us . " The location of a device is a numeric quantity, similar to an Internet IP address.

[0011] As IP addresses may be conceptually decomposed into subnet addresses, CoCo network locations have a hierarchical decomposition. There are, however, significant differences between CoCo network locations and IP addresses. Internet servers and devices have fixed IP addresses. Even computers connecting through a dial-up phone line are assigned an IP address that remains fixed for the duration of the session. A CoCo device's location may change over time as the device moves. If nearby devices move, a CoCo device's location may change even if it remains geographically stationary.

1 he Purpose ot the Naming system

[0012] A CoCo device name is a concatenation of dot-delimited strings such as "police . seatt Ie . wa . us , " while a device's network location has the form: [Dn, Dn-1 , . . ., D1 , DO]. CoCo device locations may change over time as they move, or even as devices around them move. The CoCo Protocol Routing Layer RFC explains how a node S routes packets to another node D given its location. Before this can happen, S must first learn the location of D, given its name. The naming system provides a method for doing so: it is a distributed lookup mechanism to obtain a device's name from its location.

[0013] CoCo networks maintain two concepts of node identity. One concept is a name. This name might sometimes be referred to as a friendly name or a network name. This document uses the term name for conciseness. Names are hierarchically structured ("dot"-delimited) ASCII strings. For example, the following might be a name of a node in a CoCo network:

coco .us .Washington. Seattle .police . frank

[0014] The CoCo network node with this name is part of a hierarchy that might be represented by a tree a fragment, as is illustrated in Figure 2.

Naming System Properties

[0015] The naming system is completely decentralized and distributed. It self- generates when a network first forms and it self-adjusts when the network topology changes. DNS, by comparison, requires human intervention to update server IP addresses and much longer delays for such changes to propagate through the network. Names in the CoCo naming system are unforgeable because they are cryptographically secure. Each device may be assigned a name when it is originally provisioned. It is possible to delegate the authority to assign names. For example, the city of Seattle may delegate to the police commissioner the authority to distribute names for the network devices used by members of the Seattle Police department. This is indicated in Figure 2 as node 202 providing authority to node 204.

I ne Naming system interface

[0016] Figure 3 illustrates a detailed view of how the naming situation relates to the other layers in the CoCo Protocol:

[0017] A naming system 302 exposes a simple application program interface to user applications 304. This interface includes a method to set a <name, location> association and another method to obtain the location of a node, given its name:

SetName (name, location) ; location GetLocByName (name) ;

[0018] The naming system, like user applications, uses a circuit layer system 306 calls to open and close circuits and to read from and write to circuits.

CircuitID Open ( ) ; Close ( CircuitlD) ; Packet Read ( CircuitlD) ; Write ( Packet , CircuitlD) ;

The naming system uses circuits to give a more useful response to requests for node locations. One variant of the method location GetLocByName (name) is the method CircuitlD GetLocByName (name) where GetLocByName returns circuit ID to the location that corresponds to name, rather than just location itself. This approach has two advantages:

1. The location of a node may change frequently. Maintaining an open circuit to it exploits the circuit layer's facility of keeping the location of circuit destinations up-to-date.

2. Typically, a user application invokes GetLocByName to establish a circuit to a node. If the naming system maintains an active circuit, the user can more efficiently send the node a request to establish a new circuit to it.

[0019] The second alternative may cause high overhead to maintain all the required open circuits.

i iip I iiGiαioiiiodi OI.IUOI.UIC ui Ii ic iMdi I ill in oysicm

Hierarchical Structure of the Namespace

[0020] The naming system is hierarchical. The discussion above presented the naming system as a single ring for simplicity. In fact, there is a hierarchy of naming rings. To extend the example name "smith. police. seattle.wa. us" introduced earlier, there is a ring for each level of the name. These rings are illustrated in Figure 4.

[0021] According to Figure 4,

• wa hashes to P1 on the us ring

• Seattle hashes to P2 on the wa.us ring

• police hashes to P3 on the Seattle. wa.us ring and

• names of the form xxx . wa . us appear on the wa.us ring

• names of the form xxx. Seattle. wa.us appear on the

Seattle. wa.us ring

• names of the form xxx. police, seat tie. wa.us appear on the police. Seattle . wa.us ring.

[0022] The expanded hierarchical view of the name space in Figure 4 illustrates the dominance of local network traffic and the desirability of a logarithmic distribution of keys in the contact table. It is reasonable to expect, for example, that most of the traffic to and from the device with name smith. police. Seattle. wa.us would involve devices with names of the form x. police. Seattle, wa.us more than with devices with names of the form x. police, beij ing. china.

The Contact Table

[0023] Each node maintains a contact table of other selected network nodes. The contact tables of all the nodes in the network are, in effect, a distributed database.

[0024] Table entries have the form <key, location> where key is the hash of the name of a node, and location is that node's network location. In practice, the

location field may be a circuit table entry for a circuit to that location, ratner than the location itself.

[0025] The key field is ordered so that a metric, or distance, may be defined on the set of keys. A convenient metric defines the distance between two keys as the arithmetic difference of the bit patterns of the keys, viewed as integers modulo the size of the key space. It is convenient to view the key-space as a ring, as in Figure 5.

[0026] Initially, the table for a node contains its own key-location pair. As a node becomes aware of other nodes, it adds them to its contact table. It adds nodes selectively, so that the keys of the table entries are distributed logarithmically through the key space, as is depicted in Figure 5. Using a logarithmic distribution keeps contact tables from outgrowing reasonable space constraints. The basis for a logarithmic distribution is a scalable trade-off between memory consumption and bandwidth usage. When a node A needs the location of another node B, it computes B's key and sends a location lookup request to the node in its contact table whose key is closest to the key of B.

[0027] To make the logarithmic distribution precise, let K = hash (N) . Then K is a value in the key-space which is the set of all 32-bit values. For each value d between 0 and the key-space size, let N _d be the number of entries in the contact list for N whose keys are a distance less than d from K = hash(N) . Then N _d has the following property:

such that|AαsA (N)-λαsλ(N')|< c log d

where c is a constant and the subtraction is performed modulo the size of the keyspace. In essence, this inequality states that the distribution of the keys of the entries in a node N's contact list occur increasingly sparsely at greater distances from the hash value of N .

[0028] Contact lists have another property. The logarithmic distribution of the equation, makes no requirement about the precise location on the key ring of any nodes in the contact list. However, for a contact list to be valid and fully functional, it should be in a state of convergence. The next section describes this property.

Naming System Convergence

[0029] The state of the naming system is converged if the contact table of each node contains that node's two nearest neighbors (with respect to the metric). To achieve convergence, the naming system uses a distributed message passing algorithm. This algorithm employs a primitive called introduce, which has the format introduce ( source , dest , new) .

[0030] The effect of introduce is to make the node dest aware of new's name and location. The node source "introduces" new to dest by sending a message to dest over a permanent circuit. IN some embodiments, each node maintains dedicated circuits to the nodes in its contact table. Node dest may add new to its contact list if doing so improves the approximation of logarithmic distribution in the list.

[0031] The convergence algorithm uses a series of introduce messages among selected nodes in the network. A node X that wishes to join the network initially detects the presence of another node Y and establishes a link to it, as described previously. To integrate itself into the naming system, node X sends an introduce message to introduce itself to Y. Node Y then finds the node in its contact table closest to X and clockwise from Y. This causes a limited cascade of messages from one node to another, proceeding in a clockwise sequence around the ring with each node in the sequence being introduced to X. When the process ends, X knows the identity of its immediate clockwise and counterclockwise neighbors.

Primitives and Algorithms for Naming System Convergence [0032] The naming system can utilize the following primitives.

node NextClockwise (Node X) , node NextAntiClockwise (Node X)

[0033] This method returns the node whose key is closest to X's key in either the clockwise or counterclockwise direction.

Introduce (Node source, Node dest , Node new)

[0034] Node source introduces dest and new to each other by sending a message to dest. This message is sent over a dedicated circuit. Node dest may

then choose to add new to its contact list. It will typically do so, unless its contact list is already full, and correctly (logarithmically) distributed.

[0035] A AddToContactList (Node source) method causes a node to add source to its contact list.

[0036] Using these primitives, the following algorithm provides a simplified implementation of convergence in various embodiments.

JoinNetwork (node X) // X calls this on itself to join network

{

Y = node whose presence the X's link layer detects introduce (source = X, dest = Y/ about X) // X introduces itself to Y }

// performed by dest when it receives introduce (source, dest, about) onRecievelntroduction (from, new)

{ if (contact list becomes more logarithmically distributed)

AddToContactList (from)

Y = NextClockwise (new) if Y < from / / in this context , from is X; " < " = " clockwise of " introduce ( self , from, Y) }

Managing A Densely-Populated Broadcast Communication Medium

[0037] A second application of the algorithms used for the CoCo Naming Layer is to manage a densely-populated broadcast communication medium. By creating a Contact Table populated with the transport-specific node addresses and executing the Naming System Convergence algorithm, we gain strong link security in trade for changing network efficiency from linear to logarithmic-scale. Utilizing a second Contact Table of most recently contacted destinations, the node can increase best- case efficiency to match the underlying network. Because network nodes tend to communicate with extremely regular local-routing behavior, the average case is similar to the best case. Thus, for relatively small efficiency cost in most situations, we avoid entirely a class of network security failings.

[0038] The CoCo Protocol provides a communications technology with several innovations: greatly increased security; flexible support for a wide range of Quality of Service (QoS) requirements; interoperability with standard transport mechanisms such as WiFi, cellular, Ethernet, and satellite links; and a novel addressing system that enables scalability in an environment consisting of a large number of mobile devices.

[0039] The following publications are incorporated herein by reference in their entireties:

1. J. Bruestle, P. Mark, CoCo Protocol Architecture White Paper, CoCo Communications Corp., 2004.

2. J. Bruestle, R. Eller, P. Mark, CoCo Protocol Link Layer RFC, CoCo Communications Corp., 2004.

[0040] Coco has filed the following patent applications, the disclosures of which are incorporated herein by reference in their entireties:

1. U.S. Patent Application Serial No. 10/542,824, filed on July 20, 2005.

2. U.S. Provisional Patent Application Serial No. 60/640,810, filed on December 29, 2004.

3. U.S. Provisional Patent Application Serial No. 60/655,808, filed on February 23, 2005.

[0041] A method performed by a computer system for publishing information in a distributed network without a central management infrastructure is described. In various embodiments, the method receives an indication of a new node and a destination node, the new node omitted from a contact list associated with the destination node, the contact list having an approximately logarithmic distribution of neighboring nodes; introduces the new node to the destination node via a permanent circuit; and causes the destination node to add the new node to the contact list when adding the new node improves the logarithmic distribution of neighboring nodes.

[0042] From the foregoing, it will be appreciated that specific embodiments of the invention have been described herein for purposes of illustration, but that various

modifications may be made without deviating from the spirit and scope of the invention. Accordingly, the invention is not limited except as by the appended claims.