TL;DR
This paper proposes a new network architecture called the Identifier-Locator Network Protocol (ILNP) that separates network addresses into distinct "Identifiers" (for end-to-end identity) and "Locators" (for routing). This separation aims to elegantly unify mobility, multi-homing, and Network Address Translation (NAT) support, while also enhancing security, without requiring fundamental changes to the core Internet routing infrastructure.
What Problem Does This Paper Address?
The internet's current addressing scheme, where a single IP address serves both as a unique identifier for a node and a locator for its network position, creates significant challenges. Existing solutions for mobility (like Mobile IP), multi-homing (connecting to multiple networks), and Network Address Translation (NAT) are often complex, poorly integrated, and not widely deployed. This architectural limitation hinders the development of new network capabilities and makes it difficult to seamlessly integrate these essential features. For instance, Mobile IP's reliance on a "home agent" and "triangle routing" adds complexity, while NAT breaks many application protocols that embed IP addresses. The paper argues that the current IP address is overloaded with too many responsibilities, leading to these integration and deployment issues.
Methods
The core of the proposed solution is the Identifier-Locator Network Protocol (ILNP). ILNP introduces a fundamental architectural shift by splitting the traditional IP address into two distinct components:
- Identifier (I): A stable, non-topological name that uniquely identifies a node or service end-to-end. This is analogous to a domain name but operates at the network layer.
- Locator (L): A dynamic, topological name that indicates the current network location of a node. This is akin to an IP address prefix used for routing.
The paper proposes an instance of this architecture called ILNPv6, which adapts IPv6's 128-bit address space by dedicating 64 bits to the Locator and 64 bits to the Identifier. This separation allows:
- Mobility: A node can change its Locator (network location) without affecting its stable Identifier, meaning upper-layer protocols (like TCP) remain unaware of the change. This eliminates the need for complex Mobile IP mechanisms like home agents and triangle routing.
- Multi-homing: A node or site can have multiple Locators simultaneously, representing different network connections. ILNP can manage these multiple Locators seamlessly.
- NAT Tolerance: Since NAT primarily manipulates the Locator part of the address, and upper-layer protocols are bound to the Identifier, ILNP is inherently more tolerant of NAT. The paper suggests NAT might even be used for traffic engineering purposes with ILNP.
- Security: Security associations (like those used in IPsec) can be bound to the stable Identifier, making them independent of network location changes. This simplifies security management for mobile and multi-homed nodes.
The paper also proposes a new networking API that abstracts away the complexities of IP addresses, focusing instead on domain names and service names, further simplifying application development and promoting compatibility with ILNP.
Datasets
Not discussed in the paper.
Key Insights and Ideas
- The "aha moment" is the separation of concerns: The fundamental insight is that by cleanly separating the "what" (Identifier) from the "where" (Locator), many of the Internet's persistent problems with mobility, multi-homing, and NAT can be solved elegantly and integratedly.
- Mobility, Multi-homing, and NAT are fundamentally the same problem: The paper argues that with a proper Identifier/Locator split, the mechanisms required to handle mobility and multi-homing become unified, simplifying network design and operation.
- Incremental deployability is key: The proposed ILNPv6 is designed to be backward compatible with existing IPv6 networks, allowing for gradual adoption without a disruptive "flag day" transition.
- Simplifying application development: A new API that abstracts network-layer details and focuses on domain names can make it easier for developers to create network-aware applications, especially those that need to be mobile or multi-homed.
- Identifier as a stable anchor: The stable Identifier provides a consistent point of reference for end-to-end communication, security associations, and application state, regardless of network location.
Key Results
The paper doesn't present quantitative results from a deployed system but rather outlines an architectural proposal and its expected benefits. The key "results" are the conceptual advantages and engineering feasibility demonstrated:
- Unified solution: ILNP provides a single architectural framework to address mobility, multi-homing, and NAT tolerance simultaneously.
- Simplified mobility: Eliminates the need for complex Mobile IP agents and triangle routing, allowing nodes to move seamlessly.
- Improved NAT compatibility: Protocols that embed IP addresses are less likely to break with ILNP, as NAT primarily affects the Locator.
- Enhanced security integration: IPsec can bind security associations to stable Identifiers, simplifying security for mobile nodes.
- Incremental deployment: ILNPv6 can be deployed incrementally on existing IPv6 networks.
Limitations and Future Work
- Operational Scalability and Performance: The authors acknowledge that further work is needed to address operational scalability and performance optimization of ILNP.
- Implementation: A proof-of-concept implementation is planned to demonstrate the proposal's efficacy and identify practical challenges.
- DNS Enhancements: ILNP relies on enhanced DNS functionalities (like DNSSEC and Dynamic DNS) for mobility support, which are not yet universally deployed.
- New API Adoption: The success of the proposed new networking API depends on its adoption by application developers.
- Network Realms and Interworking: While ILNP can handle NAT gracefully, interworking with diverse "network realms" (e.g., non-IP networks) might still require middleboxes or gateways.
- Specific Engineering Details: While the paper provides architectural details, specific engineering challenges for implementing ILNPv6 in existing systems (like kernel modifications) are still under investigation.
- Anycasting: Anycasting is mentioned as a subject for future study within the ILNP framework.
Summary updated