Protocol Scope Limitations in Network Systems

17 concepts15 papers1 proposal

In an ideal world, network measurement tools, security architectures, and testbeds would be completely protocol-agnostic. They would observe or route traffic seamlessly, whether it consists of standard TCP streams, UDP datagrams, low-level ICMP diagnostics, or specialized routing protocols like MPLS.

In reality, network systems are bound by strict protocol scope limitations. When engineers deploy software into production environments, they frequently encounter hard boundaries imposed by security sandboxes, legacy infrastructure, or hardware design complexities. Most often, these constraints force systems into a "TCP-only" or "IP-only" silo. Understanding why these limitations arise—and how they create critical blind spots—is essential for designing resilient network infrastructure.

Why Systems Face Protocol Scope Limitations

Protocol limitations are rarely a deliberate design choice; instead, they are a byproduct of the environment in which a system is forced to operate. We can trace these constraints to three primary root causes.

1. Sandboxed APIs and Access Restrictions

To measure the internet at scale, researchers want to run code directly on end-user devices. However, modern client environments are heavily sandboxed to protect user privacy and security.

In Opportunities and Challenges of Ad-based Measurements from the Edge of the Network, Patricia Callejo et al. demonstrate a method for executing network measurements globally by embedding code within online advertisement creatives. While this approach provides massive geographic reach, the authors explain that the execution is strictly bound by browser JavaScript APIs. Because browsers do not expose low-level raw socket access, the framework is fundamentally blocked from performing direct UDP or TCP socket manipulations, limiting the types of network diagnostics it can run.

A similar barrier exists on mobile platforms. As Viet-Hoang Tran et al. observe in Voice-Activated Applications and Multipath TCP: A Good Match?, public mobile testbeds typically deny root or kernel access to experimental applications. Without this access, researchers cannot deploy custom network stacks or modify protocols directly within the host operating system, restricting experiments to standard, high-level application layers.

2. Infrastructure and Tunneling Constraints

Even when researchers control the measurement nodes, the intermediate network infrastructure can strip away protocol capabilities.

For example, in MOR: Monitoring and Measurements through the Onion Router, Demetres Antoniades et al. build a distributed network monitoring tool by leveraging the Tor anonymity network. However, because Tor is designed to multiplex traffic exclusively over stateful TCP streams, the entire measurement framework becomes constrained to TCP. It cannot emit ICMP pings or custom raw packets, which restricts the depth of network diagnostics the tool can perform.

Similarly, when connecting disparate research environments, engineers rely on tunneling. In A wired-wireless testbed architecture for network layer experimentation based on ORBIT and VINI, George C. Hadjichristofi et al. note that standard IP-only tunneling solutions make it exceptionally difficult to evaluate next-generation protocols that rely on non-IP or layer-2 broadcast mechanisms.

3. Hardware Offloading and State Complexity

As network speeds climb, systems must offload protocol processing from the CPU to dedicated hardware or automated software synthesizers. However, tracking the state of complex protocols in hardware introduces severe engineering bottlenecks.

In Beehive: A Flexible Network Stack for Direct-Attached Accelerators, Katie Lim et al. design an open-source, FPGA-based network stack. To maintain high throughput, the initial implementation restricts its protocol support to core functionalities like Ethernet, IP, and UDP, leaving complex behaviors like IP fragmentation for future development. Furthermore, the authors note that while their hardware TCP engine handles full bandwidth across active connections, it initially lacks advanced features like connection initiation and congestion control due to the complexity of managing state transitions in hardware.

This pattern repeats across high-performance networking architectures. In High-throughput and Flexible Host Networking for Accelerated Computing, Athinagoras Skiadopoulos et al. introduce ZeroNIC to decouple data and control paths for GPU computing. Although the architecture is theoretically protocol-agnostic, the practical prototype defaults to the standard Linux kernel TCP stack, leaving user-space and alternative protocol optimizations as future engineering milestones.

Even in software-defined configuration tools, protocol support lags behind theory. The configuration synthesis tool presented by Ahmed El-Hassany et al. in NetComplete: Practical Network-Wide Configuration Synthesis with Autocompletion automates network setups but initially excludes advanced mechanisms like MPLS and route redistribution due to the sheer complexity of modeling those protocol behaviors symbolically.

The Cost of Protocol Blindness

When a system is constrained to a single protocol family, it suffers from "protocol blindness." This limitation creates significant gaps in both network visibility and security enforcement.

Measurement Blind Spots

If an internet measurement framework only tracks TCP traffic, its conclusions are inherently incomplete. In Lost in Space: Improving Inference of IPv4 Address Space Utilization, Alberto Dainotti et al. demonstrate that passive analysis focused primarily on TCP traffic misses a substantial portion of active internet addresses. To achieve a comprehensive understanding of address space utilization, integrating UDP and other protocols is a strict necessity.

Protocol scope limitations also hinder the detection of network neutrality violations. In FairNet: A Measurement Framework for Traffic Discrimination Detection on the Internet, Vinod S. Khandkar et al. design a framework to catch ISPs that throttle specific streaming services. The system relies on parsing the Server Name Indication (SNI) field inside TLS handshakes over connection-oriented protocols like TCP. However, because it depends on this specific protocol structure, the authors note that FairNet cannot be directly applied to services running DTLS over UDP, creating a clear bypass opportunity for traffic discriminators.

Testing and Security Gaps

Security and testing frameworks frequently struggle to adapt when deployed outside of standard enterprise TCP/IP environments.


Bypassing Constraints with Algorithmic Innovation

When the deployment environment restricts your protocol toolkit, how do you make progress? The most elegant solutions in networking do not try to force a protocol into an environment that rejects it. Instead, they change the mathematical framing of the problem to extract deeper insights from the limited protocols available.

A classic example of this shift can be found in Network Tomography—the science of inferring internal link loss rates using only end-to-end path measurements.

    [Source]
       |
    [Link 1]
       |
    (Router)
    /      \
[Link 2]  [Link 3]
  /          \
[Dest A]    [Dest B]

Historically, theoretical approaches to network tomography relied heavily on multicast traffic. If a source sends a multicast packet to Dest A and Dest B, the packet travels down the shared Link 1 as a single copy before splitting at the router. If the packet drops on Link 1, both destinations miss it simultaneously. This shared fate creates a strong statistical correlation that makes calculating the loss rate of Link 1 mathematically straightforward.

However, as Hung Xuan Nguyen et al. point out in Network loss inference with second order statistics of end-to-end flows, multicast is not universally supported across the public internet due to deployment constraints. A tool that requires multicast is practically useless for wide-area internet diagnostics.

Instead of trying to force routers to support multicast, the authors changed the mathematical approach. They proved that by analyzing the second-order statistics (the variances and covariances) of standard, independent unicast flows over time, they could uniquely identify individual link performance.

Because unicast packets do not share a single physical packet structure like multicast, they do not have instant "shared fate" correlations. However, if Link 1 experiences a period of high congestion, the loss rates of independent unicast packets passing through it during that window will covary. By tracking these second-order statistical relationships over time, the authors' framework reconstructs the internal topology's health using standard, universally supported unicast probes.

This breakthrough highlights the ultimate lesson of protocol scope limitations: when an environment restricts your protocol access, the solution often lies in extracting deeper statistical or contextual insights from the protocols you can use.

Key Foundational Work

MOR: Monitoring and Measurements through the Onion Router (2010) — contributed 1 concept elements to this direction.
Network loss inference with second order statistics of end-to-end flows (2007) — contributed 1 concept elements to this direction.
A wired-wireless testbed architecture for network layer experimentation based on ORBIT and VINI (2007) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_protocol_scope_limitation_cb5e95ec,
  title = {protocol scope limitation},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/protocol-scope-limitation/rdr_cb5e95ec921b45129a46d6c3180bbc19},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_protocol_scope_limitation_cb5e95ec_v1,
  title = {protocol scope limitation},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/protocol-scope-limitation/rdr_cb5e95ec921b45129a46d6c3180bbc19/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake