Internet relies on a stack of technical layers, each fulfilling a specific role. Understanding how the Internet works means first measuring how protocols, physical cables, and servers cooperate to transport data from one point to another on the planet. What gaps separate these different layers in terms of function, speed, and reliability?
Network protocols and packet routing: what distinguishes each layer
The model that structures the Internet overlays several protocols. Each manages a distinct scope, and confusing them amounts to ignoring where the points of fragility in a connection lie.
| Layer | Main Protocol | Role | Transported Unit |
|---|---|---|---|
| Application | HTTP / HTTPS | Exchange of web pages, API | Requests and responses |
| Transport | TCP / UDP | Reliability (TCP) or speed (UDP) | Segments / datagrams |
| Network (Internet) | IP (IPv4 / IPv6) | Addressing and routing | Packets |
| Link / Physical | Ethernet, Wi-Fi | Transmission over the medium | Frames |
TCP breaks data into packets, assigns them an order number, and then checks upon arrival that nothing is missing. UDP sacrifices this control for the sake of latency, making it suitable for video streaming or online gaming.
Routers read the destination IP address inscribed in each packet and choose the shortest (or least congested) path to the next network. This routing mechanism explains why the Internet has no single center: each router makes a local decision, and the sum of these decisions forms a global routing.
For those who wish to understand the Internet with CGI Network, the principle remains the same: a network of interconnected networks where each node participates in the distribution of information.
DNS and IP addresses: name resolution at the heart of the connection
Typing a URL in a browser triggers a request to a DNS server. This server translates the domain name into an IP address usable by network protocols. Without this step, communication between devices would be limited to manual entry of numerical sequences.
The DNS functions like a distributed directory at multiple levels: root server, top-level domain server (.fr, .com), then authoritative server for the concerned domain. The result is cached locally for a defined period (TTL), reducing the number of repeated requests.
IPv4, IPv6, and address exhaustion
IPv4 uses addresses coded on 32 bits. The number of available addresses is limited, and their exhaustion has led to the gradual deployment of IPv6, which codes addresses on 128 bits. The coexistence of both versions remains the norm on the majority of access networks.
Access providers often assign a shared IPv4 address (via NAT) and a public IPv6 address. This duality sometimes complicates connection diagnostics for users, as a service may work on one protocol and not the other.
Resilience of physical infrastructures and points of vulnerability
The majority of intercontinental traffic passes through undersea fiber optic cables. These cables connect continents and concentrate a considerable volume of digital data. Their damage (ship anchor, earthquake, wear) can temporarily isolate entire regions.
- Ageing undersea cables pose an increasing risk to service continuity, especially in island territories dependent on a limited number of links.
- Network redundancy, often presented as a guarantee, can itself constitute a single point of failure if backup paths use the same physical conduit.
- Data centers (datacenters) host web, DNS, and application servers. Their geographical concentration creates a dependency on a few areas.
A network that is redundant on paper is not always so in practice: two distinct fibers that run in the same trench will fall together in the event of a cut.
NIS2 Directive and cybersecurity obligations on Internet components
The European NIS2 directive has expanded the scope of entities subject to cybersecurity requirements. DNS providers, domain registries, cloud operators, data centers, and content delivery networks (CDN) are now among the covered sectors.
Incident reporting: a constrained timeline
NIS2 imposes a three-step process for significant incidents:
- Early alert within 24 hours after detecting the incident.
- Detailed notification within 72 hours, including an initial assessment of the impact.
- Final report within one month, with the corrective measures applied.
These obligations directly affect the fundamental components of the Internet: DNS, cloud, CDN. In contrast, France had not completed the transposition of NIS2 by the European deadline of October 2024, creating a legal uncertainty for the concerned operators.
Risk management required
The directive requires covered entities to implement risk management that includes business continuity, supply chain security, vulnerability handling, and multi-factor authentication. These requirements apply regardless of the size of the organization as long as it operates a listed service.
The functioning of the Internet depends as much on physical infrastructure as on software protocols and the regulatory framework that governs them. The reliability of the network is at stake at every layer, from the undersea cable to the DNS server, including the reporting obligations imposed on operators. Each absent or failing link is enough to interrupt the chain.
