Shikanoko Networks (SKNK)
Back to blog

SKNK Technical Guide

What Is an IP Prefix? CIDR, Subnets, and BGP Routes Explained

Learn what an IP prefix is, how /24, /48, and /64 work, and why prefixes sit at the centre of CIDR, BGP, ROAs, and route objects.

An IP prefix is a contiguous range of IPv4 or IPv6 addresses written as a network address and prefix length, such as 203.0.113.0/24 or 2001:db8:1234::/48. In BGP, a prefix is the basic unit a network announces to describe reachability for an address range.

Definition Box: An IP prefix is a contiguous block of IPv4 or IPv6 addresses, written as a network address plus a prefix length — 203.0.113.0/24 or 2001:db8:1234::/48, for example. It describes a range, not a single device.

Most people meet the slash for the first time in an ordinary place: on an ASN application form, in an IPv6 allocation email, on a ROA form, or in the output of a route lookup. It looks like a formatting detail. It is not. The number after the slash fixes how large the block is, and that single number changes how routers interpret it, how registries record it, and what you are allowed to do with it. This guide unpacks what a prefix actually contains, how to read the common lengths, and where it fits among subnets, CIDR, BGP, and routing-security records.

1. What an IP Prefix Contains

Every prefix is two pieces of information joined by a slash:

  • The network address — the lowest address in the range, and the address that names the whole block.
  • The prefix length — the number of fixed bits, written after the slash, which sets how many addresses the range covers.

IPv4, for example: 203.0.113.0/24 means the 256 addresses from 203.0.113.0 to 203.0.113.255 (203.0.113.0/24 is TEST-NET-3, a documentation range). IPv6, for example: 2001:db8:1234::/48 means the block starting at 2001:db8:1234:: with the first 48 bits fixed — a range so large that it dwarfs any IPv4 allocation in practical terms (2001:db8::/32 is the IPv6 documentation prefix).

Two implications follow:

  • A prefix is never a single IP address. It is a boundary drawn around a range; one server's address is a single point inside it.
  • The same range can be described at different granularities. 2001:db8:1234::/48 contains many /64 subnets, and every one of those is itself a prefix.

White-Face Xiaolu clamps the leading cells of a long continuous address ribbon, separating the fixed part of a prefix from the open range.

One prefix, one range: the address names the block, the length sets its size.

2. How to Read /24, /48, and /64

Within one IP version, the rule is simple: the shorter the length, the larger the range. Fewer fixed bits means more addresses fit inside.

Example IP version Typical meaning
/24 IPv4 A common small routed IPv4 block — 256 addresses
/48 IPv6 A common organisational IPv6 allocation
/64 IPv6 The standard size for one IPv6 LAN subnet

The /64 row deserves emphasis, because IPv6 behaves differently from IPv4 in one practical way: a normal IPv6 LAN subnet is usually a /64. Stateless address autoconfiguration (SLAAC) is commonly deployed at that size, so operators do not normally subdivide a LAN further. A /48 is a common organisational IPv6 prefix: 65,536 /64 subnets to plan with. IPv4 never gets this luxury — a home or office network is usually a /24 or smaller, because IPv4 addresses are scarce.

A common mistake is comparing 203.0.113.0/24 with 2001:db8:1234::/48 as if the numbers sat on the same scale. They do not. IPv4 and IPv6 are counted in completely different address spaces, so the /24 and the /48 say nothing about each other. If you need the exact maths of splitting IPv6 blocks, IPv6 Subnetting Explained walks through the calculation.

3. IP Address, Prefix, Subnet, and CIDR: What Each Term Means

These four terms appear in the same paragraphs so often that they get used interchangeably. They are not the same thing — four words, four jobs:

Term What it is Example In the global routing table?
IP address One endpoint address — a device or an interface 203.0.113.7 Usually no — unless explicitly announced as a /32 or /128 host route
IP prefix A defined range of addresses 203.0.113.0/24 Only if it is announced
Subnet A prefix used inside an addressing plan 192.168.1.0/24 Usually not
CIDR The addressing and routing model built on prefix-length notation The slash behind the notation Not an entry — a system

The relationship in one line: CIDR is the system that expresses ranges with slash notation; a prefix is one range written that way; a subnet is a prefix put to work inside a network — same notation, different role. A private office LAN at 192.168.1.0/24 is a subnet and, in normal operation, never touches the global routing table. The block your upstream routes for you is a prefix, and it is exactly what BGP talks about.

4. Why Routers Care About Prefixes

Routers do not keep a list of every IP address on Earth — the table would be unmanageable. They forward traffic by matching prefixes: when a packet arrives, the router looks for the most specific covering entry and sends the packet that way.

Two mechanisms make this scale:

  • Route aggregation. An upstream announces one wide block instead of thousands of small ones. Instead of announcing every /64 inside it, a network announces 2001:db8::/32 and lets the more-specific routes appear only where they actually exist.
  • Longest-prefix match. When several entries cover the same address, the longest (most specific) one wins. Suppose 2001:db8::/32 and 2001:db8:1234::/48 are both in the table. A packet to 2001:db8:1234::1 matches both — the router picks the /48, because it pins down a smaller, more precise region.

White-Face Xiaolu opens a gate on the narrow branch of a broad route, sending one packet along the more specific path.

The router reads the prefix length: the more specific entry wins for addresses inside its range.

This is also why the slash carries operational weight. Announce a prefix longer than your authorization allows — a /52 under a /48 ROA with maxLength: /48, say — and ROV can mark the route Invalid; announce only the parent and you give up the granularity longest-prefix match would have used. How announcements travel and get chosen is covered in BGP Basics. This article only needs the principle: routing runs on prefixes, not on address lists.

5. A Prefix Is Not a Route, an ASN, a ROA, or an IRR Route Object

Here is the bridge to the rest of this site's routing cluster — and the most useful table in the article:

Item What it answers What it does not do
Prefix Which address range? Does not create a BGP announcement
ASN Which network identity? Does not grant an address range
BGP route How is the prefix reached? Does not prove authorisation
ROA Which ASN may originate the prefix? Does not announce the route
IRR route object What routing policy data is published? Does not make the route live

Each row is a separate layer, and each layer can be correct or broken on its own. You can hold a prefix, have no ASN, and never announce anything. You can have an ASN, a healthy session, and an accepted route — and still lack a ROA or a route object, which may cost you visibility depending on your upstream's policy. The rows answer different questions, and none of them implies the next.

For the details: What Is an ASN? explains the identity layer, What Is a ROA? the authorization record, and What Is an IRR Route Object? the registry layer that declares or records intended routing intent.

6. When a Prefix Becomes Reachable on the Internet

A prefix sitting in a registry or in a spreadsheet is not on the Internet. It becomes reachable only when a chain of independent conditions is met:

  1. You are authorised to originate it. You hold the address space directly, or a sponsoring LIR provides it — and the ROA and IRR records, if any, align with your plan.
  2. An upstream accepts the announcement. Someone with an established BGP relationship is willing to take the route and propagate it, subject to its filters and policy.
  3. BGP is configured and the route is exported. Your router configuration, not a database, decides what actually leaves your network.
  4. Validation and filtering pass. Depending on the networks in the path, ROA and IRR data may be checked — a mismatch can stop the route even when your session is perfectly healthy.

None of these steps is automatic, and none of them "causes" the next. You can have all the paperwork and no route; you can have a live route and stale paperwork. The records in databases describe intent and authorisation; reachability is a property of the routing system, not of a record.

7. Common Misconceptions

"A /48 is one IP address."

No. A /48 is a range, and in IPv6 an enormous one. Your router's LAN interface uses one address inside it.

"A larger number after the slash means a larger address block."

The opposite. Within one IP version, more fixed bits means a smaller range: /64 is smaller than /48, which is smaller than /32.

"Every prefix needs an ASN."

No. Private ranges — RFC 1918 IPv4, ULA IPv6 — work fine with no ASN. An ASN becomes relevant the moment you want to originate a prefix in BGP.

"A ROA makes a prefix reachable."

No. A ROA is an authorization record: it says who may originate the prefix. The announcement, the session, and the upstream's acceptance are separate steps.

"A route object is the same thing as a BGP route."

No. A route object is registry data that declares or records intended origin; a BGP route is a live announcement carrying the prefix and the path to it. One describes intent, the other is in motion.

8. Do You Need Your Own Prefix?

The honest answer comes in two parts:

  • A normal application on one cloud provider usually does not need its own prefix. The provider's addresses, load balancers, and DNS are the whole story — and a good story. Your own prefix only starts to matter when you want to control addressing and routing yourself.
  • An operator considering its own ASN and routed IPv6 prefix needs a real routing use case — multihoming, announcing from multiple sites, running your own BGP edge, or serving networks that require you to originate your own space — and an upstream that will accept the announcements. That combination is the genuine trigger; without both, everything else is decoration.

If that trigger is real for you, How to Get an IPv6 /48 covers getting the allocation, and How to Announce an IPv6 /48 via BGP covers the launch checklist. One note on roles: SKNK supports the sponsorship and registry workflow, including resource allocation, paperwork, and RPKI ROA management for SKNK PA resources. Upstream selection, BGP configuration, and IRR administration remain the customer's responsibility unless separately agreed in writing.

FAQ

Is an IP prefix the same as a subnet?

Not exactly. A subnet is a prefix used inside an addressing plan. Every subnet is a prefix, but not every prefix is used as a subnet — some prefixes exist only to be routed.

What does /0 mean?

Zero fixed bits: the entire address space. 0.0.0.0/0 is all of IPv4, ::/0 is all of IPv6. You will meet it in default routes and filter rules.

Is /24 bigger than /48?

Not across families. 203.0.113.0/24 and 2001:db8:1234::/48 live in different address spaces, so comparing them by address count is meaningless. Within one family the rule holds: /24 is larger than /32, and /48 is larger than /64.

Does every IP prefix need an ASN?

No. Private use — RFC 1918 IPv4 and IPv6 ULA — needs no ASN at all. An ASN is needed only to originate a prefix in BGP.

Can I announce any prefix through BGP?

No. You can only announce ranges you are authorised to originate, and even then the upstream decides whether to accept them — typically checking ROA, IRR, and its own agreements.

What is the difference between a prefix and a route?

A prefix is an address range — a noun. A route is the prefix plus the path and attributes used to reach it — the thing that is announced, accepted, and propagated. One describes space; the other describes how traffic gets to it.

Key Takeaways

  • A prefix is an address range, written as network address plus prefix length — never a single IP.
  • CIDR is the model behind the slash notation; a subnet is a prefix used inside an addressing plan.
  • Routers and BGP operate on prefixes: aggregation keeps the table small, longest-prefix match picks the winner.
  • An ASN identifies the network; BGP, ROA, and IRR use the prefix for propagation, authorisation, and registry intent.
  • A prefix recorded in a database is not live on the Internet; reachability comes from the routing chain, not from paperwork.

In One Sentence

An IP prefix is the address range at the centre of Internet routing; the rest of the routing stack decides who may announce it and how traffic reaches it.

Continue Reading

References