Multibase
Multibase is the reason CIDv1 strings are DNS-safe lowercase base32 and can therefore appear in subdomain gateway hostnames like `bafy….ipfs.example.com`. If you build URL routers or parsers that touch CIDs, you encode and decode through multibase.
What It Defines
Self-describing base-N encoding format that prepends a single-character prefix indicating which encoding (base16, base32, base58btc, base64url, …) was used. CIDs are emitted as multibase-encoded strings, which is why CIDv1 strings begin with `b` (base32 lowercase) — the prefix tells decoders how to interpret the rest. Multibase makes CIDs self-describing at the textual layer, the same way multicodec and multihash make them self-describing at the binary layer.
Canonical (Normative)
Convenient (Practical)
Open community of implementers (Protocol Labs, Shipyard, Boost, Helia, Kubo, and others) that maintains the InterPlanetary File System spec family at specs.ipfs.tech. Publishes content-addressing primitives (CID, Multihash, Multibase, Multiaddr), the IPLD data model, naming systems (IPNS, DNSLink), and the HTTP Gateway specs (Path, Subdomain, Trustless, DNSLink) that bridge content-addressed storage to ordinary browsers.
Related Specs
Every DNSLink record, IPFS gateway URL, IPNS publish, and Filecoin deal references content by CID. If you do not understand what the bytes after `/ipfs/` mean, you cannot debug deploy/cache issues, validate signed manifests, or design content-addressed pipelines.
Every CID contains a multihash. If you ever need to validate that a fetched IPFS payload matches its CID, swap in a different hash, or implement a custom indexer, you read and emit multihash bytes directly.
Path gateways break the web origin model — every CID shares the gateway origin. Subdomain gateways are the production-correct choice for hosting independent sites. This is also the only model that makes single-page apps with cookies and Service Workers function on a public gateway.