content addressing

This document defines how content is addressed in the Aevia protocol. Every canonical reference to bytes in the protocol MUST use a Content Identifier (CID) as specified here. Opaque URLs are NOT a canonical reference.

MUST, SHOULD, MAY follow RFC 2119.

Aevia adopts CIDv1 as per Multiformats. For any object O, its CID is:

where:

• the "b" prefix identifies multibase base32 without padding per the Multibase spec;
• codec is the multicodec varint identifying the byte type (see §3);
• 0x12 is the multihash identifier for SHA-256;
• 0x20 is the digest length in bytes (32);
• SHA-256(O) is the SHA-256 digest of the object bytes.

Implementations MUST reject CIDs that do not start with "b" or that use a multihash other than SHA-256 under this RFC.

Given a CID x and a byte sequence O, a verifier MUST:

1. decode x extracting (codec, digest_algo, digest_len, digest);
2. verify that digest_algo == 0x12 (SHA-256);
3. verify that digest_len == 0x20;
4. compute SHA-256(O);
5. compare the 32 bytes of the digest; accept only if identical.

A mismatch at any step MUST result in rejecting the bytes. The client SHOULD try another peer, not re-request from the same peer with the same bytes.

CID retrieval is location-independent. Any node may satisfy the request; the protocol does not privilege any origin. Clients SHOULD:

• query FindProviders(cid) in the DHT to obtain the candidate peer set;
• request the bytes in parallel from multiple candidates, accepting the first response that passes §4;
• discard invalid responses with no retry against the same peer;
• cache the (cid → bytes) mapping locally for at least the session duration.

Clients MAY use HTTP gateways as fallback when libp2p is unavailable. Gateways MUST return the exact bytes; clients MUST verify by CID even when fetching via gateway.

• SHA-256 is considered secure for content addressing over the horizon of this RFC. Transitions to SHA-3 or BLAKE3 MUST be preceded by a new revision.
• retrieval timing attack: returning “I don’t have this CID” may leak cache patterns. Provider nodes SHOULD respond in constant time to unknown CID requests.
• malicious HTTP gateways may return wrong bytes; §4 guarantees clients are not fooled, but gateways MAY be used for traffic analysis. Clients SHOULD rotate gateways.