Summarized from Github Specification
Self-describing content-addressed identifiers for distributed systems
Basically a hash with some metadata. CID is a self-describing format for referencing content, it is a form of content addressed storage.
Format: <cidv1> ::= <multibase-prefix><multicodec-cidv1><multicodec-content-type><multihash-content-address>
Where
<multibase-prefix>is a multibase code (1 or 2 bytes), to ease encoding CIDs into various bases. NOTE: Binary (not text-based) protocols and formats may omit the multibase prefix when the encoding is unambiguous.<multicodec-cidv1>is a multicodec representing the version of CID, here for upgradability purposes.<multicodec-content-type>is a multicodec code representing the content type or format of the data being addressed.<multihash-content-address>is a multihash value, representing the cryptographic hash of the content being addressed. Multihash enables CIDs to use many different cryptographic hash function, for upgradability and protocol agility purposes.
IPVM
Brooklyn Zelenka from Fission Codes on IPVM
CID-based computation also means that we can use memoization to inform us if an operation has been run before so we can optimize our efforts and copy the CIDs of those outputs into our work, saving time and compute power.
Acyclicality
Graphs that use CIDs for references are acyclical! Hashing a cycle would mean that you need to know the CID of the contents without actually traversing its contents.
This is impossible! Consider a cycle A -> B -> C -> A. To figure out the CID of A, we need to know the CID of B. To know the CID of B you need to know the CID of C. C’s CID needs to know the CID of A and we are back where we started.
Side note: I guess this could be done by brute forcing a hash collision but this is so statistically improbable we might as we well consider it impossible