The raw IPFS Merkle DAG has no concept of a file or folder. UnixFS is the format that adds one. Here is exactly how, node type by node type.

No. Unpinning a file on IPFS only removes its protection; the blocks stay until garbage collection runs. Pin types, ipfs repo gc, and unpinning on AIOZ Pin.

Content addressing names data by a hash of its content, not by the server holding it. Here is how IPFS CIDs compare with HTTP location addressing, and why.

IPFS splits content retrieval into two jobs: a DHT finds who has your file, Bitswap fetches it. Here is exactly how provider records and want-lists work.

IPFS chunking splits files into 256 KiB fixed-size blocks by default. Rabin and Buzhash cut by content instead, so an edited file reuses most of its blocks.

A CID (content identifier) is the hash-based address IPFS gives content. Decode one byte by byte: multibase, version, codec, multihash, and v0 to v1 conversion.

IPFS shards a large directory with a HAMT, not a B-tree. Kubo switches once the listing passes 256 KiB, then spreads entries over shards of 256 buckets each.

How does IPFS work? Every file becomes a Merkle DAG of hash-linked blocks, found via a DHT and moved with Bitswap. Here is the full path, block by block.

Resize images using URL parameters on AIOZ Pin: add img-width, img-height, and img-fit to the gateway link. No API key, no second upload. All 11 options.

x402 payments let ERC-8004 weight a paid AI agent interaction more heavily than free work in its reputation score. Here is exactly how that link works.

AIOZ Pin NFT API calls pin an asset and its metadata as two tracked pins under one record. Here is how to automate it directly over REST, no SDK needed.

AI agent reputation data under ERC-8004 is a score on-chain plus a hash-verified reference off-chain. Here is why that reference needs to point at IPFS.