centurymetadata.org: Long-term Bitcoin Metadata Storage
Century Metadata is a project to provide storage for small amounts of auxiliary data. As an example, this is useful for Bitcoin wallets, which can be restored from 12 seed words, but cannot know about more complex funds without additional data. On restore, your wallet would attempt to fetch this data from https://centurymetadata.org or a mirror.
We are currently in alpha, seeking feedback.
The file format is designed to be self-explanatory and use standard, long-lived primitives as much as possible. Every file contains a preamble, followed by 8192 bytes. The preamble describes the data format which follows:
centurymetadata v0\0SIG[64]|WRITER[33]|READER[33]|GEN[8]|AES[8054]
SIG: BIP-340 SHA256(TAG|TAG|WRITER|READER|GEN|AES)
WRITER, READER: secp256k1 x-only keys
TAG: SHA256("centurymetadata v0"[18])
AESKEY: SHA256(EC Diffie-Hellman of WRITER,READER)
AES: CTR mode (starting 0, nonce 0) using AESKEY of DATA
DATA: gzip([TITLE\0CONTENTS\0]+), padded with 0 bytes to 8054\0
The data itself is a series of NUL-separated title, contents pairs. Obviously this cannot be validated on the production server, but the test server (which only allows known keys) will check the file is compliant.
The BIP 32 path recommended for centurymetadata is 0x44315441'
(DATA), with /0' as the writer key,
/1' as the reader key. Of course, others can also send data
to your reader key, but you know that the record from your own writer
key can be trusted.
The types of records accepted are as follows:
bitcoin psbt, Body: base64-encoded PSBTbitcoin transaction Body: hex-encoded transactionbitcoin miniscript Body: miniscript stringThe test API endpoint can be found at testapi.centurymetadata.org.
Content-Type: application/x-centurymetadata should contain a valid
centurymetadata file.
Since we bundle records by reader prefix (e.g. all readers starting with 42a3 might be bundled together), you need to know how long the prefix is: it starts as an empty prefix and increases by one hex digit as we grow, so bundles are always a reasonable size.
Returns a JSON object with member depth containing how many hex digits of reader to use for fetchbundle.
Content-Type: application/x-centurymetadata, consisting of multiple back-to-back
century metadata files.
There is an experimental Python package to encode and decode centurymetadata files in the GitHub repository