TLDR: We propose an imperfect availability scheme for large data. It is intended as a stopgap for TrueBit-like computational markets until fully decentralised availability solutions for large data are ready.
Whenever a task giver supplies a TrueBit task with large data
D (large program or large inputs) the task giver does the following:
Don IPFS, referenced by its hash
- Produces a short Oraclize proof
Dis available on some
Uonchain for verifiers
Verifiers will be able to retrieve
ipfs.iois a faithful IPFS gateway
archive.orgis a faithful public and permanent historical mirror
For extra reliability more than one IPFS gateway (or other service that maps files to hashes) can be used simultaneously, as well as more than one storage provider.
Various bells and whistles can be used, e.g. SNARKs to remove the need for
ipfs.io or to encrypt
U for stealthiness.