Summary

The Reputation System (RepSys) is a collection of tools and services that provide relevant information to Storage Providers and Clients to make Filecoin dealmaking more successful. This document covers the overall goals and design of RepSys, and tracks related PL workstreams.

Goals

1. Improve overall Filecoin network reliability by incentivizing good behavior for SPs offering storage on the network
   • SPs reliably offer storage and retrieval dealmaking, and clients have the data necessary to make informed decisions about which SPs they choose to work with
   • There is a feedback loop for clients to capture experiences with SPs, which includes interactions that are not already captured directly on-chain or via the Protocol (i.e., conversations on Slack, compliance with regulatory policies, etc.)
   • Use DataCap (FIL+) as the incentive mechanism for good SP behavior
2. A diversity of multiple users of RepSys at large
   • Different clients have different needs, and will care about different aspects of a reputation score. For example, large clients will dig a lot deeper in to the SP's client relationships and focus on reviews, whereas long-tail clients may care more about average latency and price. A variety of producers, consumers, and RSVs should emerge to meet the needs of the market.
3. Educate the market on good behavior
   • Filecoin is a complex network, and understanding the factors that comprise good behavior is not always intuitive.
   • Help define and instill good behavior for all market participants.

KPIs

Design

The current design for the reputation system includes the following core components:

• (Reputation) Data Producers - Services that add data to the reputation datastore. This can be data, metadata, quantitative, or qualitative. One instance of a data producer is a dealbot, which attempts to make storage and retrieval deals on the network, publishing statistics of those deals to the reputation datastore.
• (Reputation) Data Consumers - Clients that read data from the reputation datastore. Clients can be anything from a web site, storage intermediary, to a lotus instance.
• Reputation Service (RSV) - A service that combines a variety of data sources (from the datastore, or elsewhere) to produce reputation scores for storage providers. These scores will be published to the reputation datastore. filrep.io is an example of a reputation service. An RSV will be both a producer and consumer of the reputation datastore. The distinction is that the RSV must produce reputations scores, at a minimum.
• Reputation Datastore, aka Pando - a sidechain which serves as the reputation datastore for the RepSys
  • Producers can input their data onto Pando.
  • Consumers can read data from Pando to inform dealmaking decisions on the Filecoin network.
  • All data entries must be cryptographically signed via public keys to verify authenticity.

We are currently working towards the canonical RepSys (MVP) for the Filecoin Network, which is made up of the collection of all RSVs and other components that interact within it.

• There is a single sidechain datastore, Pando, which makes all the data available from RSVs, dealbots, and other producers of deal/SP reliability data, for RSVs or other services
• There can be any number of RSVs, which are community-run and independent in their methodologies, features, etc. filrep.io will be the canonical RSV implementation.
  • RSVs address two problems:
    1. Help clients/deal brokers identify storage providers on the network that meet their needs
    2. Close the feedback loop between clients and storage providers, thereby incentivizing storage providers to serve the needs of clients in the market
• Dealbots (and reputation data providers) - RSVs may choose to run their own dealbots services (leveraging the open source Dealbots Project) and/or consume data from existing dealbots maintained by other entities in the Filecoin ecosystem