Types of Oracles: Publisher and Reporter Networks

The Pyth Network is a first-of-its-kind data oracle network that introduces a new blockchain oracle design to maximize access to information in a safe way on-chain that strips out middlemen costs. This design enables maximum scalability and growth of on-chain financial market data with exciting potential for off-chain as well.

This blog post explores the implications of Pyth's architecture. In so doing, this article also addresses a common question we receive: How is Pyth different from traditional oracles?

Publisher and Reporter Networks

Pyth is the largest publisher oracle network in the world with over 40 original data sources publishing directly to the network. In this network, the nodes own and publish data directly on-chain.

Most legacy oracles are reporter oracle network where nodes carry data over the last mile from API endpoint to on-chain consumption. In this network, the nodes must purchase the data from first-party sources or other middlemen.

Why is Pyth Designed as a Publisher Network?

When data owners are publishing their data directly on-chain, the network has the lowest possible constraints and is able to charge the lowest prices for the data and to update at the fastest possible intervals. This arrangement gives maximum flexibility in margin compression (the data providers decide whether or not to stream given the returns).

In addition, entirely new features can be introduced that do not exist in current data supply chains. The primary example of this is the Pyth Network’s introduction of confidence intervals which are unique to Pyth.

A Publisher Oracle Network is Not Without Trade-Offs

One of the advantages of reporter networks is that they are much more nimble. Given the adaptability of the independent nodes to fetch unstructured information and then normalize, reporter networks should be faster in expanding to a wider variety of data including generalized queries. This works really well with information that is in the public domain.

While they are more generalized and adaptable, reporter networks are always constrained by the cost and output format of the data they purchase to report. These constraints are economic (the data costs something — middlemen aggregators charge additional fees, and the nodes themselves need to earn something) and non-economic, such that purchasers do not have control of update speeds, or have limited ability to request new features. In fact, many existing data supply chains will not permit their data to be reported to the blockchain because it is not possible to limit the distribution to only paying subscribers.

The biggest challenge with a publisher network is overcoming the initial cold start of onboarding data providers to fill the shelves of the marketplace.

Being the first person in a network carries quite a bit of risk both reputationally as well as technologically. If the network fails to pick up traction, the early participants will have spent political capital and development resources that they could have used elsewhere. Once the supply side is sufficiently robust such that it can comprehensively cover necessary markets, the network needs sufficient demand to encourage further development and incentivize the suppliers.

At some point, a critical mass is reached whereby the risk profile switches such that it is more reputationally or economically damaging to not be a part.

Launching Pyth

When Pyth first launched, in time for the spring Solana Hackathon (May 2021) on devnet, there were 4 data providers announced which were publishing across 30+ symbols covering crypto, equities, FX, and commodities. Like any new network, there were growing pains and it was not unusual for 1 or more publishers to have unscheduled maintenance and the number of providers dropping down to 2 or 3. There were also the standard new technology learning curve quirks along the way, where some publishers did not have optimized confidence intervals.

Over time, as more data providers announced themselves, the reputation risk continued to be reduced for the next data source. Data providers span from regulated US equity exchanges, some of the largest crypto exchanges, and nearly every large trading firm. As each data source gets added, this increases the network’s resiliency. On mainnet beta today, there are 25 independent first-party data sources for the Pyth BTC/USD price. Every day, Pyth becomes more secure because of new data sources getting added as well as integrations incorporating confidence intervals.

Examples in Digital Music

Napster can be described as a reporter network because it involved nodes making music available for digital consumption mostly by performing the last mile task of uploading from CD to digital. Napster created a store where music could be downloaded by anyone (for free, at least initially).

Napster did not work, mostly because they did not have the appropriate licensing for the distribution of its product, so the suppliers eventually cut them off. More accurately, the suppliers spam attacked the network by making the percentage of corrupt music files to un-corrupt files unattractive to customers. Eventually, there was a lawsuit and Napster attempted to convert its business model into a more inclusive one where the record companies received some revenue. This model did not work, partially because they were too early and the record labels were not comfortable doing deals for the bulk sharing of music.

Spotify is an example of a publisher network because it involves nodes publishing music directly to the Spotify platform. It allows publishers to share in the network revenue based on usage. Network contributors are both record labels, as well as artists who choose to publish directly to Spotify and eliminate the dependency on record labels or other middlemen. As a result of the reduced middlemen constraints, Spotify was able to scale.

There is no single right way to run an oracle network.

Pyth differs considerably from all legacy oracle networks by ensuring that the data owners have a stake in the network’s success. Pyth is making the bet that as the DeFi ecosystem grows, the stakes will continue to go higher for the data that is securing the assets. By removing upstream middlemen dependencies, Pyth is designed to scale without pricing or speed constraints, and enable the next generation killer Web3 apps.

We can’t wait to hear what you think! You can join the Pyth Discord and Telegram, and follow us on Twitter. You can also learn more about Pyth here.