It is sometimes good to visualize a moment of silence before making a decision.
Consider a lending protocol in the middle of the night. Markets are not sleeping but peaceful. Prices change, reports are updated, reserves are changed somewhere off-chain. Smart contracts are ready on-chain. They do not guess. They do not pause. They do not do anything unless something tells them to do it.
That “something” is an oracle.
On the simplest level, an oracle links blockchains to the external world. It introduces prices, reports, and signals which cannot be obtained by smart contracts. However, any person who has dealt with decentralized systems is aware that this connection is not always straightforward. Data can arrive late. Sources can disagree. A value can be right in itself but lead to an issue when applied at the wrong time.
Here, oracle design silently turns into a matter of opinion, not only delivery.
Transport is easy. Trust is not.
Majority of the population visualizes oracles as pipes. Information is fed in at one end and is fed out at the other. Practically that pipe consists of two very different parts.
The first part is transport. This is the technical procedure of retrieving data off-chain and providing it on-chain. This aspect is quite dependable with modern infrastructure. Monitoring helps. Redundancy helps. Failure in this case is normally conspicuous.
The second section is verification and here things become slow and interesting.
Verification does not concern itself with the existence of a number. It is whether that number ought to be permitted to be altered state at the present time. Was the source qualified then? Is the time of the day recent enough to use this? Is the data still reasonable in terms of the current liquidity and market conditions?
These questions cannot be answered by a smart contract. The oracle system should offer a structure of doing so.
APRO's hybrid approach
APRO employs a so-called hybrid architecture. Part of the processing occurs off-chain, with data able to be collected, verified, and contrasted effectively. The findings are subsequently checked on-chain, where transparency and finality are important.
This division is feasible. Off-chain systems are efficient in complexity and volume. On-chain systems implement regulations after decisions are reached. Combined, they enable protocols to be fed actual-world information without assuming that all judgments should occur within a smart contract.
APRO allows both data push and data pull models. In one instance, contracts demand data when required. In the other, information is presented at specified times. Both models are not universally superior. There are applications that require urgency. Still others appreciate speed less than consistency. What is significant is that the protocol is aware of what model it is basing on and the reason.
Where verification is to be found.
One is tempted to believe that after the data has been checked by an oracle network, the issue is resolved. As a matter of fact, verification does not stop at the oracle layer.
An oracle may give confidence indicators, timestamps and consensus results. Nonetheless, it is the protocol that determines what to do with them. A single application can stop when the confidence decreases marginally. The other one can record the incident and move on. Before taking action, a third may need two consecutive updates.
All these options are not wrong in themselves. The issues are problematic when they are not consistent.
This contradiction develops silently. One threshold may be applied to governance logic. Another may be applied by liquidation logic. Keeper bots may impose their safety margins since they do not entirely believe the defaults. With time, the system acts differently according to the path activated.
It is arbitrary when the users notice. On the inside, it seems that all is going on.
The risks that remain
There is no oracle system that eliminates risk. Hybrid models come with their trade-offs. Off-chain processing involves a belief in infrastructure and operations. On-chain verification may be inflexible in terms of subtlety. Validation with the aid of AI filters noise, but may cloud the decision-making process when audit trails are not transparent.
There is also a human risk. Teams can over depend on defaults. Abstraction simplifies system construction, but it may blur responsibility. When there is a conflict and the sole reason is that it is what the oracle said and the reason is that, then verification has been postponed not specified.
Another risk is markets themselves. Liquidity which is thin, slow reporting, or fast-evolving conditions may create technically valid but practically unsafe data. Oracles are able to bring out these signals, but not make decisions on their own. That remains the responsibility of the protocol.
An unobtrusive infrastructure.
Oracle systems are visible when they are working well. Nobody rejoices on a price update that did not create an issue. Nobody observes a collateral rule that evolved in a fair and smooth manner. The success case is the lack of drama.
The design of APRO is indicative of this fact. It is not so much about spectacle as about structure. Claims are introduced into the system through transport. Verification establishes the circumstances in which such claims can be significant.
It is between those steps that trust is built or lost.
And in decentralized systems, speed is seldom the basis of building trust. It is constructed through explicit rules, regular conduct and the discipline to determine not only what data comes, but why it is permitted to behave.
