Why can’t a blockchain guarantee truthful inputs?

A refrigerated truck submits a signed record saying that a shipment stayed within its required temperature range. A blockchain accepts it. What you can establish first is that the system received and accepted this claim.

Checking the shipment requires more: where the sensor was placed, whether it was calibrated, whether power failed and whether it traveled with these goods. A sensor left in a warehouse could produce correctly signed, accepted records that say nothing about the shipment. This is a hypothetical example of an evidence boundary, not a deployed project.

Who observes the world before data reaches the protocol?

When a smart contract, a program that executes rules on a chain, needs information from outside its chain, an oracle supplies it. The name does not imply that the mechanism knows the truth. It might rely on one source or compare several. Their independence and the handling of disagreements remain design questions.

Shared records can help participants examine who submitted a claim, when it was accepted and whether a correction followed. What can be checked depends on the protocol and retained evidence. Better traceability cannot replace the original measurement.

After hearing “it is on-chain,” ask how the input was produced, how an error would be detected and what evidence would justify a correction.

Book and verification sources

This lesson develops a question from 《区块链实战》, Chapter 6, with the references below. It is a new explanation, not a book excerpt.