Blockchains benefit from guarantees of immutability and reliability due to their high redundancy and distributed nature. They show their value especially when operating between untrusted parties. Their functionality can be extended programmatically by smart contracts, but are limited by high costs of on-chain computation and only being able to truly trust data which is directly included on-chain. To attempt to bridge this limitation, blockchain oracles are introduced as a conceptual solution to act as a trusted source of information within the blockchain. The Oracle Problem emerges as we consider how one can introduce trusted information into a trust-free environment without compromising the validity of the blockchain. Many promising designs for oracle mechanisms have been proposed, but it is not readily apparent how one should assess the applicability of a given mechanism, nor the strengths and features between mechanisms. To be equipped to assess and categorize oracles, we must consider not just the possible answers, but the questions to which these oracles are trying to speak. Categorizing questions by their possible answering populations, we propose a framework for considering oracle questions and the context with which they are posed. We observe that there are limitations to what an oracle can hope to achieve, depending on the nature of the question, while noting the context in which a question exists can change what is viewed as true.