LIPIcs.OPODIS.2024.19.pdf
- Filesize: 0.91 MB
- 17 pages
The concept of extension-based proofs models the idea of a valency argument which is widely used in distributed computing. Extension-based proofs have been shown to be limited in power: there is no extension-based proof of the impossibility of a wait-free protocol for (n,k)-set agreement among n > k ≥ 2 processes. Previous work used a restricted class of reductions to show that there are no extension-based proofs of the impossibility of wait-free protocols for some other distributed computing problems. It is known that for a restricted class of reductions, if a task 𝒯 reduces to 𝒮 and 𝒯 has an augmented extension-based proof that it is impossible to solve in the NIS model, then so does 𝒮. We introduce multiple-instance extension-based proofs and show that, if 𝒯 reduces to multiple instances of 𝒮, instead of just one instance and 𝒯 has an augmented extension-based proof, then 𝒮 has a multiple-instance extension-based proof that it is impossible to solve in the NIIS model. We introduce a new version of extension-based proofs that can further our understanding of extension-based proofs and their limitations.
Feedback for Dagstuhl Publishing