The locality tax: reading is state
A companion paper constructs, in the stable non-closed class of iterated revision, a family of skeletons of clause arity two whose collapse time is 2^((n+1)/2+O(log n)), and explains the exponent by an accounting: a trigger of codimension j must read j coordinates, a width-j conjunction at arity two costs j-1 sentences, and sentences are state. A second paper replaces the arity two in both halves by a parameter k and shows that the tax is exactly 1/k per coordinate, the exponent (k-1)/k being attained from both sides within the detect-and-drain architecture. This note argues that the parameterised statement, and not the number 1/2, is the object worth keeping. Three theses. First, the tax is an exchange rate between sentences and codimension, so the interesting questions are about the rates different architectures offer, not about whether the exponential can be beaten. Second, the accounting bounds designs, not classes: the two maxima are not known to differ in either direction, a search over the class finds instances well above the construction we have, and the literature --- including our own prose until very recently --- has been careless about the difference. Third, the obstruction to beating the tax is dynamical order, an orbit clinging to a cylinder, not sentence counting; the open problem must therefore be posed about orbits and cylinders. We say what the tax does not claim, why the natural attack on it (a family that merges states gradually) is refuted by a three-line packing argument, which escape routes remain, and which five problems would move the programme.
Authors
- Rongjun Feng
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-12
- DOI
- https://doi.org/10.5281/zenodo.24152360
- Primary Topic
- Advanced Graph Theory Research
- Type
- preprint