Lemma 1.2 — duality:
ProvedKServerConj.lemma_1_2competitive-analysisk-serveronline-algorithmsp2o-gran-per-chapterp2o-plan-paperp2o-strengthenedp2o-v1work-function-algorithm
Let be a finite pseudometric space, , a configuration of points, requests and . Let be the work function before the request and the work function after it. Then
where the maximum and both minima range over all configurations of points.
The extended cost of a request is thus the increase of the quantity ; the potential of the paper is built to combine these quantities for all . This is the duality lemma of Koutsoupias and Papadimitriou, expressed in terms of minimum values.
Differences from the source:
- generalization: a metric space (M, d) → a pseudometric space (
PseudoMetricSpace M); reason: the paper, p.1: "our proof works also when d is a pseudometric"; every metric space is a pseudometric space - encoding: w a work function and w′ the work function after a request at r → w = w_{t−1} and w′ = w_t for some C₀, r and t ≥ 1, with r = r_t; reason: every work function is w_{t−1} for some initial configuration and request sequence
- encoding: k servers and a configuration C₀ → hypotheses 1 ≤ k and C₀.card = k; reason: configurations are k-element sets
- encoding: max and min over all configurations → iSup/iInf over the finite nonempty type of k-element subsets; reason: a supremum over a finite nonempty type is attained
Preamble
import Mathlib import Definitions.Def_KServerConj_extCost
Formal statement
namespace KServerConj
/-- **Lemma 1.2 (Duality).** Coester, Koutsoupias, Zbysiński, *The k-server conjecture is true*,
arXiv:2609.15979 (version of 2026-11-03, sha256 7ca0380a…), printed p.2 (PDF p.3): "Let w be a
work function and let w′ be the work function after a request at r. The extended cost is
max_X {w′(X) − w(X)} = min_Y {w′(Y) − ∑_{y∈Y} d(r, y)} − min_Y {w(Y) − ∑_{y∈Y} d(r, y)}, (4)
where the maximum and both minima range over all configurations."
**Formalization Note.** A work function is `w = w_{t−1}` for some initial configuration `C₀` of `k`
points and some request sequence `r`, and `w′ = w_t` is the work function after the request
`r = r_t`, for `t ≥ 1`. `M` is finite (standing assumption, printed p.1); the maximum is `extCost`
and the minima are infima over the finite nonempty type of `k`-point configurations. -/
theorem lemma_1_2 {M : Type*} [PseudoMetricSpace M] [Fintype M] (k : ℕ) (hk : 1 ≤ k)
(C₀ : Finset M) (hC₀ : C₀.card = k) (r : ℕ → M) (t : ℕ) (ht : 1 ≤ t) :
extCost k C₀ r t =
(⨅ Y : {Y : Finset M // Y.card = k},
(workFn k C₀ r t (Y : Finset M) - ∑ y ∈ (Y : Finset M), dist (r t) y)) -
(⨅ Y : {Y : Finset M // Y.card = k},
(workFn k C₀ r (t - 1) (Y : Finset M) - ∑ y ∈ (Y : Finset M), dist (r t) y)) := by sorry
end KServerConj
Source
Coester, Koutsoupias, Zbysiński, The k-server conjecture is true, arXiv:2609.15979 (version of 2026-11-03, sha256 7ca0380a…), Lemma 1.2, p.2 (pdf p.3)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.