Pith. sign in

REVIEW 4 major objections 7 minor 30 references

ESBT bounds sequence-CRDT identifiers with a Stern–Brocot hierarchy so concurrent editors stay fast and compact without tombstones.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-31 17:47 UTC pith:7TOGVTDI

load-bearing objection Solid engineering attack on Logoot/LSEQ identifier bloat with a clear layered allocator and strong harness numbers; proofs and baselines are the soft underbelly, not the core idea. the 4 major comments →

arxiv 2607.28101 v1 pith:7TOGVTDI submitted 2026-07-30 cs.DC

ESBT: A Scalable and Deterministic Sequence CRDT for Distributed Collaborative Editing

Moulay Driss Mechaoui , Abdessamad Imine This is my paper
classification cs.DC
keywords collaborative editingsequence CRDTidentifier allocationStern–Brocot treeStrong Eventual Consistencyscalable distributed systemsRed–Black tree
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Long-running collaborative editors that use sequence CRDTs often slow down and bloat memory because position identifiers grow without bound under concurrent inserts, especially when many people type in the same spot. This paper claims that extending the classical Stern–Brocot tree of rational fractions with a small stack of fallback layers—sequence numbers, short integer paths, and a site id—plus a fixed bound on fraction size yields unique, totally ordered, dense identifiers that stay compact. Documents are stored in a red–black tree keyed by those identifiers, so inserts, deletes, and lookups stay logarithmic, and deletes free identifiers instead of leaving tombstones. A lightweight per-insert counter tracks only direct insert–delete dependence, avoiding replica-sized vector clocks. On workloads of up to 100,000 operations across 50 sites, the authors report large gains in responsiveness and identifier memory versus Logoot and LSEQ, including under adversarial middle insertions. If the claim holds, large-scale decentralized text editing can keep Strong Eventual Consistency without the usual identifier-growth tax.

Core claim

The Extended Stern–Brocot Tree (ESBT) produces unique, totally ordered sequence identifiers by taking mediants of neighboring fractions until a configurable size bound is hit, then disambiguating with sequence numbers, a lexicographic sequence path, and a site identifier, so identifier growth stays bounded while deterministic order and Strong Eventual Consistency are preserved, without tombstones.

What carries the argument

ESBT weight: the quadruple ⟨fraction f, sequence number sn, sequence path sp, site id δ⟩ ordered hierarchically (f, then sn, then sp, then δ), allocated by CREATE_WEIGHT between adjacent weights and integrated in a red–black tree.

Load-bearing premise

The fixed knobs for fraction bound, digit base, and path depth, plus the local tracker and path fallback, keep identifiers compact under real long-lived concurrency, and the authors’ matched Java reimplementations of the baselines fairly represent those systems’ costs.

What would settle it

Run the same 50-site pure-insert and mixed workloads (including sustained middle inserts beyond 10k ops) with production Logoot/LSEQ and ESBT under identical conditions: if ESBT’s responsiveness and identifier-memory advantages disappear, or identifiers still explode past the claimed bounds, the central scalability claim fails.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Collaborative editors can drop tombstone GC for sequences while still reclaiming deleted positions.
  • Identifier comparison and storage cost need not grow with session length under localized concurrent typing.
  • Causal correctness for insert/delete can be kept with a single counter per operation instead of replica-sized clocks.
  • The same allocator can key ordered structures beyond plain text (e.g., hierarchical or graph-backed documents) under SEC.
  • Adversarial midpoint-insert workloads no longer force heap blow-ups that halt baseline variable-id CRDTs.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If path serialization elides the default empty path, end-insert memory parity with Logoot/LSEQ should improve without changing the allocator.
  • Adaptive D_max tuned to live contention could shrink the rare deep-path cases the fixed depth-256 cap is meant to contain.
  • The same bounded mediant-plus-layers pattern may transfer to ordered CRDTs outside text (priority queues, timelines) where dense unique keys are the bottleneck.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 7 minor

Summary. The paper proposes ESBT, a sequence CRDT whose position identifiers are composite weights ⟨f, sn, sp, δ⟩ built from a Stern–Brocot mediant fraction f (capped by a bound D_max), a local sequence number sn, a lexicographic sequence path sp (NEWSEQ), and a site id δ. Document state is stored in a Red–Black tree ordered by these weights, yielding O(log n) insert/delete/lookup. Deletions are tombstone-free; a per-insertion monotonic counter carried on deletes distinguishes weight reuse and enforces insert→delete causality without vector clocks. The authors give uniqueness/total-order/convergence arguments as proof sketches (§6, Thms. 2–5) and report Java microbenchmarks (≤100k ops, 50 sites) against reimplemented Logoot and LSEQ (Boundary/Random/Mixed), claiming 28–88% (pure insert) and 59–74% (mixed) responsiveness gains, 50–75% smaller identifiers on beginning/random patterns, and ~86.5% / ~92.8% gains on a 10k middle-insertion workload.

Significance. Unbounded identifier growth under localized concurrency is a genuine, well-known scalability limit of Logoot/LSEQ-style allocators, so a deterministic hierarchical scheme that keeps the rational component bounded while remaining tombstone-free and SEC is a worthwhile systems contribution. The Stern–Brocot grounding, the explicit CREATE_WEIGHT/NEWSEQ algorithms, the lightweight dependency counter, and the consistent experimental trends across insertion patterns are concrete strengths. If the formal uniqueness argument under concurrent fraction/sn exhaustion and weight reuse can be completed, and if the latency/memory gains survive comparison to stronger modern baselines (RGA/Fugue/Eg-walker) and released artifacts, ESBT would be a solid incremental advance for large-scale collaborative editing and related ordered CRDTs. As written, significance is conditional on closing those gaps.

major comments (4)
  1. [§6 Theorems 2–5; also §5.2–5.3, Alg. 1–2] §6, Theorems 2–5 are proof sketches only. The load-bearing uniqueness claim (Thm. 4 / Thm. 1) must discharge concurrent multi-site exhaustion of sn for the same fallback fraction, interaction of the local Tracker (Def. 4) with remote allocations the site has not yet seen, NEWSEQ’s midpoint and site-id fallback (Alg. 1) under simultaneous path extension, and the (w,c) reuse counter of §5.2–5.3 after delete-then-reinsert at the same neighbors. Without a fuller argument (or machine-checked proof) that no two distinct insertions can produce the same ⟨f,sn,sp,δ⟩ under asynchronous delivery and weight reuse, the SEC/convergence claim is not established at the level expected for a CRDT correctness section.
  2. [§8.2–8.3; Table 3; abstract] The headline performance claims (§8.2–8.3; abstract: 28–88% / 59–74% / 86.53% time, 50–75% / 92.81% ID memory) rest solely on from-scratch Java reimplementations of Logoot and LSEQ. Related work and Table 3 discuss RGA, Fugue, and Eg-walker, yet none appear in the evaluation. Middle-insertion is capped at 10k ops because baselines OOM (§8.2.1, §8.3), so the strongest adversarial comparison is on a truncated workload. No code, traces, or harness are released. These choices make the “addresses the principal scalability limitations” claim overstated relative to the evidence; at minimum the evaluation should add at least one tombstone-based and one recent hybrid baseline, report full middle-insert curves for ESBT alone beyond 10k, and release artifacts for independent verification.
  3. [§8.1 Tables 1–2; Lemma 1; §3.2 Bounded Allocation Policy] §8.1 tunes base = 2^31−1 and depth = 256 on the worst-case middle-insertion workload, then freezes those knobs for all comparisons. Table 2 shows average/max path depths of ~1613/3226 after 100k middle inserts—so “bounded identifier growth” applies strictly to the fraction component (Lemma 1 / D_max), not to total identifier size. The paper should (i) state the boundedness claim precisely (fraction vs full weight), (ii) report sensitivity of latency and memory to D_max, BASE, and DEPTH on the same workloads used for claims, and (iii) discuss long-session behavior when sp paths become deep under sustained midpoint contention. Without that, the central scalability narrative over-claims what the hierarchical design guarantees.
  4. [§4.2, §5.3 Alg. 3, Table 3] §4–5 assert tombstone-free deletion with safe weight reallocation, yet the document is a Red–Black tree of live elements plus a delete log L and CounterMap. The paper does not specify bounds on L, when (w,c) entries may be GC’d, or how a late-joining replica reconstructs enough dependency state to apply buffered deletes correctly under the epidemic model of §4. If L or pending queue Q can grow without bound under churn, the “lightweight / low metadata” advantage claimed in Table 3 and the introduction is not fully substantiated. A short liveness/GC argument or experimental metadata footprint under join/leave would close this.
minor comments (7)
  1. [Title page] Title and running header contain a line-break typo: “Distri buted”. Fix throughout.
  2. [§8 Figures 4–8] Figure 4–8 captions and legends mix LSEQ strategy names (ALTERNATE/HYBRID vs Boundary/Random/Mixed). Align legend labels with the strategies named in §8.2.
  3. [§8 figures] Identifier size axes are labeled “Identifiers Size (Byte)” while the text discusses MB; clarify units consistently (Fig. 5, 7, 8).
  4. [§3.2.2 and passim] §3.2.2 “In contract” → “In contrast”; several Unicode/encoding artifacts appear (e.g., )uni2205(var), /u1D45B-style fragments in the source) and should be cleaned for production.
  5. [Table 3] Complexity column in Table 3 writes O(log N) for ESBT local/remote but O(n)/O(n log n) style entries for others; state clearly whether n is document size and whether remote cost includes network or only integration.
  6. [§9] Related work omits direct quantitative comparison discussion with Fugue and Eg-walker beyond the qualitative table; a short paragraph on why their trade-offs differ (fixed vs variable IDs, DAG growth) would help readers place ESBT.
  7. [Abstract; §8.3] Abstract reports precise percentages (86.53%, 92.81%) that should be tied to an explicit baseline choice (which LSEQ variant / Logoot) in the text of §8.3.

Circularity Check

0 steps flagged

No significant circularity: ESBT correctness follows from an explicit hierarchical construction and proof sketches; performance claims are empirical benchmarks, not fitted inputs relabeled as predictions.

full rationale

This is a systems/algorithm paper. Identifier uniqueness, total order, and SEC are argued from the ESBT weight definition (fraction, sn, path, site id), the CREATE_WEIGHT/NEWSEQ allocation rules, and short proof sketches in §6—not from a quantity that is algebraically identical to a fitted target. Allocator knobs (D_max, base 2^31−1, depth 256) are calibrated on worst-case middle insertions in §8.1 and then frozen for baseline comparisons; that is ordinary engineering parameter selection, not a first-principles “prediction” forced by the fit. Classical Stern–Brocot citations and Logoot/LSEQ baselines are external. Self-citations (e.g., MobiRDF) are peripheral and not load-bearing for the central claims. No self-definitional loop, no uniqueness theorem imported from the authors as an external fact, and no renaming of a known empirical law presented as derivation. Residual risks (proof sketches only, reimplemented baselines) are correctness/evaluation concerns, not circularity.

Axiom & Free-Parameter Ledger

3 free parameters · 6 axioms · 3 invented entities

Load-bearing content is mostly standard math (Stern–Brocot mediants, total orders, Red–Black bounds) plus domain CRDT/SEC assumptions and a few paper-specific allocator constants and structures. No new physical entities; invented pieces are data-structure constructs whose only evidence is the paper’s algorithms and microbenchmarks.

free parameters (3)
  • D_max (fraction numerator/denominator bound) = configuration-dependent; illustrated with examples such as 5 and 10
    Hard threshold that stops mediant refinement and forces sn/sp layers; chosen by design and used throughout allocation and Lemma 1 / Theorem 2.
  • sequence-path base (BASE) = 2^31 - 1 (final experiments)
    Digit alphabet size for NEWSEQ; selected after memory sweep on worst-case middle inserts (§8.1 Table 1).
  • maximum sequence-path depth (DEPTH) = 256 (final experiments); observed max depth up to 3226 in unconstrained worst case
    Caps path growth / fallback; chosen after depth statistics on middle inserts (§8.1 Table 2).
axioms (6)
  • standard math Stern–Brocot mediants of adjacent reduced positive rationals are unique, strictly between parents, and dense under repeated refinement.
    Foundation of the fraction layer (§2.2, §3.2.1); classical tree properties.
  • domain assumption Strong Eventual Consistency: replicas that delivered the same set of commuting updates agree, independent of delivery order.
    Correctness target stated in intro/background and Theorem 5.
  • domain assumption Reliable epidemic/broadcast delivery eventually delivers each update to all sites; sites may join/leave.
    Coordination model §4; required for convergence claims under asynchrony.
  • standard math Red–Black tree operations are O(log n) for search/insert/delete under the ESBT total order.
    Document representation §4.1 and complexity §7.
  • ad hoc to paper Pairing each insertion with a local monotonic counter and carrying that counter on deletes suffices to distinguish reused weights and enforce insert→delete dependency without vector clocks.
    Sync layer §5.2–5.3 and Algorithm 3; load-bearing for tombstone-free reuse safety.
  • ad hoc to paper When fraction and sn collide, NEWSEQ’s midpoint/lexicographic construction always yields a finite path strictly between neighbors (with site-id fallback).
    Algorithm 1 and uniqueness Theorem 1/4; argued by construction, not machine-checked.
invented entities (3)
  • ESBT weight ⟨f, sn, sp, δ⟩ no independent evidence
    purpose: Composite totally ordered position identifier with bounded primary fraction and hierarchical disambiguation.
    Core postulated structure (Definition 1); evidence is algorithmic + experimental only.
  • Site Tracker map f ↦ (sn_L, sn_R) no independent evidence
    purpose: Local left/right sequence-number allocation when mediants exceed D_max.
    Definition 4 / CREATE_WEIGHT; paper-internal mechanism.
  • NEWSEQ sequence-path allocator no independent evidence
    purpose: Lexicographic path between neighbors when sn space is exhausted.
    Algorithm 1; last-resort uniqueness layer.

pith-pipeline@v1.2.0-daily-grok45 · 37136 in / 3920 out tokens · 83328 ms · 2026-07-31T17:47:57.219238+00:00 · methodology

0 comments
read the original abstract

Modern collaborative editing systems require efficient mechanisms for managing concurrent updates across distributed replicas. Sequence Conflict-free Replicated Data Types (CRDTs) have become the de facto standard for supporting decentralized collaboration; they enable decentralized replicas to apply operations in arbitrary order while converging to a common state. Although existing sequence CRDTs guarantee Strong Eventual Consistency (SEC), they often suffer from uncontrolled identifier growth and increasing memory consumption during long-running, highly concurrent editing sessions, which severely limits their scalability and performance. This paper presents the Extended Stern-Brocot Tree (ESBT), a mathematically grounded identifier allocation scheme for distributed collaborative text editing, providing a dense, deterministic, and compact identifier space. The proposed allocation strategy bounds identifier growth while preserving deterministic ordering and Strong Eventual Consistency. Experimental evaluation using workloads of up to 100,000 concurrent operations generated across 50 collaborating sites shows that ESBT improves responsiveness by 28 to 88% under pure insertions and 59 to 74% under mixed insertion/deletion workloads, while reducing identifier memory consumption by 50 to 75% in beginning and random insertion patterns compared with the best-performing baseline sequence CRDTs (Logoot and LSEQ). Under the adversarial middle-insertion workload (10,000 operations), ESBT further improves responsiveness by 86.53% and reduces identifier size by 92.81%. These results demonstrate that ESBT effectively addresses the principal scalability limitations of existing sequence CRDTs and provides an efficient and scalable foundation for large-scale collaborative editing systems.

Figures

Figures reproduced from arXiv: 2607.28101 by Abdessamad Imine, Moulay Driss Mechaoui.

Figure 1
Figure 1. Figure 1: The Stern–Brocot tree structure [13]. recursively as depicted in [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The Extended Stern–Brocot tree structure (with = 5 ). of the weight space and avoids further disambiguation. Example : Let 1 = ( 1 3 , 0, [0], 1 ) , 2 = ( 1 2 , 0, [0], 1 ) , and max = 10. The computed mediant is : 1+1 3+2 = 2 5 . Since (2 < max) and (5 < max), the mediant is valid. The function returns the new weight: ( 2 5 , 0, [0],1 ). Fallback Fraction Selection (Lines 13–16) If the mediant exceeds max… view at source ↗
Figure 3
Figure 3. Figure 3: Consistency Issues Scenarios again would be redundant. To address this issue, each user maintains a local of delete operations. Incoming delete operations targeting previously deleted weights are recog￾nized and safely ignored, guaranteeing idempotent deletion and replica convergence. Scenario 3: Reinsertion at the Same Position [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Responsiveness time (ms) under different insertion patterns and 100% insertion workload. 1 2 3 4 5 6 7 8 9 10 x 104 0 0.875 1.75 2.625 3.5 x 106 Number of Operations Identifiers Size (Byte) LSEQ−ALTERNATE LSEQ−HYBRID LOGOOT LSEQ−RANDOM ESBT (a) Beginning insertion pattern. 1 2 3 4 5 6 7 8 9 10 x 104 0 0.5 1 1.5 2 x 106 Number of Operations Identifiers Size (Byte) LSEQ−Boundary LSEQ−Random LOGOOT LSEQ−Mixed… view at source ↗
Figure 5
Figure 5. Figure 5: Identifier size (MB) under different insertion patterns and 100% insertion workload. lowest execution time, remaining close to 56 ms through￾out the experiment. The Random insertion pattern (Fig￾ure 6(c)) remains the most demanding scenario. Logoot, LSEQ-Random, and LSEQ-Mixed require approximately 400 ms after 100 000 operations, while ESBT requires only about 107 ms, corresponding to execution-time reduc… view at source ↗
Figure 6
Figure 6. Figure 6: Responsiveness time (ms) under different insertion patterns and Mixed workloads. 1 2 3 4 5 6 7 8 9 10 x 104 0 0.5 1 1.5 2 x 106 Number of Operations Identifiers Size (Byte) LSEQ−ALTERNATE LSEQ−HYBRID LOGOOT LSEQ−RANDOM ESBT (a) Beginning 80% insertion pattern. 1 2 3 4 5 6 7 8 9 10 x 104 0 3 6 9 12 x 105 Number of Operations Identifiers Size (Byte) LSEQ−Boundary LSEQ−Random LOGOOT LSEQ−Mixed ESBT (b) End 80… view at source ↗
Figure 7
Figure 7. Figure 7: Identifier size (MB) under different insertion patterns and Mixed workload. Compared with the baseline approaches, ESBT reduces ex￾ecution time by approximately 86 to 93%, corresponding to a speedup ranging from 7× to 14×, depending on the allocation strategy. b) Identifier Memory Footprint [PITH_FULL_IMAGE:figures/full_fig_p015_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: ESBT and baseline CRDTs performance under a middle-insertion pattern with pure insertion workloads. 86.5% and identifier memory consumption by up to 92.8% relative to the evaluated baselines. Finally, ESBT can be naturally extended to support very large-scale collaborative editing environments. The current implementation of frac￾tion uses 32-bit integers, which are sufficient for the work￾loads evaluated i… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

30 extracted references · 9 canonical work pages

  1. [1]

    Time, clocks, and the ordering of events in a distributed system

    Leslie Lamport. Time, clocks, and the ordering of events in a distributed system. Commun. ACM , 21(7):558–565, 1978. doi: 10.1145/359545.359563

    arXiv 1978

  2. [2]

    Virtual time and global states of di stributed systems

    Friedemann Mattern. Virtual time and global states of di stributed systems. In Cosnard M. et al., editor, Proc. Workshop on Parallel and Distributed Algorithms, pages 215–226, North-Holland, 1989

    1989

  3. [3]

    Formal design and verification of operational transf ormation algorithms for copies convergence

    Abdessamad Imine, Michael Rusinowitch, Gérald Oster, a nd Pascal Molli. Formal design and verification of operational transf ormation algorithms for copies convergence. Theoretical Computer Science , 351(2):167–183, 2006. doi: https://doi.org/10.1016/j.t cs.2005.09. 066

    doi:10.1016/j.t 2006

  4. [4]

    Logoot: A scalable optimistic replication algorithm for collaborative editi ng on p2p net- works

    Stéphane Weiss, Pascal Urso, and Pascal Molli. Logoot: A scalable optimistic replication algorithm for collaborative editi ng on p2p net- works. In ICDCS, pages 404–412, 2009

    2009

  5. [5]

    Approaches to conflict-free repli cated data types

    Paulo Sérgio Almeida. Approaches to conflict-free repli cated data types. ACM Comput. Surv., 57(2), 2024. doi: 10.1145/3695249

    doi:10.1145/3695249 2024

  6. [6]

    Silva, and Nuno P reguiça

    António Barreto, Hervé Paulino, João A. Silva, and Nuno P reguiça. Ps-crdts: Crdts in highly volatile environments. Future Generation Computer Systems, 141:755–767, 2023. doi: https://doi.org/10.1016/ j.future.2022.12.013

    2023

  7. [7]

    Victor B. F. Gomes, Martin Kleppmann, Dominic P. Mulliga n, and Alastair R. Beresford. Verifying strong eventual consiste ncy in distributed systems. Proc. ACM Program. Lang., 1(OOPSLA), 2017. doi: 10.1145/3133933

    doi:10.1145/3133933 2017

  8. [8]

    Data consistency for P2P collaborative editing

    Gérald Oster, Pascal Urso, Pascal Molli, and Abdessamad Imine. Data consistency for P2P collaborative editing. In Proceedings of the 2006 ACM Conference on Computer Supported Cooperative Work, CSC W 2006, Banff, Alberta, Canada , pages 259–268. ACM, 2006. doi: 10.1145/1180875.1180916

    arXiv 2006

  9. [9]

    Replicated abstract data types: Building blocks for collab orative applications

    Hyun-Gul Roh, Myeongjae Jeon, Jinsoo Kim, and Joonwon Le e. Replicated abstract data types: Building blocks for collab orative applications. J. Parallel Distributed Comput. , 71(3):354–368, 2011. doi: 10.1016/J.JPDC.2010.12.006

    work page doi:10.1016/j.jpdc.2010.12.006 2011

  10. [10]

    Geo-crdt: Geometry-aware collabora tive spatial editing with robust topology preservation.ISPRS International Journal of Geo-Information, 15(7), 2026

    Pengcheng Zhang, Zhongbo Shao, Lin Xu, Jingju Gao, Tian Yu, Jifa Chen, and Ling Hu. Geo-crdt: Geometry-aware collabora tive spatial editing with robust topology preservation.ISPRS International Journal of Geo-Information, 15(7), 2026. doi: 10.3390/ijgi15070302

    work page doi:10.3390/ijgi15070302 2026

  11. [11]

    Tombstone transformation functions for ensuring consiste ncy in col- laborative editing systems

    Gérald Oster, Pascal Molli, Pascal Urso, and Abdessama d Imine. Tombstone transformation functions for ensuring consiste ncy in col- laborative editing systems. In CollaborateCom, pages 1–10, 2006. doi: 10.1109/COLCOM.2006.361867

    arXiv 2006

  12. [12]

    LSEQ: an adaptive structure for sequences in di stributed collaborative editing

    Brice Nédelec, Pascal Molli, Achour Mostéfaoui, and Em manuel Desmontils. LSEQ: an adaptive structure for sequences in di stributed collaborative editing. In ACM Symposium on Document Engineering 2013, DocEng ’13, Florence, Italy, pages 37–46, 2013. doi: 10.1145/ 2494266.2494278

    arXiv 2013

  13. [13]

    The stern-brocot tree

    Bruce Reznick. The stern-brocot tree. American Mathematical Monthly, 97(5):395–402, 1990

    1990

  14. [14]

    Exact arithmetic on the stern–brocot tree

    Milad Niqui. Exact arithmetic on the stern–brocot tree . Journal of Discrete Algorithms , 5(2):356–379, 2007. doi: https://doi.org/10. 1016/j.jda.2005.03.007. 3https://docs.yjs.dev/api/internals 4https://automerge.org/

    2007

  15. [15]

    Mobirdf: A cloud-based collaborative editing service for m obile rdf data sharing

    Nadir Guetmi, Abdessamad Imine, and Moulay Driss Mecha oui. Mobirdf: A cloud-based collaborative editing service for m obile rdf data sharing. Journal of Web Semantics , 86:100864, 2025. ISSN 1570-8268. doi: https://doi.org/10.1016/j.websem.2025 .100864

    doi:10.1016/j.websem.2025 2025

  16. [16]

    Achieving Convergence, Causality-preservation and Intention- preservation in real-time Cooperative Editing Systems

    Chengzheng Sun, Xiaohua Jia, Yanchun Zhang, Yun Yang, a nd David Chen. Achieving Convergence, Causality-preservation and Intention- preservation in real-time Cooperative Editing Systems. ACM Trans. Comput.-Hum. Interact., 5(1):63–108, 1998. doi: 10.1145/274444. 274447

    work page doi:10.1145/274444 1998

  17. [17]

    N. M. Pregui, J. M. Marqu, M. Shapiro, and M. Letia. A comm utative replicated data type for cooperative editing. In ICDCS, pages 395– 403, 2009

    2009

  18. [18]

    Introduction to algorithms

    Thomas H Cormen, Charles E Leiserson, Ronald L Rivest, a nd Clifford Stein. Introduction to algorithms . MIT Press, 4th edition, 2022

    2022

  19. [19]

    Kshemkalyani and Mukesh Singhal

    Ajay D. Kshemkalyani and Mukesh Singhal. Necessary and sufficient conditions on information for causal message ordering and t heir optimal implementation. Distrib. Comput., 11(2):91–111, 1998. doi: 10.1007/s004460050044

    work page doi:10.1007/s004460050044 1998

  20. [20]

    P. Th. Eugster, R. Guerraoui, S. B. Handurukande, P. Kou znetsov, and A.-M. Kermarrec. Lightweight probabilistic broadcast. ACM Trans. Comput. Syst. , 21(4):341–374, November 2003. ISSN 0734-2071. doi: 10.1145/945506.945507

    arXiv 2003

  21. [21]

    A s calable sequence encoding for collaborative editing

    Brice Nédelec, Pascal Molli, and Achour Mostéfaoui. A s calable sequence encoding for collaborative editing. Concurr. Comput. Pract. Exp., 33(8), 2021. doi: 10.1002/CPE.4108

    work page doi:10.1002/cpe.4108 2021

  22. [22]

    Wooki: A P2P wiki-based collaborative writing tool

    Stéphane Weiss, Pascal Urso, and Pascal Molli. Wooki: A P2P wiki-based collaborative writing tool. In Web Information Systems Engineering - WISE 2007, Nancy, France , volume 4831 of Lecture Notes in Computer Science , pages 503–512. Springer, 2007. doi: 10.1007/978-3-540-76993-4\_42

    work page doi:10.1007/978-3-540-76993-4 2007

  23. [23]

    Evaluating crdts for real-time doc ument editing

    Mehdi Ahmed-Nacer, Claudia-Lavinia Ignat, Gérald Ost er, Hyun- Gul Roh, and Pascal Urso. Evaluating crdts for real-time doc ument editing. In Proceedings of the 2011 ACM Symposium on Document Engineering, Mountain View, CA, USA, September 19-22, 2011 , pages 103–112. ACM, 2011. doi: 10.1145/2034691.2034717

    arXiv 2011

  24. [24]

    Strong eventual consi stency of the collaborative editing framework woot

    Emin Karayel and Edgar Gonzàlez. Strong eventual consi stency of the collaborative editing framework woot. Distributed Computing , 35:145–164, 2022. doi: 10.1007/s00446-021-00414-6

    work page doi:10.1007/s00446-021-00414-6 2022

  25. [25]

    Preguiça, Joan Manuel Marquès, Marc Shapiro, an d Mihai Letia

    Nuno M. Preguiça, Joan Manuel Marquès, Marc Shapiro, an d Mihai Letia. A commutative replicated data type for cooperative e diting. In 29th IEEE International Conference on Distributed Computi ng Systems (ICDCS 2009), Montreal, Québec, Canada , pages 395–403,

    2009

  26. [26]

    Concurrency effects over variable-size identi fiers in distributed collaborative editing

    Brice Nédelec, Pascal Molli, Achour Mostéfaoui, and Em manuel Desmontils. Concurrency effects over variable-size identi fiers in distributed collaborative editing. In Proceedings of the International workshop on Document Changes: Modeling, Detection, Storag e and Visualization, Florence, Italy, 2013

    2013

  27. [27]

    Collaborative tex t editing with eg-walker: Better, faster, smaller

    Joseph Gentle and Martin Kleppmann. Collaborative tex t editing with eg-walker: Better, faster, smaller. In Proceedings of the Twentieth European Conference on Computer Systems, EuroSys 2025, Rot ter- dam, The Netherlands, pages 311–328, 2025. doi: 10.1145/3689031. 3696076

    doi:10.1145/3689031 2025

  28. [28]

    Multi-core accelerat ed crdt for large-scale and dynamic collaboration

    Weiwei Cai, Fazhi He, and Xiao Lv. Multi-core accelerat ed crdt for large-scale and dynamic collaboration. The Journal of Supercomput- ing, 78:10799–10828, 2022. doi: 10.1007/s11227-022-04308-7

    work page doi:10.1007/s11227-022-04308-7 2022

  29. [29]

    The art of the fug ue: Minimizing interleaving in collaborative text editing

    Matthew Weidner and Martin Kleppmann. The art of the fug ue: Minimizing interleaving in collaborative text editing. IEEE Trans. Parallel Distributed Syst. , 36(11):2425–2437, 2025. doi: 10.1109/ TPDS.2025.3611880. First Author et al.: Preprint submitted to Elsevier Page 18 of 18

    arXiv 2025

  30. [2009]

    doi: 10.1109/ICDCS.2009.20

    work page doi:10.1109/icdcs.2009.20 2009