Nishi FamilyCompare › Sovereign Office Suite

Nishi Compare · measured, not asserted

Sovereign Office Suite

Nishi vs the field — every Nishi cell is measured against real organ source at emit time; each gap names the watch contract that will close it.

Nishi vs Microsoft 365 Copilot and Google Workspace Gemini and LibreOffice and OnlyOffice

Layer 1 · Executive

Where we are. Measured 2026-08-19. Every format organ is present and gate-proven (docx, odt, xlsx with formulas, charts, pivots, CF and DV, macros, pptx, coedit CRDT, sharing, proofing, spell) and the one measured exceed is tamper-evident additive version history. Presence is not parity: the matrix itself says the incumbents lead on formula breadth and on AI quality, the charts and pivots are HTML renders rather than in-file parts, and cloud sync stays OPEN BY DECISION (operator call, deliberately not built).

Where we need to go. Close the in-file parity gaps the incumbents are judged on (function breadth, chart and pivot parts, presence), then raise AI quality by routing to the strongest local seat under the referee -- and keep the local, no-telemetry moat where the open-source leaders win.

The unit. 1 u = one measured session-leg. Calibration from landed rungs: the mangagen panel compositor went from existing substrate to shipped and live-verified in ONE leg (2026-08-13); the citations rung went from 3 to 55 domains in one leg across seven seats (2026-08-18); a greenfield engine with a bite-proven gate has measured 2 to 4 legs. Estimates recalibrate as rungs land and PR7 actuals write back.
Cost to sheet parity in the file: 4 u. Through M0: the registry first (it is the foundation the other two compose), then chart and pivot parts.
Cost to collaboration polish: 5 u. Through M1.
Cost to AI quality: 7.5 u. Everything below. Routing is cheap; the layout generator is the creative rung.

23 of 33 capabilities measured|1 of them measured exceeds|10 open|coverage 696/1000|adoption 14 full / 9 partial

Layer 2 · Roadmap

Do this next — computed by the ranker, never chosen by a seat

Order from nx_compare_rank (nx_dr_ocm: (deficit + cost-of-delay + option + enables) x sponsor x self-sufficiency x momentum / cost). FINISH rows are rungs whose symbol is present but whose organ is short of full adoption: the cheapest closures on this board, listed before any new work. Stamp: # asof=1787883495 domain=office target_version=0.1 rungs=6 done=0 open=6 finish=0 ranker=nx_dr_ocm

#StageRungPriorityDerivation
#10.1Function registry as data (R0) xl_fn_registry2800v=14 m=1 c=5
#20.1Pivot parts in the file (R2) xl_pivot_part1333v=4 m=1 c=3
#30.1Chart parts in the file (R1) xl_chart_part800v=4 m=1 c=5
#4laterCo-editing presence and cursors (R3) cd_presence1400v=14 m=1 c=10
#5laterRoute AI work to the strongest local seat (R4) of_llm_route1400v=7 m=1 c=5
#6laterDeck theming and layout generation (R5) of_ai_deck_layout857v=6 m=1 c=7

Critical path — contract, done-rule, executor, cost

RungCloses withDefinition of done (pre-declared)ExecutorEst.
Function registry as data (R0)xl_fn_registryEvery spreadsheet function is a registry row with name, arity and a KAT; the evaluator dispatches from the registry; adding a function adds a row and a KAT, never a branch; breadth published as a count against the incumbents' 400 plusOrgan1.5 u
Chart parts in the file (R1)
after R0
xl_chart_partDrawingML chart parts emitted into the xlsx so Excel and Calc render the chart natively; the existing HTML render stays as the sovereign preview; a third-party reader opens the partOrgan1.5 u
Pivot parts in the file (R2)
after R0
xl_pivot_partpivotCacheDefinition and pivotTable parts emitted so the incumbents can refresh the pivot; values equal the resolved-grid pivot already shippedOrgan1 u
Co-editing presence and cursors (R3)cd_presencePer-user presence and cursor positions rendered on the RGA-synced document with two concurrent client processes; the convergence gate stays GREENOrgan1 u
Route AI work to the strongest local seat (R4)of_llm_routeDrafting, sheet insight and deck steps route by task to the strongest local model the cascade offers; the referee grades output quality against the 0.5B baseline with a PRE-DECLARED rule before the route is wiredLocal model1 u
Deck theming and layout generation (R5)
after R4
of_ai_deck_layoutSlide layouts and a theme generated for a whole deck from its outline; rendered to pptx through the existing emitter; a layout judge scores the result, not the eyeballLocal model1.5 u

Milestones

MilestoneRungsCumulative
M0 · Sheet parity in the fileR0,R1,R24 u
M1 · Live collaboration polishR35 u
M2 · AI qualityR4,R57.5 u
Layer 3 · Engineering
How this is scored. Every Nishi mark is measured: the generator reads the real organ source on disk and requires the implementing symbol to exist (no self-grading). A watching tag names the organ and symbol contracted to close a gap — the mark flips itself on the next compare beat when that workstream ships, and the comparewatch- plane row flips with it. The flip is necessary, not sufficient: it proves the symbol exists, never that the capability is good. The bar is the rung's pre-declared done-rule, proven by its gate — a symbol shipped without the behaviour behind it is a defect, and the flip is exactly what makes that defect visible instead of quiet. Competitor marks record documented capability presence — presence, not depth or scale. Adoption is measured too: every measured row carries where its organ stands on the estate's ladder (source → built → promoted → registered → invoked; libraries by importer reach minus validation importers; gates by the execution surfaces that run them). A row is fully adopted only at the top of its ladder; anything short is tagged partial with the exact remedy, so a build nobody promoted can no longer read as shipped. Census stamps: importers asof 1787849099, gate census asof 1787855507 (unix seconds; -1 = census absent).

Capability matrix — measured against source

leads / measured exceed present partial absent · click any capability for its evidence

CapabilityNishiMicrosoft365WorkspaceLibreOfficeOnlyOffice
OOXML .docx read and writeMeasured: docx_write exists in runtime/_hdl_build/nx_docx.nx, verified at emit. Our own OPC/ZIP writer with real CRC-32, no third-party code [ecma-376]; proven by a third-party zip reader opening what we write [onlyoffice] Adoption: RUN-BY:fork:nx_doc_view_gate — fully adopted (top of its ladder).
ODF .odt read and writeMeasured: text:p exists in runtime/_hdl_build/nx_odt.nx, verified at emit. Open format both ways so there is no lock-in [odf-1-3]; LibreOffice is the ODF reference [libreoffice] Adoption: REGISTERED-DARK — PARTIAL: callable, authorised, no MCP invocation on record (a direct fork logs the runner, so this is not proof it never ran); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by unpromoted-source -- a mention, not a run.
adoption REGISTERED-DARK
Rich run formatting bold italic headingsMeasured: w:rPr exists in runtime/_hdl_build/nx_docx.nx, verified at emit. Run-properties emitted per paragraph kind Adoption: RUN-BY:fork:nx_doc_view_gate — fully adopted (top of its ladder).
TablesMeasured: w:tblPr exists in runtime/_hdl_build/nx_docx.nx, verified at emit. Bordered tables from consecutive T rows Adoption: RUN-BY:fork:nx_doc_view_gate — fully adopted (top of its ladder).
Inline images with real dimensionsMeasured: dx_emit_drawing exists in runtime/_hdl_build/nx_docx.nx, verified at emit. PNG embedded as an OPC media part BYTE-IDENTICAL to source; display size read from the file's own IHDR at 9525 EMU per px, capped 6in with aspect preserved. A missing or non-PNG file is skipped rather than emitting a dangling r:embed, which would make Word call the whole document corrupt Adoption: RUN-BY:fork:nx_doc_view_gate — fully adopted (top of its ladder).
Page layout headers and footersOpen — no implementing organ is measured for this axis yet. Real sectPr with pgSz and pgMar plus header1 and footer1 parts
Tracked revisionsMeasured: dx_emit_track exists in runtime/_hdl_build/nx_docx.nx, verified at emit. Real w:ins and w:del accept-or-reject revisions; deletions use w:delText because Word silently drops the text otherwise. Date is a FIXED literal so builds stay byte-deterministic Adoption: RUN-BY:fork:nx_doc_view_gate — fully adopted (top of its ladder).
Comments and annotationsMeasured: dx_cm_add exists in runtime/_hdl_build/nx_docx.nx, verified at emit. commentRangeStart and End plus commentReference anchored to the preceding paragraph by lookahead Adoption: RUN-BY:fork:nx_doc_view_gate — fully adopted (top of its ladder).
FootnotesMeasured: dx_fn_add exists in runtime/_hdl_build/nx_docx.nx, verified at emit. Separator and continuationSeparator pinned at ids 0 and 1 so authored notes start at id 2; Word calls the file corrupt if that is wrong Adoption: RUN-BY:fork:nx_doc_view_gate — fully adopted (top of its ladder).
Spreadsheet xlsx with cell typingOpen — no implementing organ is measured for this axis yet. Numeric v cells vs inlineStr text
Formulas evaluated and emittedMeasured: xl_build_sheet_g exists in runtime/_hdl_build/nx_xlsx.nx, verified at emit. SpreadsheetML f cells carrying the computed value; incumbents ship 400 plus functions and are the reference here Adoption: PROMOTED-UNREGISTERED — PARTIAL: a real binary nobody can call over MCP: /api/tools/register it; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by unpromoted-source -- a mention, not a run.
adoption PROMOTED-UNREGISTERED
ChartsMeasured: xl_chart_html exists in runtime/_hdl_build/nx_xlsx.nx, verified at emit. Native render from resolved sheet data including formula values; in-file DrawingML chart parts are a named follow-up Adoption: PROMOTED-UNREGISTERED — PARTIAL: a real binary nobody can call over MCP: /api/tools/register it; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by unpromoted-source -- a mention, not a run.
adoption PROMOTED-UNREGISTERED
Pivot tablesMeasured: xl_pivot exists in runtime/_hdl_build/nx_xlsx.nx, verified at emit. Group-by SUM COUNT AVG with TOTAL over the resolved grid; in-file pivotTable parts and drag-drop UI are named follow-ups Adoption: PROMOTED-UNREGISTERED — PARTIAL: a real binary nobody can call over MCP: /api/tools/register it; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by unpromoted-source -- a mention, not a run.
adoption PROMOTED-UNREGISTERED
Conditional formatting and data validationMeasured: xl_dir_is exists in runtime/_hdl_build/nx_xlsx.nx, verified at emit. Data-driven CF and DV directives carried in the sheet source plus a real styles.xml dxfs, because a cfRule pointing at a missing dxf is a hard corrupt-file error Adoption: PROMOTED-UNREGISTERED — PARTIAL: a real binary nobody can call over MCP: /api/tools/register it; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by unpromoted-source -- a mention, not a run.
adoption PROMOTED-UNREGISTERED
Macros and scriptingMeasured: xl_macro_apply exists in runtime/_hdl_build/nx_xlsx.nx, verified at emit. SET FILL CLEAR over the grid, run BEFORE formula evaluation so a macro-written cell feeds SUM. DELIBERATELY NOT Turing-complete and with NO I/O and NO exec: a spreadsheet macro that can open sockets is the worst hole in the incumbents and copying that power would copy the defect Adoption: PROMOTED-UNREGISTERED — PARTIAL: a real binary nobody can call over MCP: /api/tools/register it; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by unpromoted-source -- a mention, not a run.
adoption PROMOTED-UNREGISTERED
Presentation decks pptxOpen — no implementing organ is measured for this axis yet. Titles and bullets authored to real pptx
Real-time co-editingMeasured: cd_live_sync exists in runtime/_hdl_build/nx_coedit.nx, verified at emit. RGA text CRDT [roh2011-rga] plus a live HTTP sync daemon, convergence proven with two concurrent client processes [shapiro2011-crdt]; the cloud incumbents remain the reference for polish, presence and cursors Adoption: BUILT-UNPROMOTED — PARTIAL: compiled, never promoted to the serving root: /api/promote it.
adoption BUILT-UNPROMOTED
Sharing and permissionsMeasured: of_do_share exists in runtime/nx_office_serve.nx, verified at emit. ReBAC-gated docs with per-user AND per-group grants, owner-only manage, and a reverse index; LibreOffice is a local app with no sharing plane Adoption: LIB-WIRED importers=5 nonval=2 — fully adopted (top of its ladder).
Version history tamper-evident and additiveMeasured exceed: VIZSLA-DOC-VERSION in runtime/_hdl_build/nx_vizsla_doc.nx, verified at emit. THE ONE MEASURED EXCEED: content-addressed, append-only, never-lose history. Mainstream version history is mutable and deletable; ours cannot be rewritten, and every AI and agent turn lands as a restorable version Adoption: RUN-BY:fork:nx_vizsla_docgen_gate — fully adopted (top of its ladder).
In-document AI draftingMeasured: of_llm_complete exists in runtime/nx_office_serve.nx, verified at emit. Runs on OUR OWN no-float model on loopback with zero cloud. HONEST: the seat is 0.5B so output quality is far below Copilot and Gemini; the incumbents are marked Best on quality [m365-copilot] Adoption: LIB-WIRED importers=5 nonval=2 — fully adopted (top of its ladder).
AI data analysis in sheetsMeasured: of_ai_prompt_sheet exists in runtime/nx_office_serve.nx, verified at emit. Reads the resolved grid and appends an insight as a real two-cell row, additive. Fill-with-Gemini class, 0.5B quality [workspace-gemini] Adoption: LIB-WIRED importers=5 nonval=2 — fully adopted (top of its ladder).
AI presentation generationMeasured: of_ai_prompt_deck exists in runtime/nx_office_serve.nx, verified at emit. Drafts the next slide as a real title plus bullet from the deck so far; no theming or layout generation yet Adoption: LIB-WIRED importers=5 nonval=2 — fully adopted (top of its ladder).
Agentic multi-step in-app workMeasured: of_agent_plan_parse exists in runtime/nx_office_serve.nx, verified at emit. The 2026 marquee, vs Copilot Agent Mode GA 2026-04-22 [copilot-agent-ga]. PLAN then EXECUTE each step then REFINE, state in a sidecar so it is resumable. ONE POST is ONE turn BY DESIGN because a model turn is 7 to 10s against a 15s edge timeout. Every turn lands as an additive version so the whole trajectory is diffable and restorable, which the incumbents discard Adoption: LIB-WIRED importers=5 nonval=2 — fully adopted (top of its ladder).
Grammar and style proofingMeasured: pf_check exists in runtime/_hdl_build/nx_proof.nx, verified at emit. Dictionary-free: doubled words, double spaces, lowercase sentence starts, missing terminal punctuation, unbalanced brackets and quotes, trailing whitespace. Skips markup lines so tables and directives never yield false positives Adoption: BUILT-UNPROMOTED — PARTIAL: compiled, never promoted to the serving root: /api/promote it.
adoption BUILT-UNPROMOTED
Spell checkMeasured: sp_near exists in runtime/_hdl_build/nx_spell.nx, verified at emit. Lexicon DERIVED from our own 205MB English corpus, not fetched: 75,472 words admitted only at 3 or more uses across 2 or more distinct documents, plus a 96-entry denylist because a frequency rule cannot tell a common word from a common mistake and the corpus had admitted teh. PRECISION-FIRST: a word is reported only if absent AND a 1-edit neighbour exists, so that neighbour is both the evidence and the suggestion; an inflection guard suppresses stem-prefix hits. A lexicon gap therefore costs a MISSED TYPO, never a false accusation against correct text. Verified against three neg-controls including real prose Adoption: BUILT-UNPROMOTED — PARTIAL: compiled, never promoted to the serving root: /api/promote it.
adoption BUILT-UNPROMOTED
Cloud and offsite syncOpen — no implementing organ is measured for this axis yet. OPEN BY DECISION, not by inability. The SOTA bar itself says local files and no telemetry is where the open-source leaders win, and that is where our one measured exceed lives; wiring OAuth to Drive or OneDrive would trade the moat for a checkbox. Operator call, deliberately not built unilaterally
Sovereign local no telemetryMeasured: docx_write exists in runtime/_hdl_build/nx_docx.nx, verified at emit. PARITY with the open-source SOTA, not an exceed: LibreOffice matches privacy. The cloud incumbents send data up; our format organs open no sockets at all Adoption: RUN-BY:fork:nx_doc_view_gate — fully adopted (top of its ladder).
Formula breadth as a data-driven function registryOpen — watching runtime/_hdl_build/nx_xlsx.nx : xl_fn_registry, re-measured on every compare beat. Ship that symbol and this mark flips itself; the comparewatch- plane row flips with it. Incumbents ship 400 plus functions; ours evaluates a fixed set -- the registry makes each function a row with a KAT, never a hardcoded switch
watching xl_fn_registry
Charts as in-file DrawingML partsOpen — watching runtime/_hdl_build/nx_xlsx.nx : xl_chart_part, re-measured on every compare beat. Ship that symbol and this mark flips itself; the comparewatch- plane row flips with it. Named follow-up of the charts row: native chart parts so Excel and Calc render them, not only our HTML render
watching xl_chart_part
Pivot tables as in-file pivotTable partsOpen — watching runtime/_hdl_build/nx_xlsx.nx : xl_pivot_part, re-measured on every compare beat. Ship that symbol and this mark flips itself; the comparewatch- plane row flips with it. Named follow-up of the pivot row: pivotCacheDefinition and pivotTable parts the incumbents can refresh
watching xl_pivot_part
Co-editing presence and cursorsOpen — watching runtime/_hdl_build/nx_coedit.nx : cd_presence, re-measured on every compare beat. Ship that symbol and this mark flips itself; the comparewatch- plane row flips with it. Convergence is proven; presence, cursors and per-user colour are the polish the cloud incumbents are judged on
watching cd_presence
AI quality above the 0.5B seatOpen — watching runtime/nx_office_serve.nx : of_llm_route, re-measured on every compare beat. Ship that symbol and this mark flips itself; the comparewatch- plane row flips with it. Route drafting to the strongest local seat by task (the nx_forge cascade precedent); quality graded by the referee, never by presence
watching of_llm_route
Deck theming and layout generationOpen — watching runtime/nx_office_serve.nx : of_ai_deck_layout, re-measured on every compare beat. Ship that symbol and this mark flips itself; the comparewatch- plane row flips with it. Named follow-up of the AI deck row: themes and slide layouts generated, not only title plus bullets
watching of_ai_deck_layout
On these two registers. Rows are declared in the domain's plan file and carry the debt id, which is the join key back to the sovereign debt plane — that plane, not this page, is the authority on state. Reconciling them automatically (the regen reading the plane and refreshing these rows) is a named, owed rung; until it lands, treat an id here as a pointer to look up, not a status to trust.
Honest verdict. The coverage above is capability presence measured against source — not depth, scale, or polish, where mature rivals may lead. Exceeds are claimed only where a mechanism backs them. Every open gap is a watch contract: it names the organ and symbol that closes it, and this page flips the cell itself when that workstream ships.

Person · product · place — not yet measured for this domain

Every compare carries this layer. Declare knowledge/compare/office.ppp (rows surface|nishi or c1..c4|label|url|connect naming OUR live surface and each rival's front door), run nx_ppp_probe domain office, and this section fills itself on the next beat: the same ruler on both sides — privacy and CX (third-party hosts, tracker classes, cookies, security headers), design and longevity (design hygiene, computed WCAG contrast, render-blocking resources, unsized media, script weight, theme and motion queries), findability (landmarks, skip link, on-site search, breadcrumb, headings, internal links).

References

Beyond a link list. Every reference below resolves twice — the publisher's copy and, where banked, the estate's own non-rottable library mirror with a content pin — and carries its evidence class plus the exact claim on this page it grounds. Keyed marks like [key] in the matrix notes jump here. A dash means honestly absent, never assumed.
  1. [m365-copilot] Microsoft. Microsoft 365 Copilot overview (Microsoft Learn): Copilot in Word, Excel, PowerPoint, Outlook and Teams grounded in Microsoft Graph, accessed 2026-08-18. publisher · read in our library knowledge/fetched/cmp_office_m365-copilot.html · pin h4072eb8fd246eda342b89eac04cb78afae9ec4ff0779f856b5d563b1b444099f · accessed 2026-08-18 · vendor-docGrounds: The Microsoft365 column: Best on 'In-document AI drafting', 'AI data analysis in sheets', 'AI presentation generation' and 'Formulas evaluated and emitted', and the cloud-side 0 on 'Sovereign local no telemetry' -- the reference the matrix header calls the quality bar our 0.5B sovereign seat sits far below.
  2. [copilot-agent-ga] Microsoft 365 Blog (Sumit Chauhan). Copilot's agentic capabilities in Word, Excel, and PowerPoint are generally available, 2026-04-22: multi-step app-native actions as the default experience for Microsoft 365 Copilot, Premium, Personal and Family. publisher · read in our library knowledge/fetched/cmp_office_copilot-agent-mode-ga.html · pin hde806e0d34cc2421dc60ecf1e902a3bb678a1136c91c3c272c2be1e865a62a18 · accessed 2026-08-18 · vendor-docGrounds: The 'Agentic multi-step in-app work' row: the 2026 marquee the note names (Copilot Agent Mode GA 2026-04-22) is this announcement; ours differs by landing every turn as an additive restorable version and by one-POST-one-turn against the 15s edge timeout.
  3. [workspace-gemini] Google. Google Workspace with Gemini -- AI in Docs, Sheets, Slides, Gmail and Meet (solutions page), accessed 2026-08-18. publisher · read in our library knowledge/fetched/cmp_office_workspace-gemini.html · pin hb090db57adc624936a741757c33038653c67072e377c5f8654cf1bdd930ae1cd · accessed 2026-08-18 · vendor-docGrounds: The Workspace column: Best on 'Real-time co-editing', 'In-document AI drafting' and 'AI data analysis in sheets' (the Fill-with-Gemini class the note names), Part on 'Agentic multi-step in-app work'.
  4. [libreoffice] The Document Foundation. LibreOffice -- Discover LibreOffice (product page): free open-source suite with Writer, Calc, Impress, Draw, Base and Math; the ODF reference implementation. Accessed 2026-08-18; libreoffice.org's TLS chain would not link for the sovereign client, so no mirror is banked. publisher · accessed 2026-08-18 · vendor-docGrounds: The LibreOffice column: Best on 'ODF .odt read and write' (the note calls it the ODF reference) and 'Sovereign local no telemetry' (parity, not exceed), 0 on 'Sharing and permissions' because it is a local app with no sharing plane.
  5. [onlyoffice] ONLYOFFICE (Ascensio System SIA). ONLYOFFICE Docs and Workspace -- OOXML-native online editors, self-hostable (product page), accessed 2026-08-18. publisher · read in our library knowledge/fetched/cmp_office_onlyoffice.html · pin hf3ebbc84dadb71ed5422d4d086fd217f138636c92a788bf45dd6d2f257d225b3 · accessed 2026-08-18 · vendor-docGrounds: The OnlyOffice column: Yes across the OOXML document, spreadsheet, co-editing and sharing rows, 0 on the AI rows -- the self-hostable OOXML-native comparator.
  6. [ecma-376] Ecma International. ECMA-376: Office Open XML File Formats (WordprocessingML, SpreadsheetML, PresentationML, DrawingML, Open Packaging Conventions), also ISO/IEC 29500. publisher · read in our library knowledge/fetched/cmp_office_ecma376.html · pin h6ebf45ac6b3db5afc2316f13acd0dddff5d78d78f2a8720cc7a07b2975d38c83 · accessed 2026-08-18 · published-standardGrounds: The 'OOXML .docx read and write', 'Tracked revisions', 'Comments and annotations', 'Footnotes', 'Spreadsheet xlsx with cell typing' and 'Presentation decks pptx' rows: the OPC package, w:ins/w:del, commentRangeStart, footnote separator ids and sectPr parts our own writer emits are this standard's, and the corrupt-file traps the notes name are its constraints.
  7. [odf-1-3] OASIS. Open Document Format for Office Applications (OpenDocument) Version 1.3, Part 1: Introduction. OASIS Standard, 2021 (also ISO/IEC 26300). publisher · read in our library knowledge/fetched/cmp_office_odf13.html · pin h4891a74a8772d7678c4fd720ebb5a54a810a4d98811b9b68ecb2e3bcf468e91f · accessed 2026-08-18 · published-standardGrounds: The 'ODF .odt read and write' row: the open format the row names as the no-lock-in path is this specification; text:p is its paragraph element.
  8. [roh2011-rga] Roh, Jeon, Kim, Lee. Replicated abstract data types: Building blocks for collaborative applications. Journal of Parallel and Distributed Computing 71(3), 354-368, 2011 -- the Replicated Growable Array (RGA). Bibliographic record confirmed via dblp; the Elsevier DOI page refuses the sovereign fetcher, so no mirror is banked. publisher · accessed 2026-08-18 · published-paperGrounds: The 'Real-time co-editing' row: the RGA text CRDT nx_coedit implements (cd_live_sync, convergence proven with two concurrent client processes) is this paper's data type.
  9. [shapiro2011-crdt] Shapiro, Preguica, Baquero, Zawirski. Conflict-Free Replicated Data Types. Stabilization, Safety, and Security of Distributed Systems (SSS 2011), LNCS 6976, Springer. publisher · read in our library knowledge/fetched/cmp_office_shapiro2011-crdt.html · pin hc415220561443c7ee82abb37b03819eddccfb2c2fca935e8ce38952c86bf540f · accessed 2026-08-18 · published-paperGrounds: The 'Real-time co-editing' row: the strong-eventual-consistency guarantee (convergence without coordination) that the co-editing gate proves is the property this paper defines for CRDTs.

generated by nx_swcompare_matrix (sovereign NishiLang organ) from knowledge/compare/office.matrix · every Nishi cell verified against organ source at emit time · watch cells re-measured on every compare beat · zero JS, zero trackers