The Living Town

Shroud of the Avatar sold town ownership as a product, around 440 times over, and it delivered the deed: a governor really does own a town, place its housing lots, and set its rules. What never shipped is the town game. Governance hangs off a lot sign's right-click menu and three slash commands; there is no in-game list of towns anywhere; and a town on day 1,000 plays identically to day 1. This document surveys the current machinery, walks the genre from Ultima Online's emergent villages to modern settlement systems, and proposes twenty-four improvements, most of them client-side composition over server plumbing that already exists.

Executive summary

Every predecessor treats some layer beyond land as the actual town game, and each proved a different piece of it. Ultima Online shipped zero town mechanics and still grew real towns, because player vendors gave strangers a reason to walk in; its governor system, added sixteen years later, sustains per-city communities to this day on almost nothing but titles and a council seat. Star Wars Galaxies shipped the full civic apparatus, elections, specializations, citizen registries, and then watched its citizen-count rank thresholds kill cities in cascade the moment populations dipped. New World's town project board proved a passer-by will happily donate to your town if a visible bar fills. Rift proved a pure-UI directory creates tourism. And Terraria proved the civic threshold that matters can be two people, not thirty. SotA's own players predicted the ghost-town outcome on Steam in 2015 and today hand-maintain a forum directory because the game has none.

Measured against that history, SotA's gaps cluster in three places:

1. The civic layer — a governor's powers are real (roles, spawn point, message of the day, lot policy, eviction) but scattered across a lot-sign context menu, one roles window, a separate spawn window and three chat commands. There is no console, no citizen identity, no crier, no way to give a town a voice or a face.
2. The discovery layer — 440 towns share 23 template map plates, and the only live datum a stranger can see from outside is an overworld population sprite. Finding a live town with open lots is folklore; players literally look for chimney smoke.
3. The progression layer — nothing in a town accumulates. No project ever completes, no contribution is recorded, no title is granted. The one progression that was sold, tier upgrades, is an out-of-game support ticket.

Twenty-four proposals follow. Fourteen are client-only or near-client-only work over ops the server already exposes. Seven need real server work, and each of those is flagged with its exact data cost. Every one of them is designed small-population-first: it must work at one governor and four residents, no citizen-count thresholds, no upkeep or decay retrofits, cosmetic and civic rewards only. Two in-flight programs are deliberately bridged to rather than duplicated: the guild and town web companion (the town's public web page, calendar, boards, web lot map) and dynamic town management (owner, size and template changes without a client build).

The town you govern today

The bones are real, and worth crediting before criticizing. A town is a scenario layered on one of 23 shared template scenes; the governor drops lot markers as placeable items, and each becomes a working claimable lot with a sign, a deed flow, taxes, and permissions. Governance is live server data: an access ladder of Owner, Steward, Resident and Banned, a selectable arrival spawn, an admin lock, per-lot claim restriction, eviction, and a message of the day. Recent work has been steady: housing lots now draw on the town map in four claimability states, terrain flattening for lots and houses shipped along with three governor levelling tools, house and lot sales with contents are built, banned players are ejected from lots, and town criers greet visitors by town name and answer questions about free lots. None of that is re-proposed here.

What follows is what running a town actually looks like. Every governor power is real; no two of them live in the same place.

• 1Per-lot only. Claim restriction is toggled one sign at a time. A new governor opening a 60-lot town performs 60 sign interactions. → S8.
• 2The "town panel" manages one noun. The Town Access window is the closest thing to a console, and it is a roles list. → S7.
• 3The message of the day has no UI. Three slash commands, 200 characters, and most governors have never found them. On arrival, nothing greets a visitor. → S1, S10.
• 4No town has a face. No banner, no description, no tag, no title for the people in it. → S13, S21.
• 5Zero in-game discovery. The community hand-maintains a forum town directory because the game has no town list at all, and no way to answer "which towns have open lots". → S9, S18.
• 6Nothing to see coming. Lot tax expiry is already delivered to every client per lot; no governor view aggregates it, so decay is discovered as a ghost row of empty houses. → S5.
• 7Nothing accumulates. No project, no contribution record, no shared goal. Towns cannot get anywhere. → S20.
• 8Owner, size and template changes are support tickets. The dynamic town management epic (in planning) owns this fix; this page bridges to it once and proposes nothing in its scope. → S24.

The machinery every proposal plugs into

A town lot is not authored into the scene. It is a decoration: the governor places a lot-marker item on the town's single master plot, the server scans the master plot's contents at town setup and registers a claimable lot (plus its basement and dungeon), and the client builds the runtime lot, sign, and house from the same item. Move the marker and the lot is destroyed and recreated under a new identity. This one fact shapes nearly every proposal below: per-lot client features must track live add/remove events rather than caching lot ids, and per-town server features must ride the existing town operations rather than writing at the database directly, because a live town room owns its own state.

flowchart LR
  G["Governor places a lot marker\n(a decoration on the master plot)"] --> M["Master plot contents"]
  M --> S["Town setup scan\nregisters lot + basement + dungeon"]
  M --> C["Client builds the runtime lot\nsign, house, permissions"]
  C --> K["Claim path\nrestriction check, deed match,\nsquare footage budget"]
  K --> D[("Lot deed record")]
  G -. "move the marker" .-> N["New lot identity minted\nall trackers must follow events"]

Lessons: Ultima Online to modern

Ultima Online shipped in 1997 with no town system whatsoever, and grew the most storied player towns in the genre. PaxLair, founded January 1998, still holds meetings. What made UO towns real was never mechanics: it was clustered housing plus vendors. A house with player vendors was a destination; a cluster of destinations was a town. When UO finally rebuilt a city for players (New Magincia, 2011: plots by lottery, a vendor bazaar), the plots filled and the community did not follow, because a housing lottery without civic mechanics produces private houses in a city skin. And when UO finally added governors (2013), the powers that sustained real per-city communities for the next decade were nearly free: an election, a citizen title, one modest city buff, a seat at the council table. SotA should read its own history in that arc: it already ran the "sell the land and community will come" experiment 440 times.

What each game contributes

Star Wars Galaxies is the canonical player-city system and both halves of its lesson matter. The good half: mayoral elections, a citizen registry that put the city on the planetary map, militia roles, zoning, and above all city specializations, one chosen identity (the crafting town, the entertainer town) that gave strangers a reason to travel. The fatal half: rank tiers gated on registered citizen counts, with demotion wiping the city's invested progress. On the emulator servers that carry SWG today, cities dipping under the 30-citizen shuttleport threshold trigger cascade abandonment and multi-account citizenship fraud. Identity: steal. Headcount thresholds: never.

Wurm Online is the deepest settlement sandbox: purchased deeds with mayor and citizens, a fully custom permission matrix, and a highway network where a village opts in with a waystone and every connected waystone offers route finding to every other. Upkeep is sold as preservation (a funded deed suffers no decay), which players accept where rent would enrage them. Eco gates every civic capability behind a physical building: no town hall, no laws. The full law engine is infeasible under SotA's trust model, but the shape, civic features that live in a physical place, and civic decisions as scheduled communal events, transfers cheaply. New World's town project board is the single most stealable artifact in the modern genre: any passer-by grabs "deliver 50 iron for the town", a visible bar fills, and a named station tier unlocks. Its downgrade twin (failed invasions and unpaid upkeep degrade everyone's stations) was so hated the developers spent years softening it. Ship the board; amputate the punishment.

FFXIV demonstrates why adjacency matters: its housing wards are 60 physically adjacent plots, and the entire player venue culture (nightclubs, cafes, libraries) exists because strangers walk past doors. Its Island Sanctuary is the control group, rich solo content with zero neighbors, and zero community formed. SotA towns already have adjacency; they lack the address, the placard, the reason to walk the street. Albion Online is the structural warning: instanced private islands siphoned its population out of shared space for years, and the developers' answer was a lever, not a lecture: resource returns are 0% on islands versus 10 to 25% in shared and dangerous zones, so public space is simply better. 440 quiet POTs are structurally Albion islands; the fix is pull. ArcheAge contributes the residents board (a per-region list of local property for sale) and land churn: abandoned property automatically recycles. Minecraft's Towny ecosystem is the folk baseline an entire generation grew up on: join a town, get a town spawn teleport, see your town's name and borders on the shared map, pay a tiny tax. And Rift's dimension browser remains the single most SotA-shaped precedent in the genre: purely a list window with search, instant visiting and a recommend button, and it created a builder-tourism culture overnight.

The small games matter most for a small game. Terraria's pylon network activates fast travel when just two NPCs are housed nearby: the proof that civic thresholds can fit SotA's population. Starbound spawns an NPC tenant from a room's furnishings: decoration literally becomes population. Stardew Valley's community center restores a town through donation bundles at population one. Trove's club fixtures include one that spawns quest-giving NPCs, the cleanest "town as content" loop shipped. EQ2's guild-hall amenities (hireling bankers, brokers, gathering NPCs bought a la carte) kept guilds engaged for years. BitCraft deliberately made its whole empire ladder cosmetic prestige, crowns, banners and titles, so that conflict stays low-stakes: exactly the trust-model-compatible shape SotA needs.

Capability matrix

● full system · ◐ partial or emergent · ○ absent. "SotA today" includes work already shipped on the migration branch, 2026-08. The pattern to notice: SotA's empty cells are UI-and-data problems, cheap under its trust model, while the genre's famous failures (SWG's citizen treadmill, New World's downgrade loop, Eco's bureaucracy load) live in rows this page deliberately keeps at ◐.

The twenty-four proposals

Tier 1 is small client work over data and operations that already exist. Tier 2 is the substantial client work plus the first modest server touches. Tier 3 carries the flagships, and every card that changes stored data says so plainly. Throughout, "the console" refers to the governor's console proposed in S7 and drawn in full in section 4.

Tier 1 · Quick wins

Tier 2 · Core upgrades

Tier 3 · Server-backed flagships

Considered and rejected

• Territory taxes over towns — rejected with receipts on the guilds page; stays rejected. Dual Universe retrofitted territory taxes onto existing builders and paid for it in an exodus.
• Citizen-count rank tiers, or any headcount-gated progression — the SWG trap: thresholds caused cascade city death and multi-account fraud, and demotion destroyed invested progress. Every card above works at one governor and four residents.
• Governor elections — the UO and SWG precedent inverts here: SotA governorship is purchased ownership, and elections across 440 towns with a median population of a few people elect nobody. Absentee-governor pain routes to the dynamic management epic's owner transfer instead.
• Town specialization buffs — power perks by another name, and devotionals already occupy the town blessing niche. Identity ships cosmetically as S13.
• Instant teleport from the directory — bypasses the teleport scroll economy and travel friction. The directory shows the way; it does not warp you.
• Upkeep, decay, or demolition retrofits — New World's downgrade loop was the hated half of its town system, and ArcheAge-style demolition belongs to games that did not sell the land. The existing tax system stays exactly as it is; S5 only makes it visible.
• A law engine or free-form permission matrix — Eco's laws work because a server enforces them as code; that is infeasible here, and the roles ladder stays fixed, consistent with the scoping precedent set for guild permissions.
• Resource spawners — carries a standing rejection on private farming. S16 ships the loot-free half only.
• Underground connections between towns — cross-town dungeon topology fights the per-town data model. S15 keeps dungeons per town.
• Anything web — the portal, calendar, boards, wiki, web lot map, lot listings and the housing tours directory all belong to the web companion. S24 is the single bridge.
• Owner, size and template switching — the dynamic town management epic owns it end to end. S23 shares its migration primitive but ships only build-time wiring for the bespoke scenes.
• Decoration quality of life — limits, undo, layouts, patrol routes, access lists: the decorating program's twenty-six cards own all of it.
• Town star ratings or leaderboards — publicly scoring paid customers' towns invites resentment. The census stays neutral counts; prestige stays opt-in through banners and titles.
• Scheduled wipes for freshness — some modern survival MMOs keep settlements fresh by deleting them on a timer. Unusable for persistent purchased ownership, listed only to say so.

The composed surfaces

Twenty-four proposals must not become twenty-four widgets. Everything client-facing above lands on exactly three surfaces: one governor console, one discovery surface, and the in-scene objects a visitor walks past. Everything else is data and server work behind them.

Surface A · The governor's console

Absorbs the Town Access window, the spawn window, the slash commands, and per-lot policy, and hosts the new tabs as their cards land: S1 message, S2 census, S5 lot health, S8 batch tools, S19 applications, S20 projects, S21 titles.

Surface B · Discovery: the town directory

One list window opened from the map screen, and columns that grow with the catalog: population sorting on day one (S9), tags and blurbs when S13 lands, open lot counts and services when S17 and S18 land.

Surface C · In the scene: the town board and its crier

One placeable notice board, refreshed when town properties change, plus the crier of S10 as its spoken twin. The board's notice is the message of the day; the open lot count is computed in scene; the last line is the S24 bridge, shown only once the town's web page exists.

Sequencing

flowchart TD
  OPS["Existing town ops\nroles, spawn, message,\nreserve, evict"] --> S7["S7 console"]
  S1["S1 message + welcome"] --> S7
  S2["S2 census"] --> S7
  S5["S5 lot health"] --> S7
  S7 --> S8["S8 batch tools"] --> S19["S19 applications"]
  S9["S9 directory v1"] --> S18["S18 vacancy board"]
  S13["S13 identity fields"] --> S9
  S13 --> S18
  S17["S17 deed index + roster"] --> S18
  S17 --> S19
  S1 --> S10["S10 crier"]
  S7 --> S20["S20 donation board"]
  S7 --> S21["S21 titles"]
  WEB["Web companion epic"] -. bridge only .-> S24["S24 bridge"]
  DPM["Dynamic town mgmt epic"] -. shared migration .-> S23["S23 bespoke towns"]

The spine: S7 is the hub nearly everything docks into. S17 is the one hard data prerequisite, shared with the web companion so it lands once. S13 feeds the directory its texture. And S1 is the deliberate first move: one small window and one welcome panel, shippable in a day, that make every town in the game feel slightly more alive the moment they exist.

Appendix

Where each proposal's data already lives

Constraints honored

• A town lot is a decoration whose identity changes when moved: every per-lot feature tracks live add and remove events, and nothing persists a lot id.
• Town state is written through live town operations only, never at the database directly, because a running town room owns its state.
• Small-population-first: every card works at one governor and four residents; no headcount thresholds anywhere; nothing downgrades.
• The trust model: no competitive stakes ride on anything client-enforced; rewards are cosmetic and civic throughout.
• Stored-data changes are explicit: S13 (three fields), S17 (one index), S19, S20, S22 (one small collection each), S21 (one field), S23 (per-town migration), each landing with its migrations ledger entry.
• Owned elsewhere, bridged only: the web companion (web presence, calendar, boards, listings, web lot map) and dynamic town management (owner, size, template). Shipped work is credited, not re-proposed: lot map markers, terrain flattening and levellers, house and lot sales, ban ejection, crier town greetings, decoration attribution, and the decorating program's twenty-six cards.

Test surface

Existing fixtures this work extends: the lot map marker model tests, town management conversion tests, lot sign and lot list conversions, lot sale conservation, tax payment models, ban ejection, and the terrain flattening suites. New pure-logic targets, one per model: town message, census (with the de-duplication rules pinned), lot health, console, directory, batch policy, identity validation, application lifecycle, donation conservation, title grants. The counting traps in S2 are exactly the kind of thing a small EditMode test pins forever.

Sources

Ultima Online: UOGuide on player towns, the New Magincia plot guidelines, and the City Loyalty governor system, with the standing forum complaints about absentee governors. Star Wars Galaxies: the SWG wiki's player city pages, Game Update 8's specialization list, and the emulator-community threads documenting citizen-threshold cascade failures. Wurm Online: Wurmpedia on settlements, permissions and highways. Eco: the official government wiki and election dev blogs. New World: the town projects wiki and the settlement retrospectives. FFXIV: the player housing wiki and lottery guides. Albion: the private island debates and return-rate design. ArcheAge: housing and tax guides. Towny: the plugin wiki and shared-map integrations. Rift: the dimension browser guide. Trove, EQ2, Terraria, Starbound, Stardew, Animal Crossing, BitCraft, Pax Dei: their respective wikis and design blogs. SotA: the 2015 Steam thread predicting the ghost-town outcome, the community's forum town directory, the 2020 development update promising expanded placed NPCs, and the player wishlist sweeps of 2025 and 2026.

General comments

No account needed. Every proposal above also has its own Discussion box for talking about that one alone; this thread is for the plan as a whole. Comments are plain text, held to basic decency, and may be trimmed.