#Carthage — map specification

Status: draft 2. Written at fbcfe65 for the three agents building the map (terrain and integration, walls, city fabric). Everything here is a number you can build to. Where the sources disagree or the archaeology is thin the entry says so and still gives the number to use, because a builder cannot act on "uncertain".

Draft 2 adds: the cubit module the Byrsa quarter is actually laid out on (§7.1, read off the excavation plan); the excavated 125 m admiralty island, against the 130 m draft 1 derived (§6.2); Appian's sixty steps verified (§5.2); the forum (§7.6) and the Megara (§7.7); and reference/carthage/, 15 licence-verified files catalogued in ASSETS.md.

This is the counterpart to src/city/rome.ts + src/city/layout.ts. Read those two first if you have not: this document deliberately reuses their idioms — a survey in real metres, an anisotropic affine projection onto the battlefield, and cited dimensions — so that Carthage is falsifiable in the same way Rome is.

#0. How to read this

Every dimension carries a tag:

Units. All real-world dimensions are metres. All elevations are metres above ancient mean sea level; if the terrain profile needs a non-negative sea bed, add one constant offset to everything in §3 including waterLevel, and say so in the profile.

Ancient units used in conversion. Attic/Roman foot 0.296 m; Attic cubit 0.444 m; Punic cubit c. 0.515 m; stade 185 m. Where a cubit choice changes the answer materially, both are given.

#1. The moment: spring 146 BC

Hold everything to 146 BC, the fourth year of the Third Punic War and the final assault under P. Cornelius Scipio Aemilianus.

Why this year and not another:

• It is the only moment when both of the map's two attack routes exist at once. In 149 BC the triple wall is intact and Manilius breaks himself on it, but there is no Roman lodgement anywhere near the harbour. In 147 BC Scipio has Megara and the blockade but not the quay. In spring 146 the triple wall is still unbreached on the isthmus — Scipio never took it — while the Romans hold the harbour quay and are one push from the cothon. An attacker on this map can therefore choose the front the Romans failed at or the flank they succeeded at, and both are period-correct on the same day. §8 is built on that.
• The 147 BC siege works are on the ground and they are good props: Scipio's mole across the harbour mouth, the Carthaginians' answering channel cut to the open sea, and the Roman brick quay-fort. [A] All three are attacker assets or defender escapes and all three are gone from a 149 BC map.
• The street fight is the payoff. Appian's six days and nights from the forum up the three streets to the Byrsa, houses six storeys high, Romans crossing roof to roof on planks. [A] That fight only happens in 146.

What 146 BC costs us, stated plainly.

• There are no Carthaginian war elephants in this battle, and there cannot be. The 201 BC peace forbade Carthage to train them; in 149 BC the city surrendered its arms wholesale before refusing to abandon the site. The elephant stalls inside the wall (§4.4) are fourth- or third-century architecture standing empty in 146 — which is not a problem, it is a detail: a wall built for an empire, held by a city that no longer has one.
• The shipped Carthage roster in src/units/roster.ts is dated 218–202 BC by its own header comment and fields war elephants. Nothing needs changing. The engine already lets any army fight on any map, and a 218 BC Punic army defending its own walls is a perfectly ordinary Total War counterfactual. If the elephants are on the field, the stalls are period dressing for them. Do not renumber the roster to suit the map, and do not renumber the map to suit the roster. State the map's year in the blurb and let the player make the anachronism if they want it.

Map metadata to register.

On the sun. Latitude 36.85 with declination +13 puts the noon sun at 66° altitude — 32° higher than Rome's November noon and far flatter. docs/VISUAL-RUBRIC.md and the blind-deck record both say relief is load-bearing, so do not ship a noon default. Open at 16:30 (sun ~30°, bearing WSW — straight down the attacker's approach and straight into the outer face of the triple wall, which is the one surface the whole map is about). A 08:00 preset back-lights the wall and rakes the Byrsa's west scarp; that is the second-best hour and worth having.

#2. The frame, and the projection

#2.1 What the engine fixes and we cannot move

From src/maps/types.ts and src/sim/scenario.ts, verified in the tree at fbcfe65:

So the map moves the world under a fixed order of battle. −Z is the attacker's side; +Z is the city. That is a direction, not a compass bearing, and Carthage uses it differently from Rome.

#2.2 Compass orientation of the map

Map −Z = true west. Map +Z = true east. Map +X = true north. Map −X = true south.

Rome's map has −Z = north. Carthage's is rotated 90° anticlockwise from that, and it has to be: the only land approach to Carthage is from the west, across the isthmus, and that approach must lie along −Z where the attacker deploys. Consequences you get for free:

• the Gulf of Tunis is the +Z map edge — the defender's back is against the sea;
• the Lake of Tunis is the −X (south) edge, and the Taenia sandspit runs along it;
• the Sebkhet Ariana salt flat is the +X (north) edge;
• the wall's two ends both die on water. There is no flank march on this map.

#2.3 Survey origin and the projection

Follow rome.ts exactly: author positions in real metres from a fixed monument and compute the battlefield coordinates. Getting Carthage wrong should require getting the survey wrong.

Origin: the summit of the Byrsa, 36.8528 N, 10.3233 E   [MOD]
  e = metres EAST of the origin
  n = metres NORTH of the origin
  111,132 m per degree of latitude
   89,100 m per degree of longitude at 36.85 N   (111,320 · cos 36.8528°)

Projection (the Carthaginian analogue of worldOf):
  x = KN · n            KN = 0.45      (across the map — true north)
  z = Z0 + KE · e       KE = 0.22 ,  Z0 = 945    (into the map — true east)

Why those two numbers.

• KE = 0.22 is not a taste decision, it is the largest value that fits the whole city between the wall and the sea inside the map. The triple wall must sit forward enough that the attacker gets a real approach (the crest lands at z ≈ 494–615, so ~640 m of open ground from the deployment box — Rome's is ~620, so the two maps read at the same tempo), and the coastline must land inside z = 1374 with visible water behind it. The city is 2.95 km deep west-to-east and the budget is 650 world metres. It also happens to be exactly Rome's KZ, which means the two maps compress depth identically and a player's sense of distance transfers.
• KN = 0.45 is set by the wall: the isthmus front is 4.43 km of real wall (§4.1) and it has to fit across a 2800 m map with both ends on water and some lagoon showing beyond each. 0.45 puts it at 1,984 world metres. It is also within 1.5 % of Rome's KX (0.443).
• Anisotropy is therefore 2.05×, against Rome's 2.00×. The two cities are distorted the same way by the same amount. Nobody has to learn a second mental correction.

#2.4 The rule that stops this going wrong

Positions compress. Cross-sections do not.

rome.ts already works this way and nobody wrote it down: a monument's position goes through the affine map, its height does not, and its plan footprint takes a separate PLAN_SCALE. For Carthage, state it up front:

• Compressed (through KN/KE): every position; the length of the wall along its own line; the extent of a district; the plan of the harbours; the footprint of the Byrsa hill.
• Not compressed (true metres in world space): wall height, wall thickness, ditch width and depth, tower height and footprint and spacing, street widths, storey heights, insula dimensions, ship-shed dimensions, the height of the Byrsa above the lower town.

Concretely: the wall is 4,434 real metres long and 1,984 world metres long, but it is 9.1 metres thick in both. Towers stand 59.2 world metres apart because that is the real interval, which means the modelled stretch carries 33 towers where the real wall carried 75. That is exactly what Rome does with WALL.towerSpacing = 35.5.

And there is a third category, which is where this rule bites. Anything whose slope matters cannot take a compressed run against an uncompressed height, because the gradient comes out wrong by the compression factor. There are exactly two such things on this map and both are overridden explicitly:

• the Byrsa — §5.1a; the projection gives it a 30° face against a real 1:7, so its world footprint is set from the gradient instead. Rome already does this: RISE_RUN = 175 is a chosen world number, not a projected one;
• open spaces that have to be fought in — the forum, the harbour quays, the wall's killing ground — which take world dimensions directly (§7.6), because a projected 180 × 120 m agora comes out 81 × 26 and is a corridor.

If you find a third case, add it here rather than quietly bending the projection.

Plan scale for monuments. Rome uses PLAN_SCALE = 0.65 because a 1:1 building in a 10×- compressed plan eats the city. Carthage's monumental load is far lighter (there is no Colosseum, no Circus, no imperial baths — Punic Carthage's largest single structures are the harbours and the Byrsa's temple platform) but the harbours are enormous. Recommend PLAN_SCALE = 0.80 for Punic monuments, and 1.00 (unscaled) for the two harbour basins, which are landscape features, not buildings, and are already compressed as areas by the projection — the same exemption rome.ts gives to soft entries.

#2.5 The survey — key positions

Longitude/latitude are modern site coordinates [MOD]; the ancient shoreline differs (§3.2).

Sanity checks that must hold after you build:

• attacker deployment (z −190) to the ditch lip (z ≈ 452 at mid-wall) = 642 m of approach;
• wall (z 527) to Byrsa summit (z 945) = 418 m of city depth on the main axis;
• Byrsa (z 945) to the shore (z 1176) = 231 m, then ~200 m of open sea to the map edge;
• modelled wall length 1,984 m against Rome's 1,781 m (WALL_X_MAX − WALL_X_MIN, computed at fbcfe65: 1150 − (−631)). Rome's wall costs 216 draw calls. Carthage's is 12 % longer, three walls deep and casemated. Budget for it early; this is the map's largest single risk.

#3. The ground

An attacker's options are set here before any wall exists, and Carthage's ground is the opposite of Rome's in the one way that matters.

#3.1 The headline

At Rome the wall stands on a hill. At Carthage it stands on nothing. Aurelian's curtain sits on a 22–34 m rise (riseAmplitude in topography.ts) so an attacker climbs 175 m of slope under fire before he reaches 6.5 m of masonry. The Carthaginian isthmus is a low, flat neck; the triple wall has no terrain advantage whatever and carries all of its defence in stone. Which is precisely why it is 16 m tall behind a 20 m ditch.

Do not "improve" this by putting the wall on a ridge. The flatness is the design.

#3.2 The peninsula

Carthage occupies a promontory with the Gulf of Tunis east, the Lake of Tunis south-west, and the Sebkhet Ariana (a salt flat) north-west. Land reaches it only from the west.

#3.3 Elevations, metres above ancient mean sea level

So the Byrsa stands ~45 m above the lower town. That is the single most important elevation on the map and §5 spends it.

#3.4 Ground conditions — the constraint that shapes a siege train

Within ~300 real metres of the Lake of Tunis and the Sebkhet Ariana the ground is soft — sabkha margin, salt marsh and silt. [MOD] [GAME as a mechanic]

Recommendation: mark those margins in the control texture as a soft channel and have it (a) halve movement speed for wheeled engines, (b) forbid engine placement entirely within 150 world metres of either water, and (c) render as cracked salt crust with standing brine in the lows. Consequence: rams, towers and heavy artillery must go up the middle of the isthmus, where the wall is strongest and where the defender knows they must come. Infantry can still work the margins. That is a real decision handed to the attacker on turn one and Rome has nothing like it.

The Sebkhet Ariana is a seasonal pan: dry crust in summer, impassable mud in winter. In spring 146 call it firm enough to walk, too soft for engines — which is the interesting setting and is defensible for late April.

#3.5 The Taenia — the flank route

Appian's taenia ("ribbon") or ligula: the tongue of land between the Lake of Tunis and the sea, running from the isthmus east toward La Goulette. [A] Censorinus assaulted along it in 149 BC. Appian names the angle where the triple wall meets it and runs on toward the harbour as "the only weak and low spot in the fortifications, having been neglected from the beginning." [A] That sentence is this map's second act.

Appian's 300 feet and the modern 1.1 km cannot both describe the same cross-section; the 300-foot figure is almost certainly the Catadas channel neck at what is now La Goulette, which is off-map to the south. Use the modelled figures. They are chosen so the corridor takes a legion in column and a battery, but not a line of battle — a 320-man cohort at the sim's 0.72 m pitch is ~35 m wide in line, so 200 m of corridor holds five cohorts abreast at most, with a lake on one flank and a wall on the other and nowhere to go.

On the map: the Taenia enters at the −X edge around z 300–500 and runs in +z along x ≈ −1100 to −1400, meeting the harbour district around z 850–950.

#3.6 The coastline

Author the shore as a polyline in the survey frame and let the projection place it. Points measured off the modern coast [MOD]; note the ancient shoreline was somewhat further west in the harbour quarter (the ports have silted and the land has prograded), which is why the harbour channel in §6.4 is short.

That last row is 43 m from the map edge, so the coast leaves the map to the north-east — correct, because the peninsula continues to Sidi Bou Saïd. The open sea occupies the +Z corner and the whole north-east of the map beyond the shore.

#4. The triple wall

This is why the map is worth building. Build it first.

#4.1 The line

Build the line with a slight bow (convex toward the attacker, sagitta ~25 world m) rather than dead straight. It gives every bay a flanking angle onto its neighbours and it stops the wall reading as an extruded rectangle at the strategic camera, which is the failure mode the blind deck already names for our architecture.

#4.2 The arrangement, front to back

Appian says only "a triple wall … the height of each 30 cubits (45 ft) without counting parapets and towers … the depth 30 ft" [A], which read flat would give three identical 13.7 m curtains and is not credible. The modern reading — Lancel, Goldsworthy, Miles — is a stepped system in which each line is lower than the one behind it, so all three can be fought from at once and the main wall dominates everything in front of it. Build that.

Distances are the clear gap between the works; cross-sections are true metres, uncompressed.

Compare Rome, exactly, so nobody quotes a flattering version of this:

The 12.4× is the real headline and the only one worth putting in the blurb. The height ratios are under 2× and the distance-to-first-building ratio is under 2×; it is the number of things between you and the city that changes, not the size of any one of them.

#4.3 The main wall — dimensions

The 7.1 m clear walk is the single most important consequence of Appian's thickness, and it is the wall workstream's headline. Rome's widened curtain gives a 2.21–4.06 m band — 4–6 ranks at the sim's 0.72 m pitch. Carthage gives 9 ranks, or four ranks plus a mounted messenger, or a scorpion with its crew and a file passing behind it. The Carthaginian wall-walk is a street.

#4.4 The casemates — yes, they are enterable

Appian, on the main wall: each wall was divided vertically by two vaults, one above the other; the lower held stables for 300 elephants with their fodder stores alongside; the upper held stables for 4,000 horses with fodder and grain; and there were barracks for 20,000 foot and 4,000 horse. [A]

Make the lower level a fully enterable interior. This is the recommendation and it is not close. A wall a player's men can be inside is a different tactical object from one they stand on top of, and no other city in this game has one.

The arithmetic says the elephants are real.

wall length (real)                        4,434 m
thickness                                     9.1 m
outer face 1.5 m + inner face 1.2 m  →  internal clear span 6.4 m
internal floor area per level        4,434 × 6.4 = 28,378 m²
300 elephants in 28,378 m²                   = 94 m² each
an elephant needs roughly                      10–15 m²

The lower vault holds 300 elephants at six times the space they need, which leaves ample room for the fodder magazines Appian puts "alongside". One stall every 14.8 m of wall [DER]. That is an internally consistent description, and internal consistency is the strongest evidence we have that Polybius was describing something he had walked through.

The arithmetic says the barracks are not.

upper level                              28,378 m²
4,000 horses at ~4 m² each             = 16,000 m²
remaining for 24,000 men               = 12,378 m²  →  0.52 m² per man

Half a square metre per soldier is not a barracks, it is a crowd. Do not model the upper level as a dormitory for 24,000. Either the 24,000 are the wall's garrison establishment quartered partly in rear ranges and in the towers, or the figure is Appian's inflation. Build the upper level as what it plausibly was: a fighting gallery — a continuous vaulted corridor with loopholes through the outer face, guardrooms at the tower bases, ready-magazines, and some stabling and bunk ranges as dressing.

Built dimensions to hand the wall agent.

Passability. The casemate corridor must be in the nav mesh at both levels, and it must be a one-file-plus-passing corridor: 6.4 m of span at 0.72 m pitch is 8 men abreast, which is a formation, not a queue. Cap it lower if it plays too easily — 8 abreast inside a wall is a very strong defensive position — but start at the real number.

#4.5 Towers and gates

Rome's towers stand 13.8 m and there are 50 of them across 1,781 m. Carthage has fewer, far bigger towers: 33 at 22.5 m. Silhouette that difference deliberately — Rome's wall reads as a serrated line, Carthage's as a row of keeps joined by a rampart.

Gates. No ancient source gives a gate count for the land wall and none has been excavated. Recommendation, [GAME] throughout, three in the triple wall:

Each gate must pass all three lines — so a gate is not a door, it is a 90 m tunnel through the whole belt: a causeway over the ditch (make it a timber bridge the defender can fire, [GAME]), a gap in the outwork offset laterally by 8 m from the gap in the middle wall, and then the main gate itself. Offset the three openings so a ram cannot see daylight through them. That single decision is worth more to this map than any texture.

Additionally, archaeologically attested and not in the triple wall: a sea gate in the Magon quarter on the east shore, with a 9 m street running to it. [ARCH] Put it at x ≈ +150, z ≈ 1200.

#4.6 What is Appian and what is not — read this before you cite anything

Appian wrote in the second century AD, ~300 years after the event, from a source generally taken to be Polybius, who was there. That makes the description first-rate by ancient standards and still not archaeology.

Load-bearing and probably sound: the triple arrangement; the two-level casemate; elephant stabling in the lower level; towers at 200 ft in four storeys; the harbour entrance of 70 ft with chains; the island with the admiral's house; the Ionic columns; the weak angle by the harbour; the three streets and the six days.

Load-bearing and probably inflated: 20,000 foot and 4,000 horse barracked in the wall (§4.4 shows why); 220 ships in the sheds where archaeology finds capacity for 160–170 (§6.3); the 33–37 km circuit.

Not located by archaeology at all: the line of the triple wall on the isthmus; the ditch; the gates; the tower plan. Every one of those is a reconstruction here and is tagged as such.

Found by archaeology and not in Appian: the seaward rampart between the Bay of Kram and Bordj Djedid; the Magon-quarter sea gate and its 9 m approach street; the Byrsa's Hannibalic housing grid; the harbour slipways.

#5. The Byrsa

#5.1 The hill

#5.1a The one place the projection must be overridden, and the arithmetic that says so

Heights are not compressed and positions are, so every slope on this map comes out steeper than it really was. On the Byrsa that stops being a stylistic issue and becomes a bug. Run the real hill through §2.3 and you get:

real footprint at the 20 m contour   700 m (E–W)  ×  550 m (N–S)
projected                            0.22·700 = 154 m in z   ,  0.45·550 = 248 m in x
relief, uncompressed                 45 m
resulting gradient into z            45 / 77   = 1 : 1.7   (30°)
resulting gradient across x          45 / 124  = 1 : 2.8   (20°)
the real gradients                             1 : 7.8  and  1 : 6.1

A 30° face is a cliff. The three streets would be unbuildable, the housing terraces would interpenetrate, and the whole point of §5.3 would be lost.

Fix: set the Byrsa's world footprint from the gradient you want, not from the projection. This is not a special case — it is exactly what topography.ts already does at Rome, where RISE_RUN = 175 is a chosen world number with the comment "short enough to read as a hill front", not a projected one. State the same thing here.

Recommended: 340 m in x by 200 m in z at the 20 m contour, which puts the built south-east face at 170 world m of run for 45 m of rise = 1:3.8. That is 1.8× steeper than the real 1:7 — the same order of distortion Rome accepts on its own rise — and it is a slope three stepped streets can climb and terraced housing can sit on.

Do not compress the 45 m. Heights are not compressed (§2.4) and this one is the reason to come.

#5.2 What stood on it

The stratigraphy of the south slope, from the published section (reference/carthage/ plan-byrsa-hill-section.png, CC0), because it tells you what the slope is: a Punic necropolis lowest, metalworkers' workshops above it, then the ground floors of the apartment blocks, and over the lot the Roman forum platform and the levelled summit. In 146 the slope is housing over industry. Put smithies, slag, kilns and workshop yards into the lower quarter rather than making the whole hill residential — it dresses the fight and it is evidenced.

#5.3 The approach, and why it matters

Appian: three streets ran up from the forum to the Byrsa, lined on both sides with houses six storeys high, and it took the Romans six days and nights to get up them. [A]

Build exactly that:

• Three streets, not one, from the forum flat to the citadel gate. They run up the hill's south-east face, i.e. predominantly in +x on this map (from the forum at x ≈ −290 to the summit at x = 0), which is the map's less compressed axis — a piece of luck from the orientation choice in §2.2 and worth checking rather than assuming. Each street is ~170 world m long and gains 45 m.
• Each 6.0 m wide [ARCH] and stepped, because even the real 14 % could not be walked in formation and the built 26 % certainly cannot. Give them treads of 1.2 m and risers of 0.17 m in flights of 8–12 with landings.
• Wheeled traffic and every siege engine is excluded from all three. There is no way to get a ram to the citadel gate. That is a historical fact and a superb constraint.
• Formation coherence must break on a stepped street. A cohort going up the Byrsa should arrive as a mob, and the defender's job is to be waiting at the top of it.

The fourth approach — round the north, up the steep scarp — should be passable to infantry in loose order only and cost heavily in cohesion. Give the player the choice.

#6. The harbours

#6.1 Can a land battle meaningfully reach them?

Yes, and the whole second half of the map is there. Appian is explicit: the Romans took the quay, then unexpectedly got onto the ring of the circular harbour, and from there Scipio seized the forum. [A] The harbours are not scenery; they are the route to the Byrsa.

#6.2 Dimensions

The island diameter, cross-checked. This document first derived 130 m from the shed counts before the excavated figure was found, and the excavated figure is 125 m. The derivation is kept because it is a useful consistency check on everything else: British excavation found 30 slipways on the island and 135–140 round the ring, 160–170 in total, at 5.9 m wide [ARCH]. A 125 m island has a 393 m circumference; 30 × 5.9 = 177 m, so the sheds occupy 45 % of it — consistent with Hurst finding them in discrete blocks rather than a continuous ring. The 325 m ring has a 1,021 m circumference; 138 × 5.9 = 814 m, 80 % of it, leaving ~200 m for the gate, the channel to the rectangular basin and the ramps. Every figure in this table is mutually consistent, which is unusual for Punic Carthage and worth trusting.

Warning about the aerial photographs. In reference/carthage/aerial-salammbo-ports.jpg the island looks like 55–60 % of the lagoon's width, not 125/325 = 38 %. That is modern silting and landscaping, plus oblique foreshortening. Build to 125 m, not to the photograph.

#6.3 The ship-sheds

That portico is the single best-looking thing on this map: a 1 km ring of columns round a circular lagoon, doubled on the island. It also happens to be a colonnade a fight can happen in, which is a kind of space this project does not currently have.

#6.4 Connection to the sea, and the 147–146 works

• The sea entrance (21 m, chained) opens into the rectangular basin from the south-east.
• The circular basin is reached only through a controlled second channel from the rectangular one, behind a double wall with a gate, so merchants could not see into the naval yard. [A] Model that channel at 21 m as well, with a gate and a chain.
• The island is reached by a single timber-decked causeway, 4.0 m wide, on the north side. [GAME] — flagged loudly. There is no good evidence for a causeway; the ancient accounts imply boats. It is here because a 4 m bridge onto a defended island is the best chokepoint on the map and because the alternative is that the island is unreachable in a land battle and therefore pointless. If the owner would rather be strict, delete the causeway and give the attacker boats.
• Scipio's mole [A]: an embankment thrown across the harbour mouth to blockade it. Model it as a rubble causeway 25 m wide at the base, 12 m at the top, 3 m above the water, running from the shore across the entrance. It is an attacker asset: it carries men and light engines to the harbour front dry-shod.
• The Carthaginian cut channel [A]: with the mouth blocked, the Carthaginians dug a new channel from the harbour straight out to the open sea and sent a new fleet through it. Model it on the east side of the circular basin, 30 m wide, freshly cut and unrevetted. It is a defender asset (reinforcement and escape by sea) and simultaneously a 30 m gap in the city's own defences that an attacker can exploit. Nothing else on either map does both.
• The Roman quay-fort [A]: on the captured quay the Romans built a brick structure as high as the city wall from which 4,000 men shot down onto the ramparts at short range. Model it as a timber-and-brick siege platform, 16 m to its fighting deck, 60 × 20 m in plan, standing on the quay of the rectangular harbour. It is the attacker's answer to a 16 m wall and it should be buildable or pre-placed depending on scenario.

#7. Street plan and housing

#7.1 What is actually known

#7.2 Road ranks — Carthage's answer to Rome's 42/24/14/8

This is where the analogy to Rome breaks, and it breaks hard.

Rome's WAY_WIDTH gives artery 42 / secondary 24 / local 14 / vicus 8 m, and layout.ts already concedes in its own comment that a real Roman via is about 4.8 m and 42 is a game compromise so a 35 m cohort can move. Carthage has no equivalent of even the honest end of that. The widest street anywhere in the excavated Punic city is 9 m, and it is called out in the literature as exceptional. The ordinary Punic street is 3–7 m.

So: use narrower ranks than Rome, and accept fewer of them.

The consequence the city agent must design for, not design around: a cohort in line is ~35 m wide. It fits on a processional street and nowhere else. Everywhere else in Carthage the attacker is in column at 4–7 m frontage or he is not moving. That is the map's infantry mechanic and it should be felt, not fixed.

Rome's own audit already found that the district generator's 374 lanes at 8 m are what make the fabric read; Carthage's are at 4–7 m and there should be more of them.

#7.3 Blocks and buildings

Author the fabric on the cubit module, not on metres. A Punic cubit is 0.515 m; the whole Byrsa quarter is laid out on it, and a generator that snaps to 30 × 60 cubits with 12-cubit plot subdivisions will produce a Carthaginian street front by construction. This is the single most useful thing in the archaeology and Rome has no equivalent.

Flat roofs are a mechanic, not a look. Appian's Romans crossed from roof to roof on planks and fought a battle up there while a second battle went on in the street below. [A] fbcfe65 made the wall traversable terrain; roofs are the same problem, and a Carthage that does not let men onto its roofs has thrown away its best scene. Specify: flat roofs of buildings 4 storeys and up on the three Byrsa streets are walkable, with 2.0–4.0 m gaps between blocks that a plank crossing can bridge.

#7.4 Grain — free authenticity

docs/HANDOFF.md records that Rome's remaining difference from the AGEA orthophoto is grain, not coverage: real blocks are smaller and punched with 1–4 courts of 10–25 m, ours are larger with one big court. Carthage's real grain is 15.5 × 31 m with a 3–5 m courtyard per house, which is finer than anything Rome has. Build to the archaeology and the grain problem solves itself. Do not scale Punic blocks up toward Roman ones.

#7.5 The military way — Carthage's pomerium

35 m clear behind the main wall. [GAME]

Rome's POMERIUM = 60 is built from three needs stacked front to back: ~20 m lateral movement corridor, ~25 m to form up facing a breach, ~15 m of slack. Carthage needs the last two and does not need the first, because the lateral corridor is inside the wall — that is what the casemate gallery is for. So 25 + 10 = 35.

Two consequences:

1. probe-nav's openGroundBehindWall min 40 is a Rome number and must become per-map. At Carthage the assertion is min 35, measured by containment against the main wall's inner face, not by nearest-centre — see the HANDOFF entry that already burned this exact instrument once.
2. The fabric comes 25 m closer to the wall than at Rome, so a breach dumps you into houses almost immediately. Intended.

#7.6 The forum — the map's one open space inside the walls

Appian: with the wall round the Cothon taken, "Scipio seized the neighbouring forum", and it is from the forum that the three streets climb to the Byrsa. [A] So the forum is the hinge between the harbour fight and the hill fight, and it is the only place inside Carthage where two armies can actually deploy against each other.

Note the exception to §2.4. An open square is neither a position nor a cross-section: put through KN/KE a 180 × 120 m real agora would come out 81 × 26 world m, a corridor. Open spaces that have to be fought in take world dimensions directly, which is the same exemption rome.ts gives its soft entries. The forum, the harbour quays and the wall's killing ground are the three places on this map where that applies.

#7.7 The Megara — a walled suburb that is not city fabric

The northern half of the walled area was the Megara: not streets and insulae but a large suburb of market gardens, orchards, hedges, ditches and irrigation channels, with scattered villas. Scipio broke into it by night in 147 BC and Appian records that the Romans found the enclosures and channels harder going than the wall had been. [A]

Build it as a third terrain class, distinct from both open ground and city:

Why this earns its place. An attacker who gets over the north end of the triple wall does not arrive in a city — he arrives in a chequerboard of walled gardens where a formation cannot hold its line, every enclosure is a small strongpoint, and cavalry is useless. It is a third kind of ground on a map that already has open plain and dense fabric, and it means the three gates in §4.5 do not lead to the same battle. It is also cheap: it is scatter and low walls, not buildings.

#8. What makes it play differently from Rome

Two besiegeable cities that play the same are one map with two skins. Here is the list, in descending order of how much it changes a player's decisions.

#8.1 Depth instead of a line

Rome is one 6.0 m curtain at 6.5 m high on top of a hill. Carthage is 74 m of belt: a 20 m ditch, a palisaded rampart, an 8 m wall, an 18 m killing ground, and then 16 m of casemated masonry. There is no single bound that takes it. A ladder party that clears the outwork is standing in a killing ground overlooked from 16 m by a wall it has not touched. A ram that breaks the middle wall has broken the second of three. The attacker has to plan a sequence, and every stage of it is a place the defender can counter-attack into.

#8.2 A wall you can be inside

Nothing else in this game has this. Concretely, four things a defender can do at Carthage that they cannot at Rome:

• Move reserves the length of the wall under cover — invisible, immune to missiles, and arriving at a threatened bay without ever showing on the walk.
• Hold a wall whose walk has been lost. Taking the top of the Aurelian wall is taking the wall. Taking the top of this one leaves 6.4 m × 4,434 m of held interior underneath you and stairs coming up behind you.
• Fight a breach in three dimensions. A hole in the outer face opens into a dark vaulted corridor 9.1 m deep, not into the city. The attacker fights along it at 8 abreast with loopholes above him.
• Burn the fodder. The magazines are full of hay and barley. A defender who cannot hold a section can deny it. (Attacker's mirror: fire in the casemate is how you clear it.)

And one thing the attacker gets: if he takes a casemate section, he is inside the wall with cover, and the defender has the same problem in reverse.

#8.3 Two axes of attack, not one

Rome presents one front. Carthage presents two, both historical, both live on the same day:

• The isthmus. Straight at the triple wall. What Manilius tried in 149 BC and failed. Firm ground for engines, but only up the middle (§3.4), and the strongest fortification in the ancient world at the end of it.
• The Taenia. Along the sandspit to the weak, low angle where the triple wall gives out (Appian's own words), through the Porta Maritima or over the harbour quay, into the cothon, up to the forum, and then the three streets. What Scipio actually did. Cheap to reach, but the corridor is 200–300 m wide with a lake on one flank, and everything past it is street fighting.

An attacker who splits between them is weak in both. That is a decision, and Rome's map does not offer one.

#8.4 No flank, no retreat, no relief

Both ends of the triple wall die on water. The map's +Z edge is the open sea. There is no way round and, for the defender, no way out — Carthage's only line of retreat is uphill to the Byrsa, and then there is nothing. Rome's Aurelian stretch has open country past both ends of the modelled curtain and a whole empire behind it.

#8.5 A vertical endgame

Rome's city is a plateau: once you are through the wall the ground is flat. Carthage's last 400 m of depth climb 45 m at 14 % up three stepped streets between six-storey blocks, with roof fighting overhead and a walled citadel at the top that no engine can reach. The battle gets harder the deeper the attacker goes, which is the opposite shape to Rome and is what makes the map worth finishing.

#8.6 A fabric that will not take a formation

Rome's 42 m artery is there so a cohort can deploy. Carthage's widest ordinary street is 7 m. Past the wall the attacker is in column, permanently, and every junction is a fight at 7 m frontage. Cavalry is close to useless inside the walls. Elephants — if the player brings them — cannot use the Byrsa streets at all.

#8.7 Water in the battle

The harbours are 14 ha of water inside the defences, crossed by a 21 m chained channel, a 4 m causeway and a 30 m freshly cut channel to the sea. A ship can arrive. A chain can be dropped. A mole can be built. None of that exists at Rome, whose only water is a river 700 m from the wall.

#8.8 Ground that decides the siege train

Soft margins on both flanks (§3.4) force every heavy engine into the centre of the isthmus. The defender knows exactly where the rams are coming and can weight the wall accordingly. Rome's approach is uniformly firm.

#8.9 Three gates, three different battles behind them

Rome's modelled circuit has one gate and everything behind it is the same city. At Carthage the north gate opens into the Megara's walled gardens (§7.7), the centre gate into the road to the forum, and the south gate into the harbour quarter. Each is a different kind of ground with a different unit type suited to it, and the attacker picks before he commits.

#9. What nobody knows, and what we are inventing

Listed so no one later mistakes our decisions for evidence.

Also flagged: Appian places the triple wall "toward the south … where Byrsa stood on the isthmus." Both halves of that are wrong — the mainland is west, and the Byrsa is not on the isthmus. It is a good example of a first-rate source being loose about geography, and a reason to trust his dimensions more than his directions.

#10. Sources

Primary:

• Appian, Roman History VIII (Punica / Libyca), esp. §§95–96 (the city, the walls, the harbours) and §§117–131 (the mole, the cut channel, the quay assault, the six days, the Byrsa). Horace White's translation is public domain.
  • https://www.livius.org/sources/content/appian/appian-the-punic-wars/appian-the-punic-wars-19/ (§95–96)
  • https://www.livius.org/sources/content/appian/appian-the-punic-wars/appian-the-punic-wars-20/ (harbours)
  • https://www.livius.org/sources/content/appian/appian-the-punic-wars/appian-the-punic-wars-24/ , /-25/ , /-26/ (the siege and the fall)
  • Perseus: http://www.perseus.tufts.edu/hopper/text?doc=Perseus:text:1999.01.0230:text=Pun. (503 at the time of writing; retry)
• Polybius — Appian's ultimate source for the city description; Polybius was present at the siege. The relevant books survive only in fragments.
• Strabo XVII — the peninsula's circumference.

Modern:

• Serge Lancel, Carthage: A History — the Byrsa excavations, the Hannibalic quarter, the street grid and insulae. Lancel and Pierre Gros directed the Byrsa sector of the UNESCO campaign 1974–1991.
• Henry Hurst — the British excavation of the circular harbour and the admiralty island; slipways, shed counts, dating.
• Lawrence E. Stager — the American excavation of the rectangular harbour and the Tophet.
  • https://isac.uchicago.edu/sites/default/files/uploads/shared/docs/ar/71-80/76-77/76-77_Punic.pdf
• Adrian Goldsworthy, The Fall of Carthage — the 9 m × 15–20 m wall and the 20 m ditch reading.
• Richard Miles, Carthage Must Be Destroyed.
• Stéphane Gsell, Histoire ancienne de l'Afrique du Nord II (1918) — public domain, the fullest nineteenth/twentieth-century topography.
  • https://www.ancientportsantiques.com/wp-content/uploads/Documents/PLACES/NorthAfrica/Carthage-Gsell1918.pdf
• 1911 Encyclopædia Britannica, "Carthage (ancient city)" — public domain; the taenia, the Byrsa at 195 ft, the harbour acreages.
  • https://en.wikisource.org/wiki/1911_Encyclop%C3%A6dia_Britannica/Carthage_(ancient_city)
• Wikipedia, Carthage / Archaeological site of Carthage / Carthage Punic Ports / Third Punic War — used as an index to the above, not as authority.
• romanports.org, "Carthage" — quay widths, the moles, the entrance.
  • https://www.romanports.org/en/articles/ports-in-focus/754-carthage.html

#11. Reference imagery — reference/carthage/

15 files, 9.9 MB, licence-verified and catalogued in ASSETS.md under "Punic Carthage reference". Full table with creator, licence, source page and SHA-256 is there; this is the index of what each one is for.

Deck eligibility: none of it, ever. This pool is layout and accuracy reference only. reference/rome2/ remains the sole blind render-quality plate pool. Mixing provenance has been got wrong twice on this project and a photograph in a render-vs-render deck measures sensor noise, not rendering.

Two cautions on using these.

1. Everything photographed is post-146. The Byrsa's visible stonework is largely Roman forum substructure standing on the Punic quarter, not Punic. The harbour lagoons have silted, been landscaped and been re-cut. Use the photographs for shape, relief, light and material; use §4–§7 for dimensions. Where the two disagree, the numbers win — see the admiralty-island warning in §6.2.
2. No reconstruction art was fetched, and none should be. A search for "Carthage" returns mostly reconstruction renders and artists' impressions of unknown provenance, and a significant fraction of what circulates is extracted from commercial games. Nothing of that kind is in this directory.

Still wanted and not yet found under an acceptable licence: a published plan of Hurst's circular harbour excavation; any measured drawing of the Magon-quarter sea rampart; a site plan showing a reconstructed line for the isthmus wall. Gsell (1918) has the last of these and is public domain, but Commons carries it as PDF, which is not a permitted format here.

#12. Build order

For the three agents, so nobody blocks on nobody:

1. Terrain and integration — §2 (the frame and the projection), §3 (the ground). Nothing else can be placed until the projection exists and the heightfield carries the isthmus, the two lagoons, the Byrsa and the coast. Also owns the MapId, the MapDefinition and making CitySystem map-aware.
2. Walls — §4, then the south wall and the sea wall. Start with the main wall's cross-section and the casemate interior, because §4.4 is the map's reason to exist and it is the highest-risk item in the whole build. Get one 200 m stretch right before running 2 km of it.
3. City fabric — §5, §6, §7. The Byrsa's relief and the three stepped streets before the general fabric; the harbours can come after, and the admiralty island last.

Cross-cutting and unowned: the soft-ground mechanic (§3.4), walkable roofs (§7.3), the harbour chain, and the per-map probe-nav threshold (§7.5). Somebody should claim these.