A link's routing metric could only be one of four presets (Prefer,
Normal, Avoid, Backup only). Links now also get a Custom option, closer
to how real networks tune IGP metrics:
- Multiplier mode scales the distance cost, (km + 60) × 0.1 to 100.
- Cost mode sets an exact cost, 1 to 65535 like an OSPF metric.
The preset buttons stay plain. Picking Custom explains how a link's cost
is worked out and lists what each preset comes to on that link. Choosing
Custom starts from the current preset and switching modes keeps the
effective cost, so nothing moves until the player changes the value.
Like the presets, it applies to every link between the two cities.
The cost formula lives in one helper, linkW(), used by compute() and the
traceroute paths. Custom links store mk/mv, which sanitize() repairs and
clamps; old saves load unchanged. The link panel skips its daily
re-render while the number field has focus, and Enter applies the value.
Tests cover routing equivalence with the presets, rerouting by cost,
mode switching, input clamping, saves, and the panel at desktop and
phone width. French, Italian and Portuguese for the new strings.
Perf on the big network: no change beyond noise, routing identical to
main.
Porto Alegre and Florianopolis are 65 px apart at the old maximum zoom,
and each city grabbed taps within 38 px, so the link between them could
not be selected at all. The maximum zoom goes from 6x to 16x (city icons
stop growing at 1.35x, so dense regions spread out), and a city now wins
a tap outright only on its icon; in the rest of its tap circle a link
right under the pointer wins. The hover cursor follows the same rule.
Dragging from a POP still draws a new link.
Wholesale alien waves and spare-pair leases already existed, but only as
random offers, and the fiber only said how many channels were sold. Each
dark fiber in service now has a 'Selling spare capacity' section: free
channels, who rents what and for how much, a pending offer with Accept,
and 'List it with a capacity broker', which creates an offer for that
fiber at once at the usual price minus a 10% broker cut (alien waves on
free channels first, then a spare pair in your own trench). One listing
per fiber at a time. Random offers use the same helpers and draw the same
random numbers in the same order, so seeded games are unchanged.
- #113: the goal was a Marseille-Singapore path and 300 Gbps served
anywhere, which broadband with local transit reached on day 0. It is now
the path plus 80 Gbps on your links between Asia and the rest of your
network for 30 days in a row (S.stats.asiaRun). A 100G route around Africa
gets there in about two months and costs about 590k a month against 2.5M
of starting cash, so it is a real challenge. The scenario card shows
today's traffic and the streak.
- #84: Poste Restante, Luxembourg's eyeball network, at LU-SICK in
Bettembourg (selective, buys transit from Tier 1s, so Tier 1 reach holds).
Citrus Belgium's blurb now says it is Tangerine's Belgian daughter.
- The wave seller and wholesale customer 'GTT (Got Tangled Twice)' used a
real company name; it is now 'G&T (Gin and Transit)'.
- The scenario start toast puts a full stop after the goal.
S.scrubAt/S.scrubTier become S.scrubs, one center per city, each with its
own size and monthly fee; old saves carry theirs over in sanitize().
An attack goes to the nearest working center with a backbone path, and
spills to the next when it is full. A remote center's flood (plus a share
of the victim's own traffic) is reserved along the backbone path after
the L2 circuits, like a circuit: it fills those links, the round trip is
shown in the NOC, and what the links cannot carry is not scrubbed and
lands on the city. Local centers behave as before.
DDoS protection customers lose mood while their attack detours 30+ ms,
with the reason in their row; a new requirement asks for a center on the
customer's continent. The Infra tab names the nearest center and its
distance, and How to play explains it.
Goals gets a Tier 1 card once it is the current goal (or the title was
lost): transit ports left, Tier 1 PNIs with the missing ones named, the
club requirements they wait for, and reach without transit with the
networks you would lose. The Networks tab gains an 'out of reach without
transit' filter and a single-network view with every PNI city and
exchange, your POPs first, then nearest. The NOC transit to-do list names
the nearest city and links to the network.
Routing counts reach without transit in the same pass. Places and
distances are cached until the POP set changes, and exchange membership
once per routing pass, so the Networks tab costs what it did.
- Upgrades take effect: the optic speed cache from #89 is now kept per cable
state (subsea, terminals in service, engineering in service), so a cable
no longer stays at 400G after both upgrades are commissioned.
- Refunds: a cable remembers what was paid for it, design and upgrades
included. Calling off a 7x low-latency cable returned 10% of its price.
'Unusable' now means no optic lights it even with both upgrades.
- Length limit: #88 counts the shortest design and both upgrades. The order
screen says which upgrade a route needs and will not order without it;
the order action checks the same. New York to Sydney needs terminals.
No leased-wave payback hint where no leased wave reaches.
- Choices start fresh when you open the own-cable view and after ordering.
- The 'consider terminals or engineering' hint only on cables you own.
- Low-latency builds take 1.7x as long (London-New York ~2.9 years), not
3.5x; optimized 1.3x. Each design is one tappable card.
- The low-latency factor is worked out once per route instead of on every
latency lookup. 'Number.isFinite' alone, no repeated guard in seaQuote,
the express factor folded into seaLatencyFactor, '+{n} days' as one
translatable string.
- Portuguese and French for the new strings; one CHANGELOG bullet (#65).
Big network vs main: no slowdowns, identical routing.
"Caisse · Banque" wrapped onto two lines in the phone's top bar at 360 px
and cut off the monthly line under the cash. "Cash · Banque" fits on one
line like the Italian and Portuguese labels do.
After #115: upgrading an existing PNI, in the Peering tab and in the NOC's
connection list, also showed only the one-off cost. Both now read
'Upgrade 100G $2.3k + $350/mo', like transit. The two old labels without a
monthly part are no longer used and leave it.js and pt.js.
Networks follow PeeringDB's 14 Düsseldorf facilities: NTT, Tell-Ya,
Cogentle, Drizzle, Voda-Phony, Doitsche Telekomm and the Crowdflare,
Amazin', Microsnoft and BiteDance clouds (NetChill and Akamaybe are
everywhere already). No nudge needed: 27+ map units from Frankfurt,
Amsterdam and Brussels.
The #16 shark test compared .land, which ignores Cairo's land2, so a
valid overland Mumbai-Cairo ask counted as crossing an ocean once the
extra city shifted the seeded draws. It now uses isSub like the game.
The buy buttons only showed the one-off setup (1500 + setup/2), so a
$1.8k cross-connect read like a monthly bill, while the real monthly
fee (PORT[i].pni) was shown nowhere. Reuse the existing translated
'{port} <small>{cost} + {mo}/mo</small>' label, like transit and IX.
CITY_PDB holds, per city, the carrier-neutral facilities and network
presences from PeeringDB (within 50 km or same city name) and the ASN
count of its country from bgp.he.net, pulled 2026-10-25.
- Transit: the regional multiplier is scaled by competition, up to 1.8x
where few networks are present. Hubs (1000+ presences) are unchanged,
and hand-set markets (Korea, Belgian red tape) keep their price.
- POPs: 1.25x to build and run with one or two facilities, 1.5x with none.
- Offers: countries with 2000+ ASNs bring extra ISP customers.
Honolulu, Guam, Koror, Majuro, Port Moresby, Honiara, Noumea, Suva,
Nuku'alofa, Apia, Pago Pago and Papeete, each its own landmass with its
real outline on the map. Exchanges only where one runs today (DRF IX,
Guam IX, Fiji-IXP, CAN'L IXP, Solomon Islands IX); PNG and Tonga have
none active, Samoa's is still planned.
Cables follow real landings: new pairs on Southern Cross, Hawaiki, AJC,
JGA, Bifrost and Humboldt (existing pairs untouched, so saves keep their
links), plus SEA-US, AAG, Echo, HANTRU-1, Coral Sea, Gondwana-1, Tonga
Cable, Tui-Samoa, ASH, Honotua, Manatua and SX NEXT as puns.
Subsea waves have no regens, so a POP on an island splits a crossing into
shorter spans: California-Sydney 400G -> 1.6T per wave via Honolulu and
Suva, LA-Hong Kong/Manila 400G -> 800G via Guam, Santiago-Sydney
400G -> 800G via Tahiti (1.2T on own fiber). LA-Tokyo stays 800G leased:
Honolulu-Guam is just over the 6,000 km 1.6T reach.
Twelve island networks, all with Tier 1 upstreams so Tier 1 players keep
their reach. PacNOG in Suva, unreliable power in Guam, Port Moresby and
Honiara, volcano cuts for Tonga, and the Stepping Stones scenario
(1.6T hop by hop from Los Angeles to Sydney).
The group split runs over an index range with one scratch array instead of
map/reduce/forEach and slice, leftover exits go into one reused array and are
sorted in place (insertion sort for short lists; both sorts are stable, so
the order is the same), flows are split without per-flow arrays, and the hot
loops use indexes instead of for...of. Without Auto-TE, exits whose path is
already full are not called at all. One routing pass on the big network
allocates 23 MB in Chromium instead of 66 MB.
Routing results are identical. The speed test's plain path now also uses the
original group split and sort, and a new test routes around a full 10G link
with Auto-TE (and a spill script) and compares both paths; planted bugs in
the split, the sort and both full-path skips are caught.
Big network vs main, two runs: headless routing -23% to -30%, Chromium
desktop -14% to -27%, phone -15% to -27%. Steps 2-4 together vs c76dd4d:
Chromium desktop routing -49%, phone -34%, a game day -26% and -30%.
Capacity only shrinks while traffic is routed, so a path with one full link
or node fits nothing for the rest of the pass: push marks it and later tries
return at once (92% of pushes on the big network went into such paths). The
nearest group of exits is found in one scan instead of two, and exits that
filled up are dropped from the shared lists once a quarter of a list is full.
Routing results are identical to before. The speed test's plain path now also
runs without these, with Auto-TE and with a RouteScript that keeps headroom on
every link; planted bugs in each of the three are caught.
Big network vs main, two runs: headless routing -51% to -58%, Chromium
desktop -22% to -31%, Chromium phone -15% to -29%.
Adds site/i18n/fr.js with all 2541 strings, registers Français in the
language list (picked automatically for fr-* browsers), caches it in the
service worker, and runs the full screen tour and the Signal-link check
in French in the tests.
npm run perf -- --browser times the same workloads in Chromium as a desktop
and as a phone (390x844 at 3x, CPU slowed 4x), one browser per version. Each
run collects garbage first and ends by forcing layout and the map canvas, so
deferred work counts. The jsdom DOM is about 30x slower than a browser's and
has no layout, so the interface rows need this.
Shared machines swing single runs by 30% on identical code, CPU time
included. A row now counts as slower or faster only when it moved 15% and a
Mann-Whitney U test gives p < 0.01 (15 runs a side by default). Identical
code: no flags in three full runs; a routing slowdown planted in a copy is
flagged in jsdom and on the phone.
The panel workload renders both selections every run instead of alternating,
which split its samples in two. CI drives the runner's Chrome through
playwright-core, no browser download.
Its fallback ladder skipped 1.2T, so between 4,000 and 5,500 km on land
(6,000-8,250 km subsea) it carried 800G while the cheaper 1.6T optic
carried 1.2T. A test now checks that a bigger optic on the same line
system never carries less than a smaller one at any distance.
800G + 400G on one route was drawn as "2 × 1.2T", which reads like two
1.2T waves. Mixed groups now list each size, largest first, e.g.
"2×800G + 400G"; equal sizes keep "4×800G".
- #88: own subsea cables longer than the best subsea optic reach are refused
with a reason; cancelling one mid-build refunds 75% of the unlaid part, and
an unusable one from an old save is refunded in full.
- #36: your own IX picks building A, B or both (second switch, +50% fabric
fee) instead of showing site and LAG buttons that did nothing.
- #105: no maintenance window or incident when swapping the only router in a
POP nothing depends on yet.
- #104: declining a loan redraws the top bar and saves.
- #50: a city with no cable landing names the nearest landing station.
The reference test from #106 now patches the loop as it looks after #96
(business-first passes) and also covers prefer, prepend and protect
customers. It catches an exit index that ignores RouteScript preferences,
which the shortcut test in speed.test.js cannot see.
npm run perf -- --save accepts files from the in-game Export save, so
scripts/benchmark-performance.js and its README section are not needed.
The performance skill documents --save, both equivalence tests and how to
handle noisy runs.
Liquid Lunch, Sea-Comedy, Angle-a, Djibouti Calling, Empty-N, Air-Tell,
Ethio-Tele-Calm and Atlas Shrugged get upstreams, so a Tier 1 with PNIs to
the other Tier 1s still reaches everyone, and their traffic stops falling
through to transit (routing on the big network: +10% to +1% vs main).
The Africa changelog bullet now opens with a short first sentence.
A chat card in the menu and How to play, a footer link on the title screen
(Italian and Portuguese included), plus README, CONTRIBUTING and a contact
link in the issue chooser.
Without an upstream, the seven networks from #92 could only be reached by
peering in Latin America, so every Tier 1 would lose reach (and the
title) once the courtesy route for new networks ran out.
Suggest protect customers only where IP customers would get unhappy
(under 95% served): it costs a second routing pass, about +65% on a big
network. Point at the exchange the network actually peers at, where you
have no port yet, instead of the city's main IX. And the draft admits it
has never been paged at 3 a.m.
Ported onto #89's routing loop: the prepend share is its own transit exit,
local-pref and prepending get their own ranks, and protect customers runs
business flows through every rank before home users. Portuguese for the
five new strings.
scripts/perf-bench.js times routing, a game day, an action, panels, the
NOC dashboard and the map on tests/fixtures/big-network.json (104 POPs,
950 PNIs, 832 caches), base and head in alternating turns, and says
whether routing results changed. perf.yml runs it on every PR against
the base; perf-comment.yml posts the table as one updated comment.
New performance skill, CLAUDE.md rule, PR template and docs.
Builds a copy of compute() with the own-exits shortcut, the first-group
scan and the per-network exit index switched off, and compares both on a
busy network with Auto-TE, preferred and backup exits, two upstreams per
city, route servers and two cache boxes per CDN. Verified it fails when
either shortcut is deliberately broken.
- prefer/avoid transit PROVIDER: per-upstream local-pref (ranks 2.9 and
3.5, so never ahead of peering).
- prepend transit PROVIDER 1-3: moves 35/60/75% of that upstream's port
to a last-resort share. Prepending is a hint, not an order.
- protect customers: business flows take every rank first, and IP
business customers are judged on their own served share.
- Optimize: drafts a script from live traffic (spill, protect, cheapest
and priciest upstream) plus comments for IX and PNI opportunities.
Cities in nineteen countries, each with a pun IX and a country code (cc).
Overland fiber between two African countries costs 2x. Kinshasa, Lusaka
and Harare get blackout, Durban load shedding.
Cables follow real landing points (submarinecablemap.com): ACE, SAT-3,
MainOne, Maroc Telecom West Africa, SAIL, SEACOM and DARE1 as puns, plus
new pairs on 2Africa, WACS, Equiano, EASSy, TEAMS, PEACE, AAE-1, Glo-1,
Medusa and SACS. All pairs are drawn from one sea track around Africa.
Existing pairs are untouched, so saves keep their links.
Brazil gets every IX.br location (ix.br/localidades/atuais), each with a
cheap nonprofit IX.brrr. Overland fiber into the Amazon costs 2x. New
national networks (Oi Oi, TIM-TIM Mobile, Brisanada) and a scenario,
Order and Progress: ports on 20 IX.brrr exchanges including Manaus.
Latin America: Montevideo, Asuncion, Cordoba, La Paz, Quito, Medellin,
Caracas, Guatemala City, San Jose, Queretaro, Monterrey, Guadalajara,
with Moviestar, Liberty Latte America, Antelope and Tiger Mobile.
SAm-1 gains a Salvador landing and Tannat a Montevideo one (new pairs
only, so old saves keep their links).
Dense clusters are spread apart with display-only CITY_NUDGE entries;
Rio no longer needs its old nudge, which put it north of Vitoria.
On a phone the NOC dashboard, scrolled past its top edge, fills the whole
screen, but the map underneath was still drawn every frame: about 4 ms per
frame on a big network on a fast PC, several times that on a phone. While a
panel covers the map completely it is no longer drawn, and it comes back as
soon as any of it shows. At the top of the dashboard the strip of map above
the card keeps moving, and on desktop, where the map shows around the
dashboard, nothing changes. No payday floaters are made while the map is
hidden, so they cannot pile up out of sight.
Four more Brazilian cities, each with an IX.brrr exchange (nonprofit, cheap
ports) and provedor offers. Clarovision and Viva Vivo serve them, and
Capybara Connect stops in Porto Alegre. The Samba Peering achievement now
names its three exchanges instead of saying "all three".
- Payday floaters were stamped with performance.now() but aged with the map's
animation clock, which stops while paused. After a pause they never faded:
they sat off-screen and were drawn every frame, about a hundred more per
payday on a big network, so long sessions got slower and slower. They are
stamped with the animation clock now.
- While paused or with animations off, the map is only redrawn when the camera,
the selection, hover or drag, the network or a new day changes it, plus once
a second and while the land picture catches up after a zoom. A paused frame
drops from ~4 ms to ~0.05 ms.
- After a day ends, the open panels update on the next frame, so one frame does
not carry both the day's simulation and the panels. With a day on every frame
they still follow, one frame behind.
The Connections tab drew every connection at once: 1,486 rows with their
buttons came to 32,000 elements, and laying them out again after every day
took about half a second while the dashboard was open (far longer on a phone).
The Connections and Customers lists now show 100 rows and a "+100 more"
button. Filters, sorting and totals still cover every row, and changing a
filter or the sort starts the list short again.
A network with a hundred POPs and a thousand peering ports froze the game for
most of a second every day and after every click, because routing checked all
~1,800 exits for each of ~14,000 flows. Now:
- compute(): exits are asked once per network which networks they reach, each
flow scans only those, and the first ECMP group is found without sorting. The
graph runs on city indices with a typed-array Dijkstra (same heap moves, same
ties), routers and per-link traffic are kept in typed arrays, the diverse-path
check reuses one live-link graph, and R.ef is put together when a link is read.
Results are bit-identical to before.
- Panels and the NOC dashboard are patched in place (morph) instead of being
rebuilt, so the browser no longer restyles 1,500 rows every tick. Buttons whose
action changed are swapped out, keeping #44's tap protection.
- The Networks tab works out each IX port's sessions once instead of once per
network.
- draw() reuses screen-point buffers, a per-frame pair index and cached label
widths, strokes the dotted fiber line on the path it already has, and uses
native roundRect: about a third fewer canvas calls per frame, same output.