You are viewing a potentially older version of this package. View all versions.
RavenIronStudios-Undertow-1.0.4 icon

Undertow

The sea gets its own motion. Currents run in a learnable shape - basin drift, coastal set, fast water between islands, slack behind a headland - and carry what floats on them. Tides, storm surge, flotsam. No HUD. One DLL, server and clients.

Date uploaded a week ago
Version 1.0.4
Download link RavenIronStudios-Undertow-1.0.4.zip
Downloads 75
Dependency string RavenIronStudios-Undertow-1.0.4

This mod requires the following mods to function

denikson-BepInExPack_Valheim-5.4.2360 icon
denikson-BepInExPack_Valheim

BepInEx pack for Valheim. Preconfigured with the correct entry point for mods and preferred defaults for the community.

Preferred version: 5.4.2360

README

🌊 Undertow

UNDERTOW VERSION CLIENT & SERVER

The sea remembers where it is going. The old steersmen knew it, and none of them dozed at the tiller with the coast downwind. Midgard's water forgot for a while. Now it remembers again.

Valheim's sea has weather. Undertow gives it water.

Waves answer the wind, storms raise them, a hull takes damage in a seaway. But nothing moves: a karve left at half sail on a fixed heading arrives exactly where geometry says it will, every time, anywhere on the map. Undertow gives the sea its own motion. Currents run across the ocean in a shape you can learn, and they carry what floats on them.

No map, no HUD, no icons. The sea shows its own motion the way water does: foam lying along the flow, moving with it, glassy where it goes dead. Nothing reads it out for you.


NEW IN 1.0.4

This is the Valheim 1.0.21 build of Undertow.

You It
Play on Valheim 1.0.21 Undertow runs on it; the sea, the drift and your settings are the same as before
Run a server update the server and every player together: Valheim 1.0.21 is network version 41

Nothing else is new in this version.


BEFORE YOU INSTALL

If… Then…
🀝 you play on a server put Undertow.dll on the server and every player. A boat feels the current on the machine that owns it, which is a player's, so a server-only install moves no boats.
🧭 you want to tune the sea for a server tune it on the server. Its sea wins: every player sails by the server's dials for the session, and nobody's own file is written.
βš“ you moor a boat and walk away it stays put. A boat with nobody aboard is left alone by default.
πŸ’Ύ you update from an older Undertow your settings are kept. A copy of your file is saved beside it first, and only values still at an old default are moved.

I

One current across the whole ocean, in a shape you can learn and plan a voyage around.

πŸŒ€ The drift slow, basin-wide water: the thing you plan a long voyage around
🏝️ Coastal set a stream that follows the shore, with a slight push toward it
πŸ’¨ Races water speeds up between close landmasses: a narrow gap is fast water, in one direction
πŸͺž Slack and eddies where opposing arms meet, the water goes dead, and things collect there
πŸ“– Read the saga β€” one sea on every machine, and nothing sent over the network

The current is built from a stream function, so the flow is divergence-free, the way real water is. Gyres, races and slack water all fall out of the same mechanism rather than being placed by hand. The narrows are read at two scales, so a wide strait registers as well as a gap between rocks.

It is the same on every machine, and no water ever crosses the network. The field is a pure function of the world seed, the position, the world clock and the season, so two players a thousand metres apart compute the same sea without exchanging a byte of it. What they do have to agree on is the tuning, and the server publishes that (see MULTIPLAYER below).


II

The current changes where you go, never how hard your hull can row.

Your boat The water
Under way, paddle or sail set going into the current costs you exactly the water's speed over the ground; going with it pays you exactly that. Your speed through the water is never touched.
Adrift carries you until you move at the water's own speed, whatever the hull
Beam-on to the current moves you less than a current off the bow, because a hull resists sideways drift more than forward drift
Nobody aboard not touched, by default
πŸ“– Read the saga β€” why a drifting karve and a drifting longship agree

A hull under way feels the current as drag relative to the water, which is what a current is. That is the same for every hull, so what a headwater costs you never depends on how small a boat's engine is.

A boat left drifting settles at the water's own speed. A karve was measured doing exactly that on the shipping game, with and without Njord and Sailing loaded, and a karve and a longship measured together earlier both came out near the water's speed, though their hulls damp very differently. The sideways resistance is not imposed by Undertow: it falls out of Valheim's own per-hull physics, so it differs from hull to hull.

Why unmanned boats are left alone. Vanilla already damps an unmanned hull almost to a stop, and a moored longship wandering off while you are away is not a feature. A crew that moors properly and wants the sea to punish the ones who do not can raise UnattendedDriftFactor.


III

The passage you know is a different passage half a tide later.

πŸŒ’ Tides a slow flood and ebb on a cycle you set. It swings how hard the open ocean runs and reverses the coastal stream, so departure time becomes a decision.
β›ˆοΈ Storm surge (with Ragnarok's Wrath) where a Devastating Storm stands, the water rises, there and nowhere else. A sheltered passage stops being sheltered while the storm sits over it.
πŸ‚ Seasons (with Ragnarok's Wrath) the season shifts the drift
πŸ“– Read the saga β€” a tide watched from one point

One coastal point was read again and again through a tide with wake field: the stream along the shore reversed between ebb and flood, under an open-water drift that never turns, and the peak-flood reading was predicted before it was taken.

A full tide takes two in-game days by default. Ragnarok's Wrath is entirely optional: without it Undertow logs its absence once and the sea runs regardless.


IV

Currents converge, so things gather: a second reason to know where the sea goes quiet.

πŸͺ΅ What washes up driftwood and cargo, and now and then what the drowned no longer need
πŸͺž Where in slack, deep water near a real player: an empty ocean stays empty
βš–οΈ How much capped, and reclaimed on a timer, so a long-running world never fills up with driftwood
🧺 Only vanilla items never a new prefab
πŸ“– Read the saga β€” the cap, the clock and the empty server

Flotsam is the server's job, and it was watched on a dedicated server: items spawned up to the cap and the cap held; each one was reclaimed on its timer, including items kilometres from the only player; the server's whole object table moved by exactly one per item; and an empty server produced nothing and still cleaned up. In single player and on a hosted game, the host's own player counts like everyone else.

The items come from lists you can change. They must be items that float in Valheim, or they sink out of reach, and wake floats lists them. Floating is Valheim's loss-prevention marker rather than buoyancy, so an iron shield floats and ore does not: choose for the story, not for physics.


V

The current carries a swimmer too, gently, and you can always out-swim it.

🏊 Carried a swimming body is set down the coast by the water, at a share of its speed
πŸ›Ÿ Capped below swim speed you can always make headway against the water and reach shore, at every setting the config allows
πŸ“– Read the saga β€” a safety property, not a balance dial

If a player can be held offshore by the current until they drown, the feature is wrong rather than mistuned, so the cap is enforced across every value the config permits. It was seen engaged in a race running a full metre per second, with the drift factor at the top of its range: a swimmer heading upstream still made headway, at their swim speed minus the cap.


VI

Faint foam streaks lie along the current on the water itself. Nothing else tells you where it runs.

On the water It means
Thick streaks a race: fast water
Sparse streaks a trickle
A few faint flecks, or almost nothing slack water
Streaks holding station alongside you are drifting with the water
Streaks streaming past at an angle under sail that angle is the set
πŸ“– Read the saga β€” reading the foam like a sailor reads a tide

The foam moves at the water's own speed and rides the swell. The streaks are symmetric end to end, so a glance gives you the line of the flow and only watching gives you its direction. There is no arrow, no number and no screen element.

Slack water keeps a scattering of faint flecks on purpose, so an empty-looking sea can be told apart from a mod that is not running: wake lines answers that. Set DriftLineSlackFloor to 0 if you want slack water strictly bare.

Night, fog, distance and a big sea dim the foam on their own, and DriftLineNightFloor sets how much of its daytime brightness it keeps at midnight. It is drawn by each player's own game, cosmetic only: it changes nothing about how a boat or swimmer moves, is never networked or saved, and a dedicated server ignores it. It keeps to a per-frame cost budget of its own, and thins itself out and says so in the log if it goes over. Two players on one deck see the same set, density and speed, but not the same individual streaks.


VII

Undertow changes the sea and nothing else.

🚫 No instruments no map, no compass, no wind gauge, no HUD of any kind: the foam is the sea, not an instrument
πŸ“¦ No new assets no new prefabs, assets or bundles
🌦️ No second weather no second wave system and no weather: vanilla and Seasonality own that ground
β›΅ No boat rebalancing no boat rebalancing and no new ship types

THE TOOLBOX

First ten minutes: put Undertow on the server and every player β†’ sail out and stop paddling in open water β†’ watch the foam beside the hull β†’ open the console (F5) and type wake here.

Command (F5) Answers
wake status what this machine is, what is running on it, which config layout your file has, and whether you sail by the server's tuning or your own
wake here the current under your keel: speed, bearing, depth, tide
wake field <x> <z> the current anywhere, loaded or not
wake drift whether the current is actually reaching boats
wake floats which items float, and so can wash up as flotsam
wake lines whether the current is showing on the water, and what it costs Β· wake lines reset rebuilds it
πŸ› οΈ See it running

A current has no sound and no icon, so the log is how you watch the mod work. Turn on VerboseLogging and the log carries a survey of your seed's ocean at start-up, plus a running readout of the drift while you sail.

πŸ› οΈ Change the sea on a live server

If you are an admin, edit a synced value on your own client while connected. It goes to the server, which saves it to its own config and tells every player. If you are not an admin, the server decides, using vanilla's own admin check rather than your client's word for it, and your edit stays in your own file.

πŸ› οΈ Choose what washes up

Type wake floats to see which items float. Put the ones you want in FlotsamCommon and FlotsamRare, comma separated, on the machine that spawns the flotsam: the server, or the host.

πŸ› οΈ Keep your settings across an update

When a release changes a default, your file is read before anything else, a copy lands beside it as .vN.bak before any value is touched, and only values still at an old shipped default are moved. Anything you set yourself is kept, and named in the log so you can see it was read and left alone. The [0 - Meta] section stamps which layout your file was written for; leave it be. wake status reports it.


MULTIPLAYER

Works with vanilla players Needs Undertow on the server Needs Undertow on both ends
a player without Undertow can join Β· their own boats feel no current and they see no foam flotsam Β· one tuning for everyone, published to every player the same sea for everyone: the current on boats and swimmers is worked out on each player's own machine, by the server's tuning

The server's sea wins. The current is worked out from the world seed on every machine, so they only have to agree on the tuning. The server publishes its thirteen gameplay dials (the shape of the current, the Ragnarok's Wrath switch, the drift and the swimmer settings) and every player sails by those for that session only. Your config file is never written to; leave the server and your own settings are exactly where you left them. Values from a server are clamped to the range your Undertow allows, so a server can make the sea faster and never impossible. The foam, the cost budgets, the logging and the flotsam settings stay each machine's own.

Update together. Keep the server and its players on the same Undertow, and update them together.


COMPATIBILITY

Mod
Ragnarok's Wrath (Raven Iron) found by itself. Storms raise the sea where they stand and the season shifts the drift. Optional: its absence is logged once and changes nothing else.
Seasonality Β· Seasons never touched. Undertow selects no environment and modifies no material.
Njord (Wubarrk) Β· Sailing (Smoothbrain) measured together: a drifting karve settles at the same speed with them as without, and Njord's per-hull speed cap and the current get on without a judder. Njord changes the hull's handling, Sailing its propulsion, Undertow the water.
Boat stat mods in general should get on for the same reason: they change the hull, Undertow changes the water.

SETTINGS

All in BepInEx/config/com.raveniron.undertow.cfg. Every system has its own switch, and every rate, cap and threshold can be tuned.

πŸ”’ Server-synced β€” set on the server; every player sails by them
Section Setting What it does
2 - Systems EnableDrift the current's effect on boats
2 - Systems EnableSwimmers whether the current carries swimmers
2 - Systems EnableWrathBridge read Ragnarok's Wrath for storm surge and the season
3 - The current MaxCurrentSpeed the size of the sea: the fastest the water runs anywhere
3 - The current TidePeriodSeconds Β· TideAmplitude how long one flood and ebb takes, and how much the tide swings the open ocean
3 - The current CoastalStrength how strong the stream along the shore runs; 0 for open-ocean drift only
3 - The current StormSurgeMultiplier how much harder the water runs under a Ragnarok's Wrath storm
4 - Drift DriftStrength how quickly a drifting boat takes up the water's speed
4 - Drift UnderWayDragFactor how much of the current a boat under way pays; 0 ignores it, 2 grips twice as hard
4 - Drift UnattendedDriftFactor how much of the current moves a boat with nobody aboard (0 by default)
6 - Swimmers SwimmerDriftFactor Β· SwimmerMaxShareOfSwimSpeed how much of the water's speed carries a swimmer, and the hard cap below swim speed
πŸ–₯️ Local β€” each machine's own
Section Settings
1 - Core TickBudgetMs Β· VerboseLogging
2 - Systems EnableFlotsam Β· EnableDriftLines
4 - Drift FieldRefreshSeconds
5 - Flotsam (read where flotsam spawns: the server, or the host) FlotsamIntervalSeconds Β· FlotsamPerHour Β· FlotsamMinDepth Β· FlotsamRingMinMeters Β· FlotsamRingMaxMeters Β· FlotsamMaxAlive Β· FlotsamTtlSeconds Β· FlotsamRareChance Β· FlotsamCommon Β· FlotsamRare Β· FlotsamWreckage
7 - Drift lines (the foam) DriftLineCount Β· DriftLineRadius Β· DriftLineOpacity Β· DriftLineMinDepth Β· DriftLineSlackFloor Β· DriftLineNightFloor Β· DriftLineLiftMetres Β· DriftLineBudgetMs
0 - Meta ConfigVersion (which layout your file was written for; leave it be)

CREDITS

Needs: BepInExPack Valheim 5.4.2360 Β· works alongside Ragnarok's Wrath when it is installed.

A Valheim mod by Raven Iron. See also Ragnarok's Wrath (the world reacts and remembers) and FireFront (fire that spreads).

Support Raven Iron. Every Raven Iron mod is free, and stays free: all of it, always. Nothing is held back for patrons, and nothing ever will be. If you'd like to help cover server hosting and test hardware:

CHANGELOG

Changelog

1.0.4 β€” 2026-11-01

Updated due to the Valheim 1.0.21 patch. Rebuilt and checked against Valheim 1.0.21 (network version 41); no code changes beyond the version string, and the wire is the same, so 1.0.3 and 1.0.4 still share a sea. The BepInExPack dependency is now 5.4.2360.

1.0.3 β€” 2026-09-25

Updated due to 1.0.16 Patch.

1.0.2

Fixes and cleanup. Nothing about the water or the drift changed, no config key moved, and the wire is the same, so 1.0.1 and 1.0.2 still share a sea.

  • Flotsam now washes up in single player and around a host's own player. Until now it only ever spawned near players connected to a server from elsewhere, so a single-player world never saw any, and on a hosted (non-dedicated) game only the guests did. The host's own player now counts like everyone else. Dedicated servers are unchanged, and an ocean with nobody in it still stays empty. A single-player world can now hold up to FlotsamMaxAlive pieces when you quit; they are vanilla items and vanilla's own despawn clears them.

  • The server now checks who really sent a config message. Valheim lets a client write its own sender id on this kind of message, and the server did not check it. So a modified client could pass itself off as an online admin and change the server's synced settings, or pass itself off as the server and hand every player a different sea for a while. The server now drops any Undertow config message whose sender does not match the connection it arrived on, and logs one warning per connection that tries it. The admin check, the ranges and the synced keys are unchanged. The check runs on the server, so the server needs 1.0.2; clients need nothing new.

  • Less garbage per physics tick on a client connected to a server. Reading a synced value and noting the last pushed hull no longer create a new string every tick.

  • The DLL no longer carries the build machine's folder path. Every DLL through 1.0.1 embedded the absolute path of its debug-symbols file, a path that included the build machine's user name. The build now maps its source folders to a neutral /_/ prefix, so neither the DLL nor its symbols name a local folder, and two builds of the same commit, from Windows clones with Git's default line endings, the same .NET SDK and the same game libraries, are byte-identical. This DLL was built from the commit tagged v1.0.2; the GitHub release names that commit and gives the DLL's md5.

  • The README's Support section links the Raven Iron website, https://ravenirongames.com, beside Patreon and Discord.

  • The store page's website link now goes to the Raven Iron website. manifest.json's website_url pointed at the GitHub repo; it now reads https://ravenirongames.com/, matching the README's own Support section.

Tested in game on 2026-09-24, on a Valheim 1.0.15 dedicated server (crossplay) plus a single-player world, with one client, all on the DLL built from 5446c9d, the head of the fix branch (md5 8e35436a412c89cbfe7d45a7aa23b2d5, 132,608 bytes) β€” the same code as this release; only documents and the store page's website link changed after it. Built against and tested on Valheim 1.0.15. Flotsam in single player, with FlotsamPerHour raised from its default 6 to 120 for the test: 6 spawns in about 3.5 minutes, 34–107 m from where the drift began, in 30 m of water with a 0.30–0.33 m/s current. Config sync on join: the client asked the server for its config and nothing was dropped (the "13 values in force" line was read on screen, not in a log). Admin push: 9 of 9 MaxCurrentSpeed changes were accepted in an earlier session the same day, on the same DLL, and the last one set the value back. Local config was restored on leaving a server. Both sides booted with Harmony patched 4; Undertow logged no errors or warnings on either side, and no exceptions appeared in either log. Not tried in game: a non-admin's config push being refused (checked only in code); the forged sender itself, which needs a modified client (the packet-header read was checked against the 1.0.15 game code; six harness cases cover the accept/drop predicate but not the patch that calls it; genuine admin traffic passing through the patch is the closest in-game evidence); a 1.0.1 client with a 1.0.2 server, or the reverse (the handshake is unchanged, read from the code); flotsam staying off on a dedicated server with an empty ocean (an unchanged path; the harness covers the origin count); flotsam around a listen host's own player with a guest connected, which needs a second player (new in 1.0.2: the harness covers the origin count, and single player ran the host branch in game); and wake drift naming the last hull, then reading back (none) once that hull is destroyed, which is the check for the per-tick garbage fix; the garbage saving itself was not measured (it is read from the code). Off-game: 495 checks, 0 failed.

1.0.1

  • A hull under way now pays the current as a speed, not a force β€” 1.0.0 stopped a paddled karve in 0.2 m/s of water. Reported within hours of 1.0.0 going up (Grishak: below a MaxCurrentSpeed of 0.25 he could paddle through anything; at 0.3 the water held him or pushed him back). Read out of the game's own hull prefabs the same hour: a karve's paddle is 0.004 m/s of speed per physics tick, and the push this mod applied against a hull driving upstream was the water's speed Γ— 0.02 per tick β€” the same 0.004 at 0.2 m/s of water. The push was an acceleration measured against the hull's engine, and vanilla's engines are tiny, so whether a current stopped a boat depended on the boat and not on the water. Every measurement before 1.0 had been a drifting hull, which the push models well, or a hull under Njord and Sailing, whose thrust is many times vanilla's.

    Now the ship's own speed setting decides. A hull with paddle or sail set feels the current as drag relative to the water β€” the exact correction to vanilla's own damping, which acts on absolute speed β€” so going into a current costs exactly the water's speed over the ground, going with it pays exactly that, and speed through the water is untouched. A hull adrift keeps the carrying push every drift measurement was taken on, so nothing about drifting changed. Watched the same afternoon on a paddled karve, verbose: 1.5–1.9 m/s of headway straight into 0.25–0.31 m/s of water, where 1.0.0 stalled; 3.0–3.5 m/s along a 0.44 m/s current running with it; mode adrift and the water taking the hull the moment the paddle stopped; zero exceptions. The harness paddles a karve and a raft tick by tick and pins both the fix and the 1.0.0 number as the defect it was.

  • UnderWayDragFactor, a new synced dial. How much of the water's own drag a hull under way pays: 1 is the physics above, 0 lets a boat under way ignore the current entirely (it is still carried when adrift), 2 grips twice as hard. DriftStrength now governs only how quickly a drifting hull is carried. Thirteen dials travel from the server instead of twelve; an older client drops the line it does not know, so the wire header did not move.

  • A boot line names every hull's real constants. Hulls (vanilla prefab constants; …): paddle force, sail factor, the three dampings, buoyancy, mass, and the Rigidbody's own drag (zero, on every hull) β€” because the class defaults a decompile shows are overridden by every prefab, and the diagnosis above needed numbers nobody had. INFO, once per boot, both roles.

1.0.0

What 1.0 means here. Not a feature. It is the point at which every sentence in the README about what the sea does has been watched happen on the shipping game β€” Valheim 1.0.15, a dedicated server and a client β€” and the point after which the config layout is a promise: from here, any default that moves is a migration with a backup beside your file, never a silent change. The last of those sightings were taken on 2026-09-21: flotsam spawning to its cap of twelve and then reclaiming fourteen times at 1801 s against an 1800 s timer, including items five kilometres from the only player, while the server's whole object table moved by exactly one per item; the swimmer's drowning guard pinned at its cap in a race running a full metre per second, with 1.58 m/s of headway straight upstream; one coastal point read sixteen times through a tide, swinging from 0.585 m/s ESE on the ebb to 0.272 m/s S on the flood, the flood reading predicted before it was taken; a karve drifting under Njord 2.0.5 and Sailing 1.1.9 at 0.99–1.00 of the water's own speed, against 0.99 in the same water with neither; and two Undertow builds meeting across the config wire, the older one refusing the server's tuning, saying so once, and sailing its own. The drift lines at night were looked at for the first time and turned out to be a defect; it is fixed below. What has not been watched is said where it stands rather than hidden here: Dive In is read from its source and never run; the longship half of the two-hull convergence rests on its pre-1.0 measurement; the storm palette for flotsam β€” wreckage instead of driftwood under a Devastating Storm β€” is built and has never had a storm over a spawn, so the README no longer promises it; a non-admin's config push has never been sent, so its refusal has been seen only in the code; and the two field-maths fixes below that only the harness has seen say so in their own bullets. 0.7.0 and 0.7.1 never reached the store; 0.7.2 did, and 0.8.0 has been on Hexium since 22:35Z on 2026-09-19 β€” so an install coming from 0.8.0 has nothing to migrate, and one coming from 0.7.2 gets 0.8.0's single rebase (DriftLineMinDepth) with a .v1.bak beside its config. Changes since 0.8.0:

  • The drift lines' night dimming works now, and it is a dial. It was designed in from 0.7.0 and never engaged: it read Unity's ambient light, and Valheim's midnight ambient is a moonlit grey (luminance 0.38, against 0.56 at noon) that never reached the "night" band, so the foam drew at full daytime brightness all night through three releases. Found by standing on the water with the clock forced to midnight and reading wake lines. The dimming now follows the fog colour, which runs 0.18 at midnight to 0.53 at noon and also darkens under a storm sky. The first working build went all the way to nothing at midnight, and that was too far β€” real foam is the most visible thing on black water β€” so how much of the day the foam keeps at full night is a new client-side setting, DriftLineNightFloor, set by eye at midnight and shipped at 0.7: a little dimmer than day, still plainly foam, and the storm-sky dimming intact. 1.0 is the old accidental look; 0 is a foam that vanishes after dark. The readings that decided the input are named constants in the code and the test fixture, and wake lines prints both luminances and the floor.

  • Slack water is an absolute speed now, so a server can raise the ceiling without losing its foam. "Slack" was defined as 12% of MaxCurrentSpeed, and MaxCurrentSpeed is a ceiling β€” raising it makes the fast places faster and ordinary water no faster at all, but it raised the slack line with it, and ordinary water quietly stopped drawing foam (measured: a 2.4 ceiling emptied 0.21 m/s water). Slack is 0.144 m/s outright, which is exactly what it was at the shipped ceiling, so nothing changes unless you turn the dial β€” and now you can. The old behaviour was measured on the water; the new definition is pinned both ways in the harness, and nobody has sailed a raised ceiling since. The same speed is what wake here calls Slack and what the foam thins at; they were one definition before and still are.

  • The coastal set's push toward land now points toward land on the ebb too. It always did on the flood. On the ebb it pointed off the shore, because the onshore share was scaled by the signed stream rather than its size β€” fifteen percent of the coastal term, the wrong way, for half of every tide, since 0.1.0. Found by reading the line while the tide reversal was being measured on the water; the reversal itself was and is correct. The fix is proven against a mutant (flood x βˆ’0.0771, ebb x +0.0771) and not yet read on the water β€” the reversal run predates it, and at fifteen percent of the coastal term it could not have seen it. This changes the water, so the config wire is at undertow-cfg/2: a client on 0.8.0 or earlier joining a server on this version will refuse the server's tuning, sail on its own, and say so in its log β€” once per session, not on every heartbeat, and wake status keeps reporting it. Update both ends.

  • Flotsam now says what it reclaims. Every verbose flotsam line ends with the size of the server's whole ZDO table ([n/12 alive, N ZDOs]), and a reclaim on the TTL β€” the half of the cap that bounds a long-running server, which used to happen in silence β€” logs by id, as does an item that was picked up or removed by the world. Verbose only; nothing changes with logging off.

0.8.0

  • The drift lines now show in all moving water. Two shipped values were keeping the foam off the sea, and both were found by standing in the water and reading wake lines rather than by reasoning about the code.

    DriftLineMinDepth shipped at 10 m, and that was wrong for the feature it constrains. Valheim's open ocean is a flat 30 m floor, so every race, strait, shelf and coastal set is shallower than that by definition β€” which meant the "fast water between islands" this mod exists to show was gated out of its own habitat. Measured in game: 0.665 m/s of genuine Race at 5.1 m depth, and wake lines reporting rejected shallow 63. The floor is now 2 m, where it stops being a filter and becomes what it was always described as: a beach guard. The number is not arbitrary β€” acceptance ramps over the 6 m above it, so it finishes at a depth of 8, exactly where CurrentField's own shallow fade finishes. Two unrelated thresholds became one curve. Foam on land remains impossible at any setting.

    Slack water is no longer bare. It used to be a cliff: below 12% of MaxCurrentSpeed the foam was strictly absent, on the argument that slack water is glassy. The argument was sound and the consequence was not β€” an empty sea is also what you see when the feature is broken, when a config is mis-set, or when you happen to be parked in a slack pocket, and a slack pocket can be over a kilometre across (measured). Three indistinguishable causes for one observation. The cliff is now a floor: below the threshold, density ramps from nothing at dead-still water up to a new DriftLineSlackFloor (default 0.08), so a glassy patch carries roughly a dozen short flecks where a race carries a hundred and sixty long lines.

    What this costs, plainly: an empty sea no longer means slack water. The contrast that makes the set learnable becomes a difference of density and streak length rather than of presence and absence. DriftLineSlackFloor = 0 restores 0.7's behaviour exactly, and the harness pins that it does. At and above 60% of MaxCurrentSpeed nothing changed at all β€” a race looks pixel-for-pixel as it did.

    No new frame cost: the field is sampled before the spawn test, so widening what is accepted adds no field evaluations. The pool is still hard-capped, and the capped state was already measured at 0.37 ms against a 0.50 ms budget.

  • The config migration ran its first destructive step, on a real server. Moving a shipped default is exactly what ConfigLedger was built for and had never done: BepInEx never rewrites a value already in a file, so changing the C# default alone would have shipped the fix disabled for every existing install, silently. Version 2 carries one rebase row. On Storm10 it logged config: version 1 -> 2: 1 value(s) moved to their new defaults: 7 - Drift lines.DriftLineMinDepth (your previous config is backed up beside it, .v1.bak), and the backup was verified byte-identical to the pre-migration file. A value you set yourself is kept and named instead. The ladder was built two releases before anything needed it, and the first rung turned out to be a data edit rather than new code on the boot path β€” which was the whole claim.

  • The server's sea, on every client. CurrentField is a pure function of seed, position, world time and season, which is why it needs no save file and no per-tick traffic β€” every machine computes the same water from facts they all already have. That argument holds only while the TUNING is also the same. A server that raised MaxCurrentSpeed and a client that did not were computing two different oceans from one seed, and because drift is applied by the peer that OWNS each hull, those two players genuinely sailed different seas. Nothing desynced, nothing errored, and nobody could tell.

    The server now publishes its twelve gameplay dials β€” the five field terms, the Wrath bridge switch, and the drift and swimmer groups β€” and a client adopts them in memory for the session. Your config file is never written, never backed up and never touched; leave the server and your own settings are exactly as you left them. Incoming values are clamped to the range your own build allows, so a server can make the sea faster and never physically impossible.

    What does NOT travel is as deliberate as what does: the drift lines, the tick budget, the field refresh cadence, verbose logging and every flotsam key stay yours. Those are your machine's frame rate, not the server's sea.

    An admin who changes a synced value on their own client pushes that one value up; the server applies it to its own config β€” so it survives a restart β€” and re-publishes to everyone. A non-admin's edits never leave their machine, and say so in the log rather than silently doing nothing. Admin identity is checked on the SERVER against its own admin list, accepting both the full platform id and the bare numeric form the way vanilla does.

    wake status reports which side of this you are on and how many values are in force.

    No new Harmony patch (the boot line still reads Harmony patched 3), nothing saved, nothing sent per frame, and with the mod alone on a machine it never speaks at all.

  • The field was profiled, and it was being computed twice for nothing. One Evaluate makes exactly nine WorldGenerator.GetHeight calls β€” counted with an instrumented probe, not read off the source, and now pinned by an assertion. The arithmetic around them is 0.23–0.34 Β΅s and irrelevant. So the only thing worth optimising is how often Evaluate is called at all, and two places were calling it and throwing the answer away: a boat nobody is aboard was evaluated before the check that UnattendedDriftFactor is zero (its shipped default), which on a dedicated server is the dominant cost and all of it waste, because a server owns exactly the boats no player is near; and the swimmer postfix had no cache at all, asking for 450 GetHeight a second per swimmer against a hull's 36. The same values come out β€” they are just no longer computed and discarded.

0.7.2

Built and tested against Valheim 1.0.15 and BepInEx pack 5.4.2350. A release-readiness pass over 0.7.1, which was packaged but never uploaded. Nothing here changes how the sea behaves; it is the pass that makes 0.7.1 fit to ship, and it takes a new version number rather than becoming a second 0.7.1 β€” the whole reason it exists is that a version number which means two different binaries is exactly the trap this release was caught in.

  • Log messages were mojibake. Twenty-four runs of double-encoded UTF-8 across four source files β€” twenty-one em dashes and three bullets β€” each one the correct bytes decoded as Windows-1252 and re-encoded, so β€” had become Ò€". Two of them are lines a player reads. One is Harmony attached NO patches, which is the single message that has to be legible on the day a Valheim update moves an API, and the other is the warning that nothing in the game floats. The README and this changelog were never affected. Found by reading the raw bytes; every console this was viewed through had been quietly showing it correctly or quietly showing it wrong, and neither could be trusted.

  • A persistent fault could drown its own log. The catch-all in the ship and swimmer postfixes logged every occurrence, and those run in the physics step β€” about fifty times a second, per hull. A fault that kept happening, which is the realistic shape after a game update, would bury the only instrument this mod has under its own noise at the exact moment it was needed. The first one is still reported immediately and in full; after that it is one line per thirty seconds, carrying the number suppressed in between so the log distinguishes "once, oddly" from "constantly". Vanilla was never at risk either way: the catch exists so our faults never reach the game.

  • Flotsam could leak past its own cap. When the timed reclaim of a piece of driftwood threw, the item was dropped from the tracking list anyway, while it may well have still been out there. Every such failure permanently freed a slot under the cap that is the only thing between a long-running server and an unbounded object table β€” and the log looked healthy throughout. The item now stays counted and is retried on the next sweep.

  • Two claims in the shipping documents were not true. The README stated that a hull resists sideways drift about twice as hard as forward drift, true of every hull. No such measurement exists: the project measured the along-current ratio only, and its own notes conclude the damping is a per-hull value where no single constant can be right. It now quotes the numbers actually measured β€” 0.86 for a karve and 0.96 for a longship against the water's 1.0 β€” and describes the sideways effect qualitatively. Separately, the project's status notes said the config migration had never run in-game and that the build referenced Valheim 1.0.12 assemblies. Both were stale: the migration's own line is in a dedicated server's log from 2026-09-18, and the reference set was refreshed to 1.0.15 hours after the note was written.

  • The release script now refuses a red harness. "Tests green" was this project's own definition of done and the one guard packaging could skip. It checks the harness's OUTPUT and not merely its exit code, because a script this machine's execution policy refuses also exits non-zero β€” a distinction that once let a mutation suite score twenty out of twenty without compiling a line.

  • The BepInEx dependency now names 5.4.2350, the pack actually in use, rather than the 5.4.2333 carried unchanged since the first release.

0.7.1

The config file migrates itself. BepInEx merges a new key into an existing file at its SHIPPED default β€” and a shipped default is chosen for a fresh install. It says what a NEW world should feel like, not what an existing one already feels like. When those differ, an owner who changed nothing gets a different sea, with no error and nothing in the log. This closes that door before it is ever opened.

  • The machinery, from the family. Wu'barrk's from Wings of the Valkyrie, by way of Valkyrie's Cargo, and matching the ports that landed in Ragnarok's Wrath and FireFront the same week. The raw file is read BEFORE any value binds, a [0 - Meta] ConfigVersion stamps the layout, and three kinds of change are expressible: a value still equal to an OLD shipped default moves to the new one, a key that is ABSENT can be given a legacy value that preserves how your world already behaved, and a key no current build binds is dropped instead of riding along forever as a BepInEx orphan. Anything an admin actually set is kept untouched and named in the log, so you can see the migration read it and leave it alone.

  • It cannot lose a setting. Before any change that would overwrite or delete something, a copy of the previous file lands beside it as .vN.bak; if that copy cannot be written, nothing is changed and the version is left unstamped, so the whole thing retries next boot instead of half-happening. A migration that fails never stops the mod loading.

  • Version 1 changes nothing, and that is the point. Undertow's config has only ever GROWN β€” no default has ever moved, no key has ever been renamed or removed, no range has ever narrowed. Checked against three independent records that agree: the source at every commit, the shipped 0.5.1 and 0.6.0 binaries, and nine real config files. So this release only stamps your file. The ladder exists before the rung that needs it, because the alternative is writing it under pressure on the day a default has to move.

  • 0.7.0's drift lines stay ON for existing installs. They were considered for a legacy value and deliberately left alone: they are client-side cosmetics that touch no world state, a dedicated server ignores them entirely, and they are the feature 0.7.0 was for. EnableDriftLines = false is one line if you want the old look.

  • wake status now reports the config layout version, and the migration's own boot line when this boot migrated anything.

  • Harness 248 β†’ 357. The decisions are pure (Core/ConfigLedger.cs) and every rule is pinned: a backfill never overwrites a key that is present, a retirement never fires on one that is absent, an admin's value is never mistaken for a default, a commented-out setting stays commented out, and the file is read before the first bind. Twenty-eight deliberate breakages were applied to the shipping source and all twenty-eight were caught by a named assertion. An independent review then went over the finished code and found six more real defects in it, each now fixed and pinned the same way.

  • Two more corrections, found by auditing the sibling ports (same day).

    • A retirement that failed was reported as harmless, and the version stamped anyway. Binding and removing a key both touch the file, so a transient lock β€” antivirus, cloud sync, a config manager, a second process in the same directory β€” takes the drop down through nobody's fault. The stamp then wrote "already migrated" and the retirement was never retried: a one-second lock made permanent. The drop reports upward now and the version is left unstamped, so the next boot tries again. Undertow retires nothing today, so this is a guard against the first rung that does β€” which will be a release, on somebody else's config file. Finish takes the plan as a parameter for the same reason ConfigLedger.Plan takes its tables: an empty shipped table must not leave live code unmeasured until the rung that uses it ships.
    • wake status reported the plan's INTENT, not what happened. The summary is written before a single step runs, and a refused step logged a warning and nothing else β€” so the one line wake status printed could claim a key was dropped that is still in the file. Refusals now correct the summary, and the count is per boot.

    Harness 357 β†’ 367, and thirteen mutations of the migration are each caught by a named assertion.

0.7.0

The current, visible. Run in-game the day it was built β€” Storm10, Valheim 1.0.15 on both sides: 0 exceptions, streaks carrying the field to a degree of bearing, height agreeing with vanilla's water to the millimetre, 0.37 ms a frame at a full pool, and a reading by eye β€” "long ways along the flow, a line instead of an arrow". Night, a storm and a long sail are still owed; docs/BACKLOG.md task 7 has the numbers and what remains.

  • Drift lines. Faint foam streaks lie along the current on the water itself, move at the water's own speed, ride vanilla's wave surface, and are absent in slack water, which stays glassy. Thick in a race, sparse at a trickle; night, fog, distance and a big sea dim them on their own. From a drifting hull they hold station; under sail they stream past at the crab angle. The streaks are symmetric end to end, so a still frame gives the line of the flow and only watching gives the sense.
  • Still no HUD. Nothing is drawn in screen space and nothing reads the current out β€” no arrow, no number, no icon. The locked "no navigation instruments" decision stands; the one sentence that changed is the premise's "never on the screen", now "never as an instrument".
  • Procedural, client-only, off the gameplay path. The particle system, its texture and its material are generated in code: no prefab, no asset, no bundle, nothing networked or saved. It adds no Harmony patch (the boot line still reads Harmony patched 3), owns only its own particles, latches itself off after a single error, and halves its own particle count when it exceeds a per-frame budget β€” and says so in the log.
  • wake lines reports state, shader, spawn accounting with rejection reasons, the streaks' mean bearing and speed against the field's own, the nearest streak's height against vanilla's Floating.GetWaterLevel, the scene inputs, and the measured cost. wake lines reset rebuilds the emitter β€” the console's first mutation, local and cosmetic.
  • Config. EnableDriftLines under 2 - Systems (default on); new section 7 - Drift lines: DriftLineCount 160, DriftLineRadius 60, DriftLineOpacity 1.0, DriftLineMinDepth 10, DriftLineBudgetMs 0.5. Per machine by nature β€” two players on one deck see the same set, density and speed, not the same individual foam.
  • Harness 162 β†’ 248. The streak maths is pure (Core/DriftLineMath.cs) and every rule is pinned: no foam in slack water or on a beach (cross-checked against the field's own Slack classification over 90 000 points), no visible rim, nothing pops, no glow at night, a texture with no head end, an injective memo key. Each assertion was proven to fail under a mutation.

0.6.0

Valheim 1.0. If you updated the game, 0.5.1 was broken and this is the fix.

  • Verified in-game on 1.0.12 (2026-09-12), client-side: loads with no exceptions, all three patches attach, the console registers, the Wrath bridge resolves, and the drift force ran on a live karve for 182 samples across all four field terms. In slack water the hull converged on the water's own speed; in a race it reached 0.91 median and never 0.98, which is recorded as an open question rather than a pass. Flotsam spawning, storm surge, the tide cycling, any second hull and the swimmer drowning-guard were NOT re-tested. See docs/BACKLOG.md task 6.
  • Rebuilt for Valheim 1.0.12. No behaviour changed β€” currents, tides, flotsam and every tuning value are exactly what 0.5.1 shipped. What changed is the game underneath: 1.0.7 added a parameter to Terminal.ConsoleCommand's constructor, and .NET resolves a call like that by its exact signature at runtime, so the wake console registration in the old binary threw MissingMethodException on a 1.0.x game. The source was always fine; the binary was compiled against an API that no longer exists.
  • Why 0.5.1 looked healthy. It compiled clean against 1.0.7 the day the update landed, so nothing asked to be fixed. A clean build proves the source matches today's game β€” it says nothing about a DLL built in August. The break was found by reading the shipped binary's own reference table and resolving each entry against the live game assemblies.

0.5.1

First release. The sea gets its own motion.

  • Currents across the whole ocean, in a shape you can learn: a slow basin drift, a stream that follows the coast, fast water between close islands, and dead water behind a headland. Built from a stream function, so the flow is divergence-free the way real water is β€” gyres, races and slack all come out of one mechanism instead of being placed by hand.
  • The same on every machine with no network traffic. The field is a pure function of the world seed, position, world clock and season. Verified by a server and a client independently producing identical readings for the same points.
  • Tides. A slow flood and ebb that swings how hard the open ocean runs and reverses the coastal stream, so a passage you know is a different passage later.
  • Boats are carried, never braked. A drifting hull settles at the water's own speed, and a raft, a karve and a longship all agree β€” the push fades as a hull takes up the water's speed rather than being calibrated against damping constants that differ per boat. Unattended boats are not touched by default.
  • Flotsam. Driftwood and cargo gather in slack water, with wreckage instead while a storm is overhead. Vanilla items only, capped, reclaimed on a timer, and spawned only near a real player, so an empty ocean stays empty.
  • Swimmers are carried gently, and hard-capped below swim speed: you can always out-swim the water and reach shore, at every setting the config permits.
  • Storm surge with Ragnarok's Wrath. Where a Devastating Storm stands, the water rises β€” there and nowhere else. Optional: without RW the bridge logs its absence once and the sea runs regardless.
  • The wake console β€” status, here, field, drift, floats.

Requires the mod on the server and every client: boat physics runs on whichever machine owns the hull, and that is a player's.