The forge announces what it finds merged, and an old merge surfacing late moved the recorded head
backwards and rebuilt everything built from that repository, once per old merge. A merge made
before the source was last seen is history; one that says nothing about when is taken as news.
The catalogue now carries when each source was last seen. The bus-records test follows #116:
a module's tools are every one under its own name.
One assignment of a module per node is now the rule, not a limitation —
the operator dropped the multi-assignment requirement, and the schema's
(node, module) key has been the decision since migration 0005. What
changed: Assign reports whether the assignment was new, and the command
says 'already runs — one node runs one of each (ADR 0115); nothing
changed' instead of printing 'is assigned' for a no-op, which read as
an action that happened. Idempotence stays: a repeat is exit 0, because
a script stating what is already true is not wrong.
Implements novox/hq ADR 0110 and 0111.
The seat set lives in internal/catalogue/seats.go: fourteen seats, each with a scope, what occupying
it delivers, and the record that made it one. A test asserts the count and a decision per entry, so
changing the set means finding the argument, as the host's vocabulary test does. The first set is
every seat already claimed — including the-private-network, which the network module claims from a
manifest composed in this repository's code, not from any module.json — plus npm-package-registry
(ADR 0109) and git (ADR 0111). A test parses every catalogue manifest and this repository's own and
fails on any refused claim, so closing the set refuses nothing in use.
ParseManifest now refuses a claim on a seat the mesh does not define, a seat claimed at another
scope, and a delivering seat claimed by a module that does not provide what it delivers. A
malformed claim is refused once, for being malformed.
Resolution: among several providers of a mesh provision, a pin still wins; then the holder of the
seat that delivers it; then the only provider; otherwise refused as before. ADR 0009's "never
guessed" holds — the seat is the choice made once, mesh-wide, rather than a pin per consumer node.
A provider now carries the module it came from, because a provider is a (node, module) pair and the
pair is what tells a holder from a neighbour on the same machine.
The planner's second pass is now given the first pass's holdings. Without them, a node consuming a
seat-delivered provision was refused there, and a refused node's own claims dropped out of what the
mesh holds — letting a second holder of one of its seats pass unrefused.
`seats [--json]` lists every seat, what it delivers, and each holder, derived from assignments
every time and never stored. Unheld seats are listed. A stored claim outside the set — possible
for a manifest registered before the set closed, since stored manifests are not re-validated — is
shown rather than hidden.
`build --self <owner>/<repo>` builds from a repository on the git seat's holder. The clone URL is
composed at build time from the holder's node and what it serves for git; the recorded source is the
path and the seat (migration 0032), never an address, so a moved forge changes nothing recorded.
Nobody holding the seat refuses self-hosted builds and says so; external URLs are unchanged. An
address passed with --self is refused rather than recorded as a path.
Replaces three foundation tests that defended the builder's carried package binding. The catalogue
removed that binding when the builder began requiring the registry through a real grant, so the
tests were already failing on main; they now assert the builder requires what the npm seat delivers
and carries no copy of its own, and that the forge holds the npm and git seats.
Verified: go vet clean; the whole suite passes against a throwaway Postgres (make postgres), the new
inventory tests included; gofmt clean apart from cmd/mesh-builder/stdout_test.go, which fails on
main too.
One name per thing, per the HQ glossary: the module/container/image/binary/repo
becomes mesh-controller, the seat the-controller, and the store+broker pair the
foundation (embedded base bundles, default template and example lock renamed with
their go:embed directives). No behaviour change — a pure vocabulary rename.
Claude-Session: https://claude.ai/code/session_01D6qtiYU3P9jk3pnAXyAFyx
The check read "declares no resources" as "runs nowhere", and those are not
the same. The private network declares no resources either — the control plane
computes them when it composes a machine's declaration — and it is assigned to
every machine that has to reach another one. Refusing it stopped a four-machine
bed at its first assignment.
The signal is narrower: it builds an artifact and places nothing. Made a
function of its own, because a judgement with a wrong answer this expensive
should be testable without a database — nothing guarded it, which is how it
shipped.
The image every module in the scripted toolchain is compiled on top of is
registered as a module so the mesh can build, version and depend on it. It is
not one: nothing about it belongs on a machine. Assigning it succeeded, the
machine was sent a declaration containing nothing of it, and everything
reported success — the operator had said run this here and the mesh had agreed
to something it cannot do.
A module whose resources are worked out per node is asked about separately, so
it stays assignable, which is the point of it.
The builder says what it built and the catalogue decides whether that was an
upgrade. Only the control plane knows which machines run the thing, so it is
the one that acts — and what it does is a choice somebody recorded, not a
behaviour compiled in: record that they are behind, or send it, one machine at
a time or together.
Recording is the absence of an action rather than a second path: a machine not
running what the mesh would send it is already something the mesh reports.
Defaulted to recording. A mesh that rolls out everything it builds the moment
it builds it is reasonable to want and a bad thing to arrive by default — the
first module to inherit it would be the control plane, upgrading itself out
from under the push applying it.
The builder cloned a repository and read the manifest at its root, which means one
repository per module. Nothing we have is shaped that way, so the builder could be
asked to build nothing that exists (novox/hq ADR 0069).
The path travels the whole way — named when asking, carried in the request, used
to read the manifest and as the context everything is produced from, echoed back
in the result, and recorded as part of where a module came from. Without that last
part the mesh could notice a module was behind its source and then be unable to
rebuild it, which is the worst of both.
A path climbing out of the clone is refused: a machine whose job is building other
people's repositories must not read whatever else is on its disk.
Claude-Session: https://claude.ai/code/session_01D6qtiYU3P9jk3pnAXyAFyx
`module forget` cascaded. The settings, the module's own secrets and the ports the mesh
chose all name the module by a foreign key that cascades, so removing the row took all
three and reported "forgotten" — an action succeeding into a state its own verify would
reject (novox/hq 04-ISSUES/017). A sealed secret is not recoverable afterwards, because
the mesh discarded the plaintext when it made it.
It now reads what it would destroy, names each thing one at a time, and refuses.
`--and-what-it-holds` is how somebody says they mean it, and the removal then reports
what went — this being the only record that any of it ever existed.
Reported as "operator settings do not persist, because re-registering a module
cascade-deletes them". Half of that is wrong, and the test now says so out loud: the
upsert is on the name, so `module add` at a new version leaves the settings, the secrets
and the ports exactly where they were. The command that destroyed them was `forget`, and
a wrong belief about which command destroys data is expensive in both directions — it
sends people looking for a fault that is not there, and leaves the real one unexamined.
Checked by internal/inventory/forget_test.go, which writes all three, re-registers the
module at a new version, reads them back, and only then tries to forget it.
Claude-Session: https://claude.ai/code/session_01LrgweAeERJYBg88c5cKDzF
Three faults, one file split. All from reading, all verified to bite.
Unassign now releases the module's ports. ReleasePorts existed, said
"for when it is unassigned" in its own comment, and was called by
nothing — so a fixed port stayed claimed in the name of a module that
was gone, and the next module needing it was refused by a ghost.
Kept-once-chosen is a promise about a module that is still here.
MachineSide reads addressed mappings. "127.0.0.1:8080:80" was split at
the first colon, "127.0.0.1" failed to parse as a port, and the mapping
was silently skipped — putting the filter back on the declared port,
the exact fault the function was written to end. The machine side is
the second-from-last part, which is the reading the host already
applies, and the substrate bundle writes that shape today.
An allocation race answers in the mesh's words. Two concurrent picks of
the same port used to surface as a Postgres constraint violation,
verbatim. The table has two keys, so the collision is one of two facts:
the racer was this same assignment — then its answer is the answer,
kept-once-chosen does not care who chose — or another module took the
machine port, and an unfixed pick is simply made again against the
moved free list. A fixed port that lost the race is refused by name.
Told apart by re-reading the row, not by the constraint's name, so this
does not couple to the migration's spelling.
And the artifact-store cycle tests moved to bootstrap_cycle_test.go;
machineside_test.go had quietly become three subjects.
novox/hq ADR 0009: a capability's presence gates an assignment and its detail
carries a value — seat: card1-DP-1, an architecture, an amount of memory. So
'can this run here' and 'what should it be configured as' are one fact read two
ways, and the mesh was keeping the first read and discarding the second.
The reason an absent capability is absent went the same way, which is the case
a person most needs: 'this machine has no container runtime' is the answer and
'docker is not installed' is why, and only the machine knows why.
`node show` says it back. Never reported and reported nothing stay different
things there — one machine has not run the host, the other ran it and can do
nothing, and those send a person to different places.
A build is work, not state. Everything else the control plane sends a
node is a declaration — this is what you should be — reconciled forever.
A build happens once and is finished. Putting it in a declaration would
mean rebuilding on every reconcile, or a declaration carrying "and I
already did this", which is state about an event rather than about a
machine.
So it travels on its own queue and the answer comes back correlated. One
queue, so several build machines share the work and each request is done
exactly once — which a per-machine routing key would not give.
mesh-builder is the program a build machine runs. Not the control plane,
which must not run commands on a machine; not the host, which would then
need a container runtime and git everywhere to do something almost no
machine will ever do. It holds its own broker credential and nothing
else.
Three properties that are decisions:
- a request is acknowledged only once the answer is away, so a builder
that dies mid-build leaves the work for another machine rather than
losing it with nobody ever hearing why
- one build at a time. Five at once against one runtime finishes all five
slower than it would have finished the first, and the queue is what
shares work between machines
- a failure is a RESULT. A build that fails silently is
indistinguishable from a builder that is not running, and those want
different responses
And `module list` is a catalogue: what exists, at which version, built
from which commit or handed over by hand or shipped with the control
plane, whether it is behind its source, and which machines run it. All of
that was recorded from the first build and none of it was shown, so "is
this current?" could only be answered by reading the database.
Proven against a real broker, registry and store: the mesh asked, a
builder consumed, built, published, answered; the manifest was recorded
with its commit; the source moved and the catalogue said "behind";
rebuilding caught it up with a new digest because the content changed.
Two different things were both written `requires`. A shell, a display
server and a private network have to be on the machine that needs them.
A database does not — it runs somewhere and is reached over the network.
Both were answered the same way, so requiring a database installed
PostgreSQL on every machine that ran a web application.
What a module provides now carries a scope, the same idea claims already
use, written short in the ordinary case:
"provides": ["shell"]
"provides": [{"name": "database", "scope": "mesh"}]
A mesh-scoped requirement is answered by finding the node already running
it — never by installing it here. Choosing a machine to put a database on
is a decision with consequences, and nothing resolving a web application
should make it silently. With nothing anywhere it refuses and says which
module to assign; with two it refuses and says how to choose.
Choosing is `pin <node> <provision> <from>`, kept per node because that
is the granularity the choice has. A pin at a machine that does not
provide it refuses rather than falling back — a fallback would quietly
move somebody's data. One provider does not overrule a pin either.
Resolving a node now needs to know what the others offer, and working
that out needs them resolved, so it is two passes: the first answers only
what each node offers, the second answers everything. Nothing is ever
declared from the first.
A node's plan says what it takes from elsewhere. It is the only part of a
set that stops working when a different machine goes away, and nothing
else in that output would have said so. It is also where a credential
will hang once there is a mechanism for handing one back.
One test found passing for the wrong reason: it read pins through a join
on the provider, which hides a dangling row whether or not it was cleaned
up. It counts rows now, and bites when the cascade is removed.
Connectivity was code beside the module system doing the module system's
job: every machine with an address was on the private network and there
was no way to keep one off.
A manifest can now say its resources are computed by the control plane,
which is what a peer list needs — it is derived from every machine at
once, so nothing could be written in advance. The network is a module
from there on: assigned, resolved, settled, and absent from a machine
nobody gave it to.
Three modules rather than one, because WireGuard is one VPN of several:
mesh-wireguard provides private-network, mesh-addressing
claims the-private-network, one per node
mesh-names provides name-resolution, requires mesh-addressing
networking requires both, and ships no files of its own
The last is the point. Most people want the network up and do not want
to choose a VPN, so `assign networking` takes the only answer to each
requirement silently. The day the catalogue holds a second one there are
two answers, the resolver refuses and names them, and choosing is
assigning the one you want. No flavor field, nothing to configure.
Names left the WireGuard declaration for their own module. They would be
identical over a different private network, and bundling them made one
module out of two things.
Three faults the walk found:
- choosing tailscale still installed WireGuard, dragged back in by the
names needing the mesh's own addresses. Caught now by a claim: running
two VPNs is fine, being *the* mesh network is singular.
- a requirement wanted by two modules was reported twice, identically.
- "this mesh has no hub" was reported when the real cause was that a
node could not be resolved at all. It now names the node and the why.
And a test that asserts the manifests actually shipped, after the claim
went missing from the real one while every test stayed green.
Managed files are generated and never edited, so somebody's intention about one
has to live where the generator can see it. It does now: the module ships
defaults, settings go over the top by key, and the file is produced from both.
Upstream can rewrite its half freely and the keys somebody chose survive.
Two layers, both from the start. The mesh's settings for a module, then one
machine's over those. A node that differs is expressed by differing, rather
than by restating everything the rest already say -- which would pin all of it
against future changes for no reason.
An override beats a default and there is nothing to resolve. A setting is a
statement about that key made deliberately; the default was only ever what to
do in the absence of one. So when upstream changes a key somebody has set,
there is no conflict, no merge markers, and nothing to ask.
Nested blocks merge and lists are replaced whole. Setting one field of a block
must not delete its siblings, or every setting would restate the whole block
and pin all of it. A list that merged element-wise could neither be shortened
nor reordered, and there is no correct guess about which element is "the same
one".
A module can keep specific keys for itself -- a socket path its own code
depends on -- and setting one is REFUSED rather than ignored. A setting quietly
dropped is somebody believing they changed something.
Settings that reach nothing are named at the moment they would be used, not
discovered later by the machine not behaving differently.
`plan --files` prints what a machine would be given before it is sent, because
"1 resource" does not tell you whether the merge landed.
One test kept with a note that it does not defend this code: output stability
comes from Go's encoder sorting map keys, so it passes with the merging
removed. Worth having as the thing that would catch a change of encoder, but it
is not evidence about anything written here, and it was checked.
Delivery is a comparison, not a pipeline: the control plane holds what source
exists and what has been built from it, and the difference is the work. Both
halves are written down now, so "is this current" is a question about two
columns rather than something you find out by building.
`status` answers "did my change go out?", which ADR 0010 names as the real risk
of replacing a pipeline with a comparison -- it is answerable today by opening
a pipeline, and something had to replace that.
zsh holds 4f2a9c1e, source has 9e3b7d2a
running on laptop
The machines are the point. A module being out of date is a fact about the
catalogue; which machines are running last week's version is the thing with
consequences.
Three things this had to get right.
A module with no source is never behind -- it was handed over directly, which
is how a one-off arrives, and saying "out of date" about it would be inventing
a comparison against nothing.
A source nobody has checked is not behind either. Reporting it as behind would
put every module on the list the moment provenance was recorded, which makes
the list say nothing. Fault injection found this: my first test passed with the
guard removed, because both halves were empty strings and compared equal. The
case that actually needed it -- a known commit and an unknown head -- was
untested.
And handing over a manifest by hand does not erase where the module normally
comes from. Fixing something in a hurry is legitimate; silently forgetting its
origin is not, because that record is the only thing that would say afterwards
that a machine is running something nobody can rebuild.
Also fixed the flag parsing, which stopped at the first positional argument and
silently ignored every flag after it -- so `module add thing.json --source x`
recorded no source at all and said it had succeeded. The host's own parser
documents this exact footgun and I wrote it again anyway.
The gap that has been named at the end of every report for a week. Until now a
declaration came from a person handing over a file; now it comes from what was
assigned, resolved against the catalogue, and the control plane is deciding
rather than relaying.
Everything from the module conversation, built and run on real machines:
assign laptop i3 -> accepted, brings xorg, because nothing else provides
it and there was no choice to make
assign laptop sway -> refused: xorg and wayland both claim the-seat
assign laptop editor -> refused: three modules provide a shell -- bash,
fish, zsh -- choose one
assign laptop zsh -> accepted, and the editor's requirement is answered
bash, fish beside it -> fine, nothing is claimed
Claims rather than pairwise exclusion, so a third display server would say what
it claims and need no edit to xorg or wayland. Scoped to node, site or mesh:
two DHCP servers at one site collide and at two sites do not, and the mesh-wide
one is the hub said as a claim instead of hard-coded.
Some conflicts cost no manifest field at all. The refusal above names the seat
AND the two files, because the mesh already holds every resource of every
module -- neither i3 nor sway knows the other exists.
Resource identities carry their module, so two modules may both call something
"config" without the second silently replacing the first. What a service
reflects is qualified the same way, or it would name a resource that no longer
exists and stop being restarted when its own configuration changes.
Nothing is sent until every node resolves. A push that configured three and
refused on the fourth would leave the mesh in a state nobody asked for, and the
fourth is exactly where a claim collision appears.
One real flaw found by using it rather than by testing it: assigning zsh did
not satisfy a requirement for a shell. Requirements were counted against the
catalogue without first asking what the set already offers, so "choose one and
assign it" named three modules and then ignored the one you chose. The remedy
was useless and every test passed.
jschoubben
jochen