`zbase32 0.1.2` is LGPL-3.0+ — the only hard copyleft dependency in the whole
Rust graph and the last remaining blocker for the MIT release
(docs/LICENSE-COMPLIANCE-AUDIT.md §2). Statically linking LGPL code into a Rust
binary obliges us to ship relinkable objects, which is impractical for a node
image.
The audit offered two routes: the MIT `z32` crate, or an original
implementation. Took the latter — z-base-32 is an alphabet substitution over a
bit stream, so ~60 lines removes the blocker while adding *zero* new
dependencies rather than trading one supply-chain entry for another.
**Byte-compatibility was the requirement, not a nice-to-have.** A `did:dht`
identifier IS this encoding of an Ed25519 public key, so any drift would
silently rotate every node's DID and orphan its already-published DHT records.
So the semantics were not guessed: I read the vendored zbase32-0.1.2 source to
extract exactly what `encode_full_bytes` and `decode_full_bytes_str` do —
including that decode truncates to the next lower byte boundary, which is why a
52-character string round-trips to 32 bytes while discarding 4 padding bits.
A model implementation was then validated against three independent sources
before any Rust was written, all five vectors agreeing:
encode(b"testdata", 64) -> qt1zg7drcf4gn (crate doctest)
encode_full_bytes("Just an…") -> jj4zg7bycfzn… (crate doctest)
decode_full_bytes("qb1ze3m1") -> b"peter" (crate doctest)
encode([f0,bf,c7]) -> 6n9hq (Zimmermann spec)
encode([d4,7a,04]) -> 4t7ye (Zimmermann spec)
The module pins all of those plus four known 32-byte keys, a 0..40-byte
round-trip sweep, a 52-char/round-trip check over 64 keys, rejection of the
characters z-base-32 deliberately omits (`l`, `v`, `2`, `0`) and of non-ASCII,
and an alphabet/decode-table consistency check so the compile-time reverse table
can't drift from the alphabet.
`did_dht.rs` gains `did_for_a_known_key_is_stable`, which pins the full
identifier string for a known key — the regression that would actually hurt,
asserted at the call site that gives the string its meaning.
Dropped from Cargo.toml and Cargo.lock (7 lines); no other user in the tree.
Verified: 28/28 network tests pass, zero copyleft crates remain in the lockfile.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The audit has claimed since 2026-07-23 that these were git-rm'd. They weren't —
only the web/dist copies went, and all of them were still tracked at HEAD nearly
three weeks later, in a repo about to be published under MIT.
Deleted (~40.7 MB):
neode-ui/public/assets/fonts/Courier_New/{CourierNew-Bold,CourierNew-Regular}.ttf
neode-ui/public/assets/fonts/Benton_Sans/BentonSans-Regular.otf
neode-ui/public/assets/fonts/Redacted/redacted.regular.ttf
neode-ui/public/packages/wireguard.apk (17 MB)
neode-ui/public/packages/atob.s9pk (23 MB)
Courier New is Monotype proprietary and Benton Sans is a commercial Font Bureau
typeface — neither is redistributable. wireguard.apk carries GPL-2.0 libwg
components, so shipping it triggers a source offer. atob.s9pk is a Start9
package of unknown license. Redacted's upstream is OFL-1.1 but no license text
was shipped; deleting was cheaper than sourcing it, since it was unused.
Verified unreferenced before deleting, not after:
- Every @font-face rule in the tree (2 in src/style.css, 2 in
public/entropy/index.html) loads Montserrat. None of these files was ever
loaded by CSS.
- The three `Courier New` hits (tailwind.config.js `mono`, two public HTML
font-family lists) name the *system* font as a fallback — they are not
@font-face sources, so rendering is unchanged.
- wireguard.apk and atob.s9pk have zero references in any tracked file.
- These live under neode-ui/public/, which Vite copies verbatim rather than
resolving, so their absence cannot break a build.
Deliberately kept: neode-ui/public/packages/archipelago-companion.apk, which IS
live (staged by .githooks/pre-push, the Android release flow, and the in-app
pairing QR); Montserrat (OFL.txt) and Open Sans (LICENSE.txt), both properly
licensed; and neode-ui/test-install.sh, which the same audit line listed but
which is not a licensing concern.
Audit updated: §1 and §3's font/package items marked closed, the false DONE
entry rewritten as a history note rather than deleted — a DONE line here is a
claim and should be re-verified with git ls-tree, which is exactly the lesson.
§2 (zbase32, LGPL-3.0+) is now the last hard blocker.
Side effect: ~40 MB off the frontend OTA tarball.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The runbook still opened with "the catalog is accepted unsigned (migration
window) and the anchor is unpinned (RELEASE_ROOT_PUBKEY_HEX = None)". Both have
been true-for-a-while false: `trust::anchor::RELEASE_ROOT_PUBKEY_HEX` is a
`Some(...)` with a verification note in its doc comment, and
`releases/app-catalog.json` carries both a `signature` and a `signed_by`
did:key.
This one matters more than a normal stale status: a reader taking the header at
face value would think the fleet still accepts unsigned catalogs and that the
one-way anchor-pinning door is still open. It isn't — pinning already happened,
so any future ceremony is a *rotation*, which is the case the doc's own warning
about mismatched-signature hard-rejection applies to most sharply.
Marked complete and kept the procedure verbatim below, since it's exactly what a
key rotation or publisher change needs. Also dropped a stale `:21` line number
from the anchor.rs citation.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The inverse of the usual drift — these understate rather than overstate, which
is just as misleading for someone deciding what is safe to change.
**dht-distribution-design.md** was headed "Status: Design (no code yet)".
`core/archipelago/src/swarm/` has five modules plus `content_hash.rs`.
**phase4-streaming-ecash-plan.md** was headed "not implemented". `swarm/paid.rs`
states in its own header that it implements "DHT distribution plan, Phase 4 step
F", and there is a `streaming::` module behind five `streaming.*` RPCs
(list-services, configure-service, toggle-service, pay, prepare-payment).
Neither is reachable in a stock build, which is presumably why the headers were
never updated — and that is the part worth documenting rather than eliding. Both
now state the gates: the `iroh-swarm` cargo feature is off by default (iroh and
iroh-blobs are optional deps pulled in only by it), `config.swarm_enabled` is off
by default, and paid serving stays free for everyone until the operator enables
the `content-download` streaming service.
Checked the other plan-only docs for the same error; these two were the only
ones. `nostr-identity-import-plan.md`, `nostr-signer-login-research.md` and
`hardware-signer-design.md` correctly say no code exists — verified: no identity
import or NIP-07 login RPC, and no TROPIC01 reference anywhere in core.
`dual-ecash-design.md`'s "in progress" is right too — the `wallet.fedimint-*`
RPCs exist, no Cashu ones do.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The 2026-07-23 status block lists as DONE: "Deleted: Courier_New/, Benton_Sans/,
Redacted/ fonts; wireguard.apk; atob.s9pk; obsolete test-install.sh (all
git-rm'd)". All seven are still tracked at HEAD and present on disk. Only the
web/dist copies went; the sources never did.
git ls-tree -r HEAD --name-only | grep -iE 'Courier_New|Benton_Sans|Redacted/|wireguard.apk|atob.s9pk'
That means a repo about to be published under MIT still carries a commercial
Font Bureau typeface and two proprietary Monotype fonts — precisely what §3 of
this audit says must not ship. An audit that reports a blocker as closed is
worse than one that never checked, so the entry is now struck through with the
file list and the verification command inline.
Deleting them is safe and I checked before saying so: nothing references the
font *files* (the three `Courier New` hits are CSS font-family fallbacks naming
the system font, not @font-face sources), and wireguard.apk / atob.s9pk have
zero references anywhere in the tree. Left the deletion itself to the operator —
it is 40 MB of tracked binaries and outside a docs pass. Removing them also
takes 40 MB off the frontend OTA tarball, which is a separate open item.
Also re-verified the rest of the remaining list:
- `zbase32` (LGPL-3.0+) is still a direct dep (Cargo.toml:113, did_dht.rs:40,49).
Still the only hard copyleft blocker.
- LICENSE (MIT), NOTICE and both THIRD-PARTY-LICENSES inventories are present —
so the headline "no license of its own" is closed; softened the verdict to say
which blockers remain rather than leaving a stale "not releasable as-is".
- The four StartOS-derived crates still exist; flagged that KEY-05 cites
core/models, so that one needs review rather than a blind delete.
- Item 6 (git filter-repo history purge) is superseded — the launch plan is a
fresh-history publish, so there is no history to rewrite.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
**ADR-009** lists six "non-negotiable" mandatory security defaults. Checked each
against `core/container/src/manifest.rs` and `core/security/src/`:
- `seccomp_profile: Default` — the string `seccomp` appears **nowhere in
`core/`**. Not as code, not as a TODO. This constraint is entirely fictional.
- AppArmor — `container_policies.rs` generates and `apparmor_parser -r`s a
profile, but its own comment reads `TODO: Configure Podman to use the
profile`. `security.apparmor_profile` parses into a manifest field that
nothing ever reads.
- `user` UID > 1000 — no UID validation exists in the runtime parser at all.
- `image_tag` pinned — preflight script only; the parser accepts `:latest`.
- `readonly_root` / `no_new_privileges` — safe defaults when omitted, but
`validate_security()` never rejects an explicit `false`, so the ADR's
"Reject manifests that violate mandatory defaults" step does not exist.
Genuinely enforced: the capability allow-list and bind-mount confinement (the
latter stronger than the ADR describes). Added an Implementation status section
saying so per-row. The decision stands; the claim of enforcement did not, and on
a security ADR that gap is the whole point of writing it down.
**ADR-004** said Tor carries *all* inter-node communication and runs as the
`archy-tor` container. Neither holds: transport priority is mesh → LAN → FIPS →
Tor (`TransportKind` 1-4, Tor as last fallback, largely because of the latency
this ADR itself lists), and Tor is the host Debian service driven by
`archipelago-tor-helper` — `container-doctor.sh` actively removes an `archy-tor`
container if it finds one, and no `apps/tor` manifest exists. Added an amendment
rather than rewriting the record. Worth flagging that both changes landed
without their own ADR.
All 10 ADRs are Status: Accepted; 001-003, 005-008 and 011 verified consistent.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Header claimed the whole 2026-04-22 plan "has been implemented". The
architecture was adopted, but checking each item against the tree:
- The `core/archipelago/src/reconcile/` module the doc lays out in detail —
desired.rs / current.rs / diff.rs / apply.rs / derived.rs / backoff.rs — was
never created. The reconciler shipped as container/boot_reconciler.rs +
container/prod_orchestrator.rs instead.
- FM2's named fix `reconcile::derived::render_bitcoin_conf` does not exist. The
drift was eliminated a different way: bitcoind runs with an explicit `-conf`
derived from secrets each start, and stale datadir configs are removed.
- FM1/FM3 are partial — companion UIs are Quadlet units, main app containers are
not, since `use_quadlet_backends` still defaults false. The "v1.7.48+ full
reconcile module / main containers become Quadlet units" step has not happened.
- **FM6 was never implemented.** There is no podman corrupt-state probe and no
`system renumber` recovery anywhere in the tree. The 2026-04 failure that made
a registry node unreachable would still require manual SSH today — which is
precisely the "zero-manual-intervention" target this doc opens with.
FM4 and FM5 did ship as described.
Replaced the blanket claim with a per-item table so the doc stays useful as
incident history without reading as a description of the code, and noted that
the unit path throughout says /etc/containers/systemd/ while units are actually
written to ~/.config/containers/systemd/ (the path is rootless).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
`test_node_key_known_answer_vs_python_verifier` pinned the node Ed25519 and node
Nostr keys, and `test_release_root_known_answer` covers the release root. The
remaining six — FIPS mesh transport, identity Ed25519, identity Nostr (NIP-06),
Bitcoin BIP-84 and LND aezeed entropy — were only asserted to be mutually
distinct by `test_full_derivation_from_known_mnemonic`.
Distinctness is satisfied by ANY change to an HKDF info string or BIP-32 path.
So redefining `archipelago/lnd/entropy/v1` — the seed behind a user's Lightning
wallet — broke no test, while invalidating every backup verification a user had
already performed against docs/SEED-VERIFICATION.md. Same for the FIPS key that
authenticates a node on the mesh.
Expected values were produced independently by the Python verifier published in
that doc, whose primitives were themselves cross-checked against bip_utils and
cryptography's own HKDF (BIP-39 seed, both BIP-32 paths, x-only pubkey, bech32
and HKDF-SHA256 salt=None all matched byte for byte). This commit closes the
loop in the other direction: the Rust implementation now agrees with those same
bytes, so the doc and the code are pinned to each other.
Verified: 26/26 seed tests pass.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
**container-lifecycle.md** told operators to read the reconciler's decisions with
`journalctl --user -u archipelago`. That returns nothing: `archipelago.service`
is a SYSTEM unit (`WantedBy=multi-user.target`) that merely runs as
`User=archipelago`. It's `sudo journalctl -u archipelago`. Easy to get wrong
because the companion Quadlet units next door genuinely are `--user`, so both
forms appear in the docs and only one is right per unit — spelled that out
inline. Swept the rest of docs/: no other instance.
**quadlet-compilation.md** — added the `Network=host` case. Podman rejects
`PublishPort` with host networking (crash-loop, exit 125), so the renderer drops
declared ports rather than emitting them
(`render_host_network_omits_publish_ports`). A developer reading the directive
list would otherwise expect a mapping that never appears.
Everything else in both docs verified against quadlet.rs / prod_orchestrator.rs /
boot_reconciler.rs: the unit dir, the DO-NOT-EDIT header, Pull=never,
DropCapability=ALL, Secret=…,type=env, TimeoutStartSec=0, RestartSec=10,
WantedBy=default.target, the render/write_if_changed/enable_now/disable_remove
four-step, uid 1000, adopt_existing, the user-stopped.json / user-uninstalled.json
desired-state gates, and the 30s tick.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The reconciler's pre-start hooks for the btcpay stack, fedimint and fmcd
chowned unconditionally on EVERY prepare — and prepare re-runs far more
often than install (every reconcile that touches the app). archi-dev-box's
journal showed the same three dirs re-chowned every ~15s. The hooks exist
to repair old installs; they now skip when ownership is already correct
(root stat probe — the daemon's rootless metadata read can't see the
subuid-owned dirs).
Co-Authored-By: Claude <noreply@anthropic.com>
Ran the doc's script rather than only reading it, and cross-checked every
primitive it implements against independent libraries (bip_utils for BIP-39
seed / BIP-32 derivation / bech32, and cryptography's own HKDF): BIP-39 seed,
m/44'/1237'/0'/0/0, m/84'/0'/0', x-only pubkey, npub encoding and
HKDF-SHA256(salt=None) all match byte for byte. The hand-rolled crypto in this
doc is correct.
Two real gaps fixed:
- **The FIPS mesh transport key was missing.** `seed.rs:227` derives it from the
same master seed via `archipelago/fips/secp256k1/v1`, and a user verifying
their backup had no way to check it — despite it being the key that
authenticates them on the mesh. Added it to the diagram and as section 2b of
the script (same shape as the node Nostr key; verified against
`derive_fips_key` and `hkdf_derive` using `Hkdf::new(None, ikm)`).
- **The "compare with your node" commands were wrong.** The RPC endpoint is
`/rpc/v1`, not `/api/rpc`, and `identity.get-node` is not a method — the real
ones are `node.did` and `node.nostr-pubkey`. Also dropped "UI: Settings >
Identity", which is not a screen that exists, in favour of the two
identity files on disk.
Verified and left alone: all five other HKDF info strings, both BIP-32 paths,
and the `node_key.pub` filename. The release-root key
(`archipelago/release/root/ed25519/v1`) is deliberately still absent — it is
derived from the project's signing seed, not a user's node seed.
Noted separately: `system.get-node-key` sits in the CSRF-exempt list
(`api/rpc/mod.rs:337`) but has no dispatcher arm, so it is an exemption for a
method that does not exist. Harmless, but it should be removed.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This doc is marked "Status: implemented ... shipped end-to-end" and then
describes a cryptographic verification chain that does not exist. On a repo
about to go public, that is the single worst kind of doc bug: it promises a
security property.
`signatures.manifest_hash` / `signatures.did_signature` appear exactly once in
the codebase — as two struct fields at `marketplace.rs:106-107`. Nothing reads
them. There is no hash comparison, no DID resolution, no signature check. The
authenticity actually delivered is the Nostr event's own NIP-01 Schnorr
signature, which proves the publishing key sent the event but says nothing about
the DID the manifest names.
Added a warning at the top, marked the "Manifest Signing (DID Layer)" section
and steps 3-6 of the verification flow as not implemented, and annotated steps
7-8 as advisory (validate_manifest returns scoring issues; it does not block
discovery or install).
The trust model was overstated in the same direction:
- "DID Verification | 30 | Manifest is signed by a valid DID key" is a
`did.starts_with("did:")` string test. Any publisher can claim any DID and
take the 30 points.
- "Relay Consensus | 20" is graduated and never zero (1 relay still scores 5).
- "Version History | 15 | multiple published versions (shows maintenance)" —
nothing counts versions; it's 10 for a 3-part semver plus 5 for a non-empty
repo_url.
Worked the arithmetic through: an unsigned manifest with a plausible DID string
and a pinned image scores 65, landing in the "Community" tier. Said so.
Other corrections:
- `marketplace.unpublish` is documented but was never implemented (the string
appears nowhere); removed it and noted why NIP-33 makes it non-trivial. Added
the two payment methods that do exist (`create-invoice`, `check-payment`).
- The schema section said marketplace manifests "follow the existing
apps/{app-id}/manifest.yml schema", contradicting the header three paragraphs
above. They are separate types.
- The security-enforcement list claimed a capability allow-list, a
host-networking ban and system-path mount restrictions. Those rules are real
but live in the runtime manifest parser for a different schema — marketplace
validation checks four things and gates none of them.
- `run_as_user` documented as "> 1000" in two places while the code checks
`>= 1000` and the doc's own example uses 1000.
- Data-storage tree listed `cache/trust-scores.json` and `config.json`; neither
is ever written.
- The 15-minute cache TTL and 30-minute background refresh don't exist —
discovery is RPC-triggered and the cache has no expiry.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The index covered users, architecture and app development but had no entry
point for "I want to work on Archipelago itself" — so eight tracked docs were
reachable only by guessing filenames, including the two that matter most to a
newcomer: developer-guide.md (how to build the workspace, the frontend and an
ISO) and LICENSE-COMPLIANCE-AUDIT.md (dependency licensing, which is exactly
what a reader checks first on an open-source repo).
Added a "Contributing to Archipelago itself" section covering those plus
bulletproof-containers, the signing runbook, the 1.8.0 hardening plan and
CLAUDE.md; filed pine-voice-commands under Getting started and demo-build-info
under contributing.
Also noted that ADR-010 was never issued — verified across all history, so the
009 → 011 gap is not a missing file — and added the two archived session logs
(HANDOVER-2026-07-02, SESSION-1.8.0-OTA-PROGRESS) to the archive table, which
already claimed to cover completed session logs but listed none.
Link check re-run across docs/: 0 broken. Only RELEASE_NOTES_BACKLOG.md is now
deliberately unindexed (internal working list).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The header says Phases 1-3 shipped, then §1 "Where we are today" described the
world before any of them: catalog carrying "version + image override only", the
manifest "never registry-distributed", counts of 48 disk manifests and 28
catalog entries. A reader hits the contradiction immediately and can't tell
which half is current.
Retitled §1 as the pre-Phase-1 baseline it is, and added the actual state:
`releases/app-catalog.json` has 66 entries and 56 embed a full `manifest` block
— one for every `apps/*/manifest.yml` in the tree (the stale counts were 48 and
28). What's genuinely left is Phase 4 (build-context apps) and Phase 5 (drop
`apps/` from the OTA rsync), which the phase list already marks ⏳.
Also:
- The install arrow claimed "render Quadlet unit"; same overstatement corrected
in architecture.md and app-manifest-spec.md — Quadlet is opt-in, the default
is podman create+start.
- §8's open question "generated_files with inline content — already supported?"
is answered: `app.files[]` takes inline `content` with placeholder rendering.
Marked answered rather than leaving a resolved question looking open.
Verified present and unchanged: `catalog_manifest_to_overlay`,
`install_stack_via_orchestrator`, `install_immich_stack`, and the catalog-wins
merge semantics.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
**COMMANDS.md** named six CSRF-exempt read-only methods, two of which
(`bitcoin.getinfo`, `monitoring.current`) are not exempt — a client trusting the
doc would send them with the cookie alone and get rejected. The real set is
twelve (`api/rpc/mod.rs:326-340`); listed all of them and said plainly that
everything else needs the header. The rest of the doc verified clean: the 480 /
200 / 160-char caps, the four `assistant_*` config keys, both default model ids,
`is_sender_allowed`, `strip_archy_trigger` / `run_node_cmd`, the three
unauthenticated HTTP endpoints, and `auth.login.totp` all match the code.
**secrets.md** said `secret_env` "sets `<key>` in the container's environment",
which reads as a plain `-e KEY=value` and undersells the design. It isn't:
resolved pairs are registered as podman secrets named
`archy-env-<app-id>-<key>` and referenced by name, precisely so the value stays
out of `podman inspect` and out of plaintext `Environment=` lines in Quadlet
units. Also documented the interpolation-taint rule — a plain `environment`
entry that expands `${SECRET}` (BTCPay's connection strings) is itself treated
as secret-bearing rather than left in the clear, which is what makes it safe to
build connection strings from secrets.
Everything else in secrets.md verified against `container/secrets.rs`: the four
kinds and their file shapes, the bare-filename rule, the every-tick idempotent
`ensure_generated_secrets`, and the atomic 0600 temp-fsync-rename writer.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Narrative pass over the manifest spec, plus one correction to the guide I
committed in ba052736.
- **`app.id` allowed `_`.** `is_valid_app_id` accepts lowercase ASCII letters,
digits and single hyphens only — no underscores, no leading/trailing hyphen,
no `--`. The spec's "alphanumeric + `-`/`_`" would have a developer write an id
that fails to parse.
- **"Must match the directory name"** is a convention, not a rule. The loader
(`prod_orchestrator.rs:1455-1474`) walks `*/manifest.yml` and keys off
`app.id`, never comparing it to the folder, so a mismatch silently registers
the app under a different id. Said so rather than implying enforcement.
- **The Quadlet claim was the same one architecture.md was corrected for**
(f55ed6bf): install does NOT compile to a `user.slice` Quadlet unit today.
`config.use_quadlet_backends` defaults false, so apps take the legacy
`podman create + start` path; Quadlet is opt-in per node and companion UIs are
the exception that already use it.
- **"no per-app installer code"** — true of installers, but
`run_pre_start_hooks` is a hardcoded `match app_id` covering seven first-party
apps (bitcoin-ui, filebrowser, lnd, archy-nbxplorer, btcpay-server,
fedimint-clientd, grafana). Documented as the caveat it is; anyone reading the
source will find it in a minute and the doc should not look like it's hiding it.
- `derived_env` now names the full closed allow-list including `{{BITCOIN_HOST}}`
and what it resolves to.
Correction to ba052736: I wrote there that an unknown `derived_env` placeholder
passes through verbatim. It doesn't — `validate_derived_template` rejects both
unknown names and unbalanced `{{`. Fixed that row in the guide.
Verified accurate and left alone: the capability allow-list, network_policy
values, `/dev/*` device rule, volume option allow-list, bind-source confinement,
the four generated_secret kinds, `hooks.pre_start` being schema-only, and the
30s reconciler interval (`BootReconciler::DEFAULT_INTERVAL`).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Narrative pass. The Security Requirements section described a blocklist where
the code enforces an allow-list, and attributed enforcement to the wrong layer:
- **"Forbidden: mounting system paths /, /etc, /var, /usr, /proc, /sys"** — the
real rule (`manifest.rs:1290-1313`) is the inverse: `volumes[].source` must be
absolute and under `/var/lib/archipelago/`, or a plain named volume, or one of
two reviewed exceptions (`/run/user/1000/podman/podman.sock`, `/var/run/dbus`).
Anything else is a parse error. The old wording also listed `/var` as
forbidden while every app in the repo binds `/var/lib/archipelago/<id>` — a
developer reading it would not know where their own data goes.
- **"enforced by the marketplace/catalog pipeline and the node"** — split by
layer instead. The capability allow-list is parser-enforced (verified against
the 9 entries at `manifest.rs:1089-1099`); `:latest` is NOT — only
`validate-app-manifest.sh` checks it, and a `:latest` manifest still installs.
readonly_root / no_new_privileges / network_policy=isolated are parser
defaults, so omitting them is safe rather than dangerous.
Also:
- `derived_env` documented `HOST_IP`/`HOST_MDNS`/`DISK_GB` "such as"; the set is
closed and includes a fourth, `{{BITCOIN_HOST}}`. Noted that unknown
placeholders pass through verbatim rather than erroring, so a typo silently
ships `{{FOO}}` into the container.
- The networking example hardcoded `bitcoin-knots`; `{{BITCOIN_HOST}}` resolves
to knots or core depending on what's installed.
- Documented the `files[].content` placeholder set, which is a different set
from derived_env and wasn't mentioned at all — notably `{{NETWORK_GATEWAY}}`
(the nginx `resolver` fix for post-restart 502s) and `{{secret:NAME}}`.
- The "check the UI" URL `/app/my-app/` is not a route; it's
`/dashboard/apps/:id` (detail) or `/dashboard/app-session/:appId` (embed).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The first drift-gate attempt called plain sudo stat, which the daemon's
privilege path doesn't answer — the probe silently failed and the chown
loop continued. host_sudo_output mirrors host_sudo (systemd-run --pipe)
but returns the process output, so the ownership check gets a real answer.
Co-Authored-By: Claude <noreply@anthropic.com>
The installer's completion screen and the login-console banner both told the
operator "Web Login password123". No release build accepts that password: no
default account is ever created (`main.rs:356-362`), and the `password123`
pre-setup path is `#[cfg(debug_assertions)]` + `dev_mode`
(`api/rpc/auth.rs:36-46`). A new user following the screen gets
"User not set up. Please complete setup first." on their first-ever
interaction with the product.
Both screens now say the web UI asks you to create a password on first visit,
which is what `Login.vue` actually does when `auth.isSetup` returns false. The
SSH line is unchanged — `archipelago`/`archipelago` really does still ship
(`install-to-disk.sh:205`), and killing that is the open half of the
"kill default credentials" hardening item.
Note on the path: `image-recipe/build-debian-iso.sh` is a thin wrapper that
copies `_archived/build-auto-installer-iso.sh` and rewrites its relative
paths, so despite the directory name the archived builder is the live one.
Same string fixed in scripts/install-tui-demo.sh, which mirrors the screen.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Narrative pass over troubleshooting.md against the code. Seven claims were
wrong, several of them actively misleading:
- **Tor is not a container.** §15/§16 told operators to run
`podman ps --filter name=tor` / `podman restart tor` and to read
`/var/lib/archipelago/tor/hidden_service/hostname`. Tor is the host's Debian
package running as `debian-tor`; Archipelago drives it by staging a torrc and
poking `archipelago-tor-helper` (`scripts/tor-helper.sh`, which does
`systemctl restart tor`). The hidden-service dir is
`hidden_service_archipelago` (suffixed), it's root-owned 0700, and the file a
normal user can actually read is the synced copy at
`/var/lib/archipelago/tor-hostnames/<service>`.
- **The USB installer has no "Repair" mode.** Cited three times as the recovery
path. The boot menu has exactly three entries: Install, Install (verbose),
Boot from local disk. Replaced with what those entries can actually do, plus
the fact that the installer prompts for a disk and requires typing `yes`, so
booting it isn't itself destructive.
- **`bitcoin-cli -datadir=/data`** — the container's datadir is
`/home/bitcoin/.bitcoin` and RPC creds are in a generated `/tmp/rpc.conf`;
the documented command could not have authenticated.
- **"edit bitcoin.conf to add addnode="** — the entrypoint passes an explicit
`-conf` and logs "ignoring legacy datadir bitcoin.conf". Flags come from the
manifest (and the signed catalog entry that overrides it).
- **"Bitcoin requires 600GB+"** — only above the manifest's 1000 GB threshold;
below it the node runs pruned at `-prune=550`.
- **`sudo systemctl restart podman`** — apps run under rootless Podman as the
`archipelago` user, so that restarts an unrelated root socket.
- **"Settings > Network"** — DNS config and disk cleanup are both on the Server
page (`/server`), not Settings.
Also: header claimed "the 20 most common issues" over 21 sections, and §16
presented Tor as required for peering when it's the last fallback after
mesh → LAN → FIPS.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The walkthrough told new users to log in with `password123` and said they'd
be "prompted to change this password immediately". Neither is true on a
release build:
- `AuthManager::ensure_default_user` is never called. `main.rs:356-362`
says so explicitly ("Don't auto-create default user — let onboarding flow
handle password setup via auth.setup"), and the function is `#[allow(dead_code)]`.
- The only `password123` login path is `api/rpc/auth.rs:36-46`, which is
`#[cfg(debug_assertions)]` AND `dev_mode` AND only fires *before* setup —
no release binary carries it.
- `Login.vue` calls `auth.isSetup` on mount and renders the "Set Up Your
Node" password-creation form when it returns false. That is the real
first-boot screen, and it is the only `auth.setup` caller in the frontend.
So there is nothing to be "prompted to change" — the user creates the
password themselves, and the doc's version taught them to look for a
default that does not exist.
Fixed in four places:
- user-walkthrough Step 8 rewritten as "Create Your Password"
- troubleshooting's "Default password is password123" solution replaced,
including the warning that deleting user.json does NOT recover a lost
password (the onboarding gate refuses auth.setup on a provisioned node)
- api-reference cURL example uses a placeholder, not the fake default
- 1.8.0 hardening plan's "kill default credentials" item now reflects that
the web half is done and only the SSH defaults still ship
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The direct metadata read can be denied in the service's rootless context even
when the directory is already correctly owned, which kept the reconciler
calling sudo chown on the same Postgres volume every minute. A root
fallback gives the guard a reliable answer on deployed nodes while remaining
much cheaper than a recursive chown.
Co-Authored-By: Claude <noreply@anthropic.com>
Verified the project-structure tree against the tree. Two stale entries:
- image-recipe/build-auto-installer-iso.sh was the old builder, now under
_archived/; the current builder is image-recipe/build-debian-iso.sh (the
release workflow drives it via scripts/build-iso-release.sh). Repointed.
- CLAUDE.md was labelled "AI development instructions"; it is now the sanitized
public contributor guide. Relabelled.
Everything else verified: run-tests.sh, first-boot-containers.sh, container.rs,
vpn.rs all exist; the add-an-endpoint / add-a-Vue-page tutorials match the
current dispatch pattern.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The overview stated apps install as user.slice Quadlet units. Verified against
prod_orchestrator.rs: use_quadlet_backends defaults to false, so regular apps
install via the raw podman path today; the companion UI containers are the ones
that run as Quadlet units (companion.rs owns them), and the Quadlet flip to
default for all apps is opt-in/held.
Reworded both places (the layer diagram and the App Platform section) to match
reality and the container-lifecycle / quadlet-compilation dev docs: the
orchestrator owns and self-heals app containers; companion UIs run as Quadlet
units, the validated path being flipped to default. Everything else in the doc
verified accurate — crate table, module map, data paths, security model, and
the note that the four orphan crates still exist.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
W1.7's regression gate: after 8329b826's tree-shake fix, this makes the
mock-quarantine load-bearing — a future change that reintroduces the mock
modules into the production graph fails the build instead of shipping
silently. The demo-site build (VITE_DEMO_CONTENT=true) is exempt by design.
Co-Authored-By: Claude <noreply@anthropic.com>
get_data_dirs_for_app had no btcpay arm — the default mapped to
/var/lib/archipelago/btcpay alone, leaving postgres-btcpay (where the
ACCOUNT lives) and nbxplorer on disk. Uninstall-with-wipe then reinstalled
to the old account still enabled. The btcpay arm now covers all three dirs,
for every alias and stack-member id. The map stays deliberately hardcoded:
deletion code must never derive its targets from a manifest at uninstall
time (a bad manifest could aim the wipe at another app's data).
Co-Authored-By: Claude <noreply@anthropic.com>
BITCOIN-RPC-PROXY-EXPOSURE.md's port claims verify against code (Bitcoin RPC on
127.0.0.1:8332, the bitcoin-ui proxy on 127.0.0.1:8334). But its incident
narrative named a specific node's LAN address (192.168.63.240, five times) on a
subnet the earlier 192.168.1.x sweep did not cover. Replaced with the RFC 5737
documentation address 192.0.2.240. The incident content — the exposure, the
probes, the fix — is unchanged and remains a legitimate public security record.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Verified the security subsystem's design-doc claims against code:
- KEY-05's foundational claims are accurate: entropy::draw_key_bytes exists,
KeyGenRng is sealed with OsRng as its sole production member, MIN_GUARDED_LEN
is 12, and core/clippy.toml bans rand::random/thread_rng exactly as stated.
- But its per-site table listed every production nonce/key site as disposition
"migrate" (pending), when all of them have since been migrated to
draw_key_bytes(OsRng) — storage_crypto, credentials/store, wallet/bdhke,
mesh/x3dh — and zero rand::random/thread_rng remain in production. Added a
completion note so the doc no longer reads as pending work.
Both KEY-05 and PSBT-SIGNING-ARCHITECTURE referenced
ENTROPY-SEED-AUDIT-2026-07-31.md five times as their evidence base — a doc that
was moved to local-only, so a public reader could not follow it. Reworded all
five to state the audit's findings inline ("the internal entropy audit found
...") without the unresolvable path. No published doc references it now. The
link-checker missed these because they were inline code, not markdown links.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
apply_data_uid ran a recursive sudo chown on every prepare_for_start, and the
reconciler re-prepares — archi-dev-box's journal showed postgres-btcpay rechowned
every ~45s despite already-correct ownership, and on framework-pt the same loop
surfaced as operator-visible 'chown failed' noise. chown_for_rootless_container
now stats the target first and returns early when the top-level owner already
matches the host-mapped uid:gid. Deep drift in a running container is still
caught by ensure_running_container_ownership's in-container write-probe, which
is the authority that actually matters (it probes writability, not stat bits).
Co-Authored-By: Claude <noreply@anthropic.com>
The guide walked a developer from manifest to install but never told them how
to validate the manifest locally first — despite scripts/validate-app-manifest.sh
existing for exactly that. A developer's first signal that their manifest was
wrong would have been an install failure on a node.
Adds a "Validate Your Manifest" step at the top of Testing, pointing at the
script (recently fixed — it had been rejecting every manifest because it shelled
out to a missing ruby). Notes the strict behaviour a new submitter hits, e.g.
an unpinned :latest tag is rejected, and that the Rust parser is canonical.
Verified: the install RPC example in this guide (id + dockerImage) matches the
handler; the cargo test target crate name (archipelago-container) is correct.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Verified all 144 documented RPC methods against the dispatcher. 143 are live;
one was fabricated: `mesh.discover` (params { timeout_secs? }, returns
{ nodes: MeshNode[] }) does not exist — "mesh discovery" appears only in code
comments as a concept, never as a method. A developer calling it gets "unknown
method".
Replaced with the real peer-listing method `mesh.peers` (no params, returns
{ peers, count }), which the frontend actually uses and which was undocumented.
Also verified: every source path cited across the docs resolves (placeholders
and a correctly-recorded deletion aside), and every documented app-manifest
field exists in the schema (no fabricated fields).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Three `MEMORY → <note>` see-also references pointed at the private agent-memory
system from public docs (demo-deployment-design.md x2, registry-manifest-design.md
x1). Removed. No tracked doc references the memory system now.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
bitcoin-multi-version-design.md carried three layers of stale internal content:
an 80-line HTML-comment work-tracking block (per-phase status with "UNCOMMITTED
on the branch", node numbers, "Next action when resuming", "Decisions still
needed from user"); a rendered "Status: design (2026-06-22)" header that was
wrong — the feature shipped, all four phases, with the downgrade guard added
today; two private `MEMORY →` references; and a node-numbered scheduling note.
Now: the comment block is gone, the status reflects reality, the MEMORY
references and node numbers are removed, and "verify on a real node" replaced
the specific fleet addresses. The design content (source-of-truth decision,
phase designs, invariants) is unchanged.
Separately, bitcoin-version-bulletproof-rollout.md was an inter-agent rollout
handoff — node numbers, branch coordination, "the other agent owns" — not a
design or reference doc. Moved to local-only (still on disk, gitignored) like
the other handoffs; its two path references (a plan doc and a script comment)
are generalized.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This 2026-04 plan has been implemented, but it still read as an active plan
("implementation started"), linked private agent-memory paths, and ended with a
stale "To resume" work block naming fleet nodes, dated fleet state, and the next
file to edit.
- Header now marks it a historical design record and points at
container-lifecycle.md for the current behaviour.
- Removed the two private ~/.claude/.../memory/ references from the header and
the entire "To resume" section (private paths, node numbers, 2026-04-22 fleet
snapshot — none of it belongs in a public design doc).
- Genericized the one remaining node-number reference in the incident narrative.
The valuable content — the six failure modes and the reconciler reasoning that
answered them — is kept intact.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The open-source plan flagged three references as "the real gaps for app
developers", and the docs index named them as not-yet-written. Written now,
each from the code rather than stubbed:
- secrets.md — generated_secrets/secret_env: the two halves, the four kinds
(hex16/hex32/base64/bcrypt) and which files each writes, the idempotent
self-healing 0600 materialisation, and the rules a developer must not break
(no hardcoded fallbacks, one canonical name, right encoding). From
container/secrets.rs and the manifest schema.
- quadlet-compilation.md — manifest -> .container unit: the full directive
mapping (including Secret= by reference, never value, and Pull=never), where
units land (~/.config/containers/systemd, systemctl --user), the
render/write/enable/disable lifecycle with write-if-changed, and how to
inspect one. From container/quadlet.rs, scoped accurately to the companion-UI
path it drives today.
- container-lifecycle.md — the level-triggered 30s reconciler: desired state
from user-stopped/user-uninstalled/manifest set, the operations table, the
self-heal-vs-respect-a-deliberate-stop rule, and migrations-never-destroy-data.
From prod_orchestrator.rs and boot_reconciler.rs.
Index updated to link all three under App development and the "known gap" note
removed. Every link across the docs tree resolves.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Follow-up to 73970cf3. Two improvements:
- Corrects a stale claim — manifests are no longer "loaded from disk, goal is
the catalog". The signed catalog has been the delivery mechanism since
2026-06-23 (origin-wins over disk), so the guide states that, plus the
consequence contributors need: editing a disk manifest alone does not change
a catalog-covered app.
- Tightens the north-star paragraph.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
CLAUDE.md was the internal agent guide: a dated "gate is GREEN" status banner
naming a specific node, pointers to now-local-only planning docs
(PRODUCTION-MASTER-PLAN, UNIFIED-TASK-TRACKER, multinode-testing-plan), the
gitea-ai push account mechanics, and references to the private memory system.
Rewritten as a contributor guide that keeps everything public-worthy — the
invariants (rootless podman, declarative apps, manifest-declared secrets,
non-destructive migrations), the build/verify notes, the commit-and-push
discipline, and the production test-gate definition — and drops the status,
node numbers, push-account specifics, and memory references. Points at
docs/ROADMAP.md and docs/README.md instead of the internal trackers.
No infra identifiers or internal mechanics remain; all links resolve.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Two concrete, verifiable documentation gaps from the open-source review:
- docs/ had no index. Adds docs/README.md grouping the 60-odd published docs by
task — getting started, architecture, app development, design docs, ADRs,
security, roadmap — in the bitcoin/bitcoin doc/ style the plan called for.
Every link in it resolves (checked). The top-level README now points at it as
the front door rather than duplicating the list.
- ROADMAP.md and tests/lifecycle/TESTING.md linked docs/multinode-testing-plan.md,
which moved to local-only (it is a fleet node inventory, not published). Those
references now describe the scope split in prose instead of pointing at a file
that is not in the public tree.
The index is honest about what is missing: it names the three app-developer
docs the plan flagged as gaps (quadlet compilation, container lifecycle,
secrets materialisation) as not-yet-written, and points at the authoritative
code for each rather than pretending they exist.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The repo ships an MIT LICENSE and the README carries an MIT badge, but the
crates themselves declared no license, so `cargo metadata`, packaging and any
downstream mirror saw "license: null". Adds [workspace.package] license = "MIT"
and inherits it in all five members via license.workspace = true. Verified with
cargo metadata: all five now report MIT.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Today's outage window came from ordering, and the ordering was baked into the
publish script itself: it pushed main — the branch nodes read the manifest
from — together with the tag, up front, then uploaded and verified assets
afterward. So the manifest advertised the new version for the entire
upload+verify window. When an upload failed inside that window, every polling
node briefly saw a v1.7.126-alpha update whose binary 500'd and whose tarball
did not yet exist.
Reordered so the manifest goes live last:
1. push the TAG only (the Gitea release and asset URLs hang off it; the tag
alone changes nothing for nodes)
2. upload assets
3. verify every asset downloads in full and matches the manifest sha256/size
4. only then push main — the step that actually triggers nodes
Also fixes a way a bad asset could slip through unnoticed: the inline
verification ran in a `while read` pipe subshell, where its `fail` (exit 1)
terminated only the subshell and let the script continue to "published and
verified". Verification now runs in the main shell via a new
check-release-assets.sh, which fails hard on the first bad asset. The same
script is the reusable by-hand verifier used to recover today's release
(both assets confirmed 200 + sha256-match before the manifest was re-published).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
13-15's two artifacts. The map names plan/wave/threat-ref per row with
today's measured results (Rust assistant suite 130/130; adapter 37; broker
25; toolConfirm, audioPlayer 11, appsConfig 13; AIUI 353/356 with the three
documented pre-existing fixture failures). Four close-out rows added
(S-invariants, evals, egress, mock-free-bundle grep). Manual-only table
discharged except the physical-handset pass, which 13-UAT records as owed
(the AIUI-06 flagged assumption wants both, and only devtools-mobile is on
record). Open questions 1/2/4 resolved with plan cites; Routstr stays
honestly open.
Co-Authored-By: Claude <noreply@anthropic.com>
Restores the signed v1.7.126-alpha manifest to main now that both artifacts are
confirmed fetchable end-to-end:
- archipelago HTTP 200, sha256 matches the manifest
- frontend tarball HTTP 200, sha256 matches the manifest
The earlier publish was rolled back (e346e552) because the manifest went live
before its assets resolved. Two separate asset faults, now fixed: the binary's
first upload landed corrupt server-side and returned 500 on download (deleted
and re-uploaded, clean); the tarball's first upload returned an empty response
and never attached (re-uploaded, 201, full 210 MB).
This is byte-for-byte the manifest the tag already carries, so its signature is
unchanged and re-verified against the pinned release root.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The v1.7.126-alpha manifest went live on main — which is where nodes read it
from — before its artifacts were reachable. The binary returns HTTP 500 and the
frontend tarball never uploaded (404), so any node polling would advertise an
update it cannot fetch.
Restores the previously published, still-validly-signed .125 manifest
byte-for-byte from 19487670, so nodes see the last release that actually
resolves. The v1.7.126-alpha tag and its signed manifest are unchanged in git
history; only what main serves is rolled back.
Publishing order was the mistake: the manifest is the trigger, so assets must
be verified downloadable before it lands on main, not after.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Signed release manifest for v1.7.126-alpha, verified against the pinned
release root before committing.
Committed by hand rather than by re-running create-release.sh: the script
regenerates the manifest at step 6, which would overwrite the signature
applied at step 6b. Its own "sign it, then re-run this script" advice destroys
the thing it just asked for.
Version bump, changelog and What's New landed earlier in 1cd068e4 — the
frontend build embeds the version via the curated What's New list, so those
must exist before the build step rather than after it.
Release gate: 8 of 9 stages passed in-run. cargo-test-weekly hit its 1500s
ceiling (exit 124, a timeout not a failure) because the non-incremental
all-targets compile does not fit on this machine; the suites were then run
separately and passed 100/100, including the downgrade guard and 30 update::
tests over the OTA apply/rollback path. Right-sizing that ceiling is a
follow-up — an override that becomes routine stops being a gate.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Live on archi-dev-box: the node still proxied web-search straight to
SearXNG :8888 unauthenticated — the repo conf was fixed in d0c9ea6e but
existing nodes' /etc/nginx/sites-enabled never gets rewritten by a source
edit. Added to the nginx self-heal battery: stale 8888/search proxy_pass →
session-gated 5678 with the Cookie forwarded (heal_stale_web_search_block,
pure + idempotent + tested). Fresh ISOs already ship the gated block.
Co-Authored-By: Claude <noreply@anthropic.com>
Mesh AssistQuery answered with a bare LLM call — no tools, no actions.
The CallerScope::Mesh variant was designed for this wiring ('the variant
exists so the shape is right when a future plan wires mesh callers into
the shared loop'); this is that plan. A trusted/allowlisted asker's prompt
now runs assistant::chat with CallerScope::Mesh { authorized } — the
operator's persisted grants cap what the model may touch (never wider),
and writes suspend on the node's own confirm gate. The reply is capped
for airtime as before, with a brevity instruction for mesh turns.
Wiring follows the blob_store pattern: RpcHandler::set_mesh_service (now
&Arc<Self>) forward-propagates an Arc<RpcHandler> into the mesh state's
new assistant_handler slot; absent (early boot) falls back to the legacy
bare-LLM answer.
Test: mesh_caller_authority_is_capped_at_operator_grants.
Co-Authored-By: Claude <noreply@anthropic.com>
Build #189 proved the IPv4 fix works (downloads all succeeded) and proved
my own edit wrong: appending inet4_only to /etc/wgetrc before wget's
package landed made dpkg's conffile prompt hit EOF, leaving wget and
debootstrap unconfigured. Moved the pin below the apt install.
Co-Authored-By: Claude <noreply@anthropic.com>
The 2912×1632 jpg painted visibly slowly over Tailscale/Tor. 1920w q82
webp is visually identical behind glass and under half the weight.
Co-Authored-By: Claude <noreply@anthropic.com>
This box (and its containers) blackhole IPv6: deb.debian.org answers AAAA
first, wget tries v6 until debootstrap's per-package timeout, and the
installer-env stage died twice today with 'Couldn't download packages'.
Probed in a debian:trixie container: v4 OK, v6 hangs. inet4_only for wget
(covers debootstrap) + Acquire::ForceIPv4 for every apt-get, including the
chroot and the rootfs Dockerfile stages.
Co-Authored-By: Claude <noreply@anthropic.com>
