plan: DEC-9/10 decided, DEC-12 TLS proposed, add WS-I TLS card, update WS-H

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
This commit is contained in:
ariandClaude Sonnet 5.5 committed 2026-10-16 06:53:26 -04:00
1 parent b5b3466e1d
commit 7b588120e0
6 files changed
+91 -27

No files matched your search

+20 -12
View File
@@ -108,27 +108,25 @@ The MCP spec requires Origin validation on Streamable HTTP servers. If the pinne
---
## DEC-9 — Port 1984 in the NixOS module, reachable from a remote server on the LAN
## DEC-9 — Port 1984 in the NixOS module, loopback by default
**Status:** Revised 2026-10-16, awaiting answers
**Status:** Decided 2026-10-16
The module defaults `port = 1984`. The binary's own default is unchanged, so non-Nix installs are unaffected.
The module defaults `port = 1984` and `bind = "127.0.0.1"`. The binary's own default is unchanged. Exposing the service to the LAN is an install-time choice made in the host's own config, never a default: the operator sets `bind`, `openFirewall` (or an interface-scoped rule) and `allowedHosts` for that deployment. Config validation already refuses auth-off on a non-loopback bind, and a non-loopback bind without TLS logs a warning (see DEC-12).
**Intended use:** another server on the LAN calls this host. So the host config sets `bind` to a LAN address (or `0.0.0.0`), the firewall opens 1984 for that traffic, and auth must be enabled (config validation already refuses auth-off on a non-loopback bind). For `/mcp`, `server.allowed_hosts` must include the hostname or IP the remote client uses, or rmcp answers 403.
**Open:** (a) restrict the firewall to the remote server's IP or subnet, which is recommended, versus the whole LAN; (b) plain HTTP versus TLS in front, since bearer tokens are readable on the wire over plain HTTP; (c) the LAN address or hostname clients use.
**Decision:**
**Decision:** Accepted by the user. The default is loopback. Installs change the relevant settings.
---
## DEC-10 — CLI wrapper for token commands
## DEC-10 — No CLI wrapper
**Status:** Proposed
**Status:** Decided 2026-10-16
The DB is in a 0700 state dir owned by the service user, so a login user cannot run `slinuxm token …` directly. The module installs `slinuxm` on `PATH` as a wrapper that re-execs `token` subcommands via `sudo -u <service user>` and passes everything else through. Alternative: ship the bare binary and document `sudo -u`.
The module installs the plain `slinuxm` binary on `PATH`. Tokens are minted with `sudo -u skyg-linux-tracker slinuxm token create <name>`, documented in the README and in the module's post-install notes. No wrapper script.
**Decision:**
Running the token commands as root is a trap: if the service has not started yet, root creates the state dir and DB owned by root, and root-owned WAL files can lock the service out. So the token subcommands refuse to run as root when the state dir or DB belongs to another user, and print the exact `sudo -u …` command to use. This is a small change in `slinuxm`, tracked as H5.
**Decision:** Accepted by the user. No wrapper. The app guards against the root mistake.
---
@@ -139,3 +137,13 @@ The DB is in a 0700 state dir owned by the service user, so a login user cannot
Tokens are minted with `slinuxm token create <name>`, printed once, and never stored in plaintext. There is no `bootstrapTokens` option and no import of pre-hashed tokens in v1.
**Decision:** Accepted by the user: print the token once, never store it.
---
## DEC-12 — TLS terminated in-process with rustls
**Status:** Proposed
Bearer tokens over plain HTTP are readable on the wire, and LAN exposure is the intended deployment. The service gets native TLS: `server.tls.enabled`, `cert_file`, `key_file`, with certificate reload on SIGHUP so ACME or sops-managed rotation needs no restart. A reverse proxy still works and nothing depends on it. rustls is already in the dependency tree and openssl stays out. When TLS is enabled the MCP Origin allowlist uses `https://`. A non-loopback bind without TLS logs a startup warning but is not refused.
**Decision:**