Blog

Development Update — June 18

Thu, 18 Jun 2026 00:00:00 +0000

The most important day in a while, and it started with the network down. The dmsg-only convergence work of the prior days, combined with a fleet-wide redeploy, walked straight into a cold-start bootstrap trap: every dmsg-server was waiting on a registration that could only succeed once another server was already accepting — a circular dependency where nobody could be first, so the whole fleet stayed down even though every binary was healthy. Untangling it took a chain of fixes that each peeled back one layer of the trap, and once the network was back up, a multi-agent code audit of the v1.3.72–76 dmsg changes turned up a long tail of leaks, races, and “looks-healthy-but-can’t-relay” failure modes — all fixed the same day, culminating in the v1.3.76 release. This post traces the arc: the outage, the root cause, the audit, and a hard-fail defense against silent post-quantum downgrades.

Development Update — June 17

Wed, 17 Jun 2026 00:00:00 +0000

The big structural change of the day was collapsing the visor’s two same-key dmsg clients into one — and the bumpy path it took to land. The visor had been running a discovery client and a direct client under the same public key, and because dmsg servers evict by newest-session-wins, the two were quietly evicting each other on every shared server: a continuous self-eviction storm measured at tens of session-deaths per minute per server. The fix went in, got reverted when it surfaced a fatal re-entrant deadlock, and was re-applied with the deadlock fixed. Around that headline: a clutch of resolving-proxy alias features (including a clickable home.dmsg directory page), a redis-starvation fix in service discovery, a dmsg-server config simplification, and the noise-fork PQ design doc catching up to the code.

Development Update — June 16

Tue, 16 Jun 2026 00:00:00 +0000

The headline is post-quantum cryptography landing in the Noise handshake. Over three pull requests — a design sketch, a self-contained crypto layer, and the live wiring — skywire’s encrypted transports gained a hybrid X25519 + ML-KEM-768 key agreement, with no new round trips, no flag to flip, and full backward compatibility with peers that haven’t updated. Around it: an end-of-month docs refresh of the reward rules to match the current dmsg-only deployment, two transport-reliability changes aimed at dead-edge detection (OS-level keepalive and a local-route-calc preference for disjoint mux legs), and the migration off the frozen noise fork that made the PQ work possible in the first place.

Development Update — June 15

Mon, 15 Jun 2026 00:00:00 +0000

A short but consequential day, with two changes that each close a gap the old code couldn’t see. The router gains a per-leg end-to-end liveness probe, so a multiplexed route whose path dies beyond the first hop is detected and dropped instead of silently black-holing traffic. And the rewards regional-saturation formula is reworked so that running more visors at a country’s IP grows the country’s pot proportionally rather than freezing it at the single-visor amount — fixing a case where adding visors actively diluted everyone’s reward.

Development Update — June 14

Sun, 14 Jun 2026 00:00:00 +0000

A day of grounding the dmsg-only default in reality: as the network leans harder on the overlay, the failure modes that used to be masked by a plain-HTTP fallback now have to be fixed properly. Two of them were boot-and-reconnect wedges — a dmsg-only visor that spun forever trying to reach the transport-discovery, and a SUDPH client that retried a closed socket until it gave up. Alongside those, a new inject_pk mode makes forwarded websites HTTP-aware (the backend finally learns who is connecting), the router’s setup-failure diagnostics are made panic-safe and then richer, and a couple of CLI/UI papercuts get smoothed.

Development Update — June 13

Sat, 13 Jun 2026 00:00:00 +0000

A day that follows the plain-HTTP retirement to its logical conclusion: if deployment services live on the overlay, you need a clean way to see them from a browser without a clearnet hop — and that way is the resolving SOCKS5 proxy. The headline work makes those .dmsg / .skynet proxies smarter: a visor asked for its own key now short-circuits in-process instead of dialing a wasteful round-trip back to itself, and a configurable alias lets skywire.dmsg mean “the local visor’s landing page.” Around it: the CLI finally drops its dead plain-HTTP fallback (the resolver is the supported replacement), a discovery-side timeout-and-hang fix on /uptimes?v=v3, a self-healing --direct route for control-plane forwards, and the real fix for a dmsg-discovery CPU storm that had been pinning the service for hours.

Development Update — June 12

Fri, 12 Jun 2026 00:00:00 +0000

The headline is persistent pty sessions: the hypervisor web terminal can now drop, reconnect, and rebind to the same shell with its missed output replayed — built up across four self-contained phases so the concurrency-critical core was validated before anything touched the host’s hot path. Around it, a receive-side route-setup race gets fixed by parking frames instead of dropping them; the CLI stops silently serving stale cache data when you asked for fresh; minimal visor configs finally seed the direct-service entries they need to reach TPD over dmsg; the web terminal works for sub-hypervisor visors; and reward lookup and hv ls --load give operators two new windows over dmsg.

Development Update — June 11

Thu, 11 Jun 2026 00:00:00 +0000

A day spent making things recover from failure instead of silently wedging on it. skysocks-client’s reconnect loop finally fires on the failure that matters most — a route group torn down router-side without an EOF — and interactive pty terminals stop dying every two minutes of idle. Underneath, a sustained adversarial audit during a live deploy-churn window turned up a string of churn-exposed leaks, races, a crash, and a transport-eviction bug that was breaking routes for minutes at a time. Plus a --cascade config flag so operators can opt into the source-driven route setup, and a documented hook for restarting out-of-tree services after an auto-update.

Development Update — June 10

Wed, 10 Jun 2026 00:00:00 +0000

A heavy day on two fronts. The first is the dmsg server: it stops booting on plain HTTP entirely — building its transit client before the server, registering and resolving strictly over dmsg, falling back to the embedded deployment config when no block is set, and self-updating its embedded server list over dmsg — and it gains a way for a non-public server to be reachable inbound via signed peer announcements. The second is the adaptive routing arc for multiplexed proxies, which lands its full observable → manual → adaptive control loop: mux-set to reconcile a leg-set, mux-auto to prune toward a latency preset, and the GROW direction so a session re-grows lost redundancy. Around those, a long run of lifecycle-audit fixes hardened the router, transport, dmsg session, and CXO layers against leaks and races found during live deploy-churn.

Development Update — June 9

Tue, 09 Jun 2026 00:00:00 +0000

A day of sharpening the tools you measure and operate the network with. The route-finder learns to hand back per-hop latency so a client can rank routes without re-probing them; the proxy mux-info view gains a live throughput-and-share readout so you can see what a multiplexed session is actually delivering right now; and the CLI takes another step off the plain-HTTP edge of the deployment services. Underneath that, two leaks and a race get root-caused — a CXO publisher heap leak that was OOM-looping the transport-discovery in hours, and a router data race that turned out to be the real cause of the linux CI failures.

Development Update — June 8

Mon, 08 Jun 2026 00:00:00 +0000

A day with two threads running in parallel: making multihop route setup tolerate the network’s churn instead of failing on it, and giving the resolving proxy real source-routing power. The route-setup work adds a per-hop id-reservation retry on transient dmsg-202 and replaces dmsgweb’s reconnect storm with a single client. The resolver work lets you name a dmsg rendezvous server or an exact transport path right in the hostname — pinning how a .dmsg or .skynet address is reached. Around them: latency-weighted route selection becomes the finder’s default, debug logs and pprof get consolidated onto one dmsg port, and a third CXO goroutine leak is closed.

Development Update — June 7

Sun, 07 Jun 2026 00:00:00 +0000

The day the dmsg session lifecycle finally got fixed — the change that was the real root cause behind “dmsg error 202” and the per-hop coin flip that made multihop route setup unreliable for months. Newest-session-wins replaces a lifecycle that rejected healthy reconnects and clung to half-dead corpses. Alongside it: bounding the mux-bandwidth RTT probe so it can’t hang the run, and a hypervisor UI pass that adds a DMSG-server IP column, fixes text contrast, and declutters the skynet forms.

Development Update — June 6

Sat, 06 Jun 2026 00:00:00 +0000

A day spent unsticking the things that quietly leak, wedge, or fall back to plain HTTP. The headline is a dmsg-only fix for the blank hypervisor network tab — a casualty of the new default config — alongside a setup-node connection pool that stopped handing out dead connections and a CXO goroutine leak that was dragging whole hosts down. Around them: the rewards calc sourcing uptime from the transport-gated tracker, and the got CLI learning to actually use its SOCKS5 proxy.

Development Update — June 5

Fri, 05 Jun 2026 00:00:00 +0000

A big day the day after a release, and the dominant thread is multihop route setup again — but this time with a hard, pragmatic pivot. Live A/B testing showed the new source-driven cascade installs routes whose control plane succeeds on every hop yet whose data plane doesn’t carry traffic, so the default flips back to legacy direct-dial route setup while the cascade’s data-plane bug is chased down, with the dial loop hardened to walk more candidate intermediates and a mux that now self-heals its degree in the background. Around it: a benevolent bandwidth-spreading routing-policy preset, a TPD fix that keeps both edges of an expired transport (worth ~half the network’s bandwidth rewards), a launcher fix that had the skysocks- and vpn-clients dead on arrival, a synchronous bbolt integrity walk that stops ARM boards crash-looping, the ssh/sshd/sshfs commands unified under cli pty, and a clutch of CXO, skychat, and logserver ergonomics.

Development Update — June 4