Handover
A new satellite, a new delay, a new bottleneck. Last epoch's samples are lies.
Teaser epoch-typed control Starlink-class paths
The language that treats Starlink like Starlink.
Classic congestion control assumes a stable path and a stable RTT. Low Earth orbit has neither. VELA makes path epochs, mobility loss, and statistical contracts first-class, so the next controller can keep more of the bandwidth you already pay for.
The break
Your dish is not a fiber drop.
Starlink-class links hand over every 12 to 60 seconds. RTT jumps tens of milliseconds. Capacity flickers. Packets vanish because a beam moved, not because a buffer filled. CUBIC treats that as congestion and collapses. BBR keeps a stale min-RTT across the hop and sizes the window for a path that no longer exists.
Mobility loss
A burst with no queue is not a full buffer. Recovery should hold, not cut.
Lucky seeds
One orbit can flatter a controller. A win that cannot survive five is not a win.
The language
Samples die with their epoch.
VELA is not prettier Python. It is a type system and a compose kernel for LEO controllers. Reading a min-RTT after a reconfiguration is a type error. Loss is a closed taxonomy. Dual-gate claims are contracts the eval harness actually runs.
- First-class path epochs and automatic sample invalidation
- Mobility vs congestive vs unknown, with different recovery
- Interval bandwidth and handover calendars, not point estimates
- Compose lists so a new chase cannot silently double-cut the window
- Multi-seed dual-gate contracts (goodput, p95, terrestrial)
lang vela 0.1
controller Horizon {
compose Detect + SoftReprobe
+ IntervalBw + PredictiveFreeze
signals:
epoch: Epoch
rtt: Sample<ms> @ epoch
bw: Interval<bps> @ epoch
p_ho: Prob
on Reconfig(e) {
invalidate min_rtt, bw
enter Reprobe(cut: 0.58)
}
on Loss(k) match k {
Mobility => hold
Congestive => cut(0.72)
Unknown => require delay_ratio > 1.35
then cut(0.72) else hold
}
}
contract DualGate vs BBRv3approx {
assert mean(goodput) >= baseline.goodput
assert mean(p95) <= baseline.p95
}What users should feel
More of the path. Fewer dead seconds.
VELA does not add satellites. It stops wasting the ones you already have. The design target is user-visible: higher sustained speed, calmer interactive delay, fewer stalls when the sky rearranges.
- After a hop, the old RTT is gone. Controllers written in VELA cannot keep a 20 ms min-RTT across a 90 ms jump. That is how you get the path back in the first one or two round trips instead of sitting in a hole.
- Beam flicker is not a traffic jam. Typed mobility loss holds the window. You do not pay CUBIC's collapse every time the beam breathes.
- Stable sky, send. Uncertain sky, ease the queue. Interval estimates and handover calendars let a controller be aggressive on a tight epoch and gentle in the last beat before a hop. That is lower p95 lag on calls and games.
- A claimed win has to survive many orbits. Dual-gate contracts refuse a README that beats BBR on one seed and loses the mean. Honesty is how a Starlink improvement stays an improvement.
Honesty
A teaser, not a firmware flash.
VELA is a lab language. It does not flash your dish. It does not promise plus-N megabits on a particular terminal tonight. The running, measured congestion controller on this network is OrbitStack / LeoAware. VELA exists so that controller can be composed, checked, and extended for space internet without unnamed flags or lying numbers.
Optional path hints (ASCENT-D) stay fail-closed. If the hint is missing or corrupt, the endpoint controller is still defined. That is the only story that matches Starlink today.
Questions
- Will this make my Starlink faster today?
- Not as a kit update. VELA is how the next controller gets written. OrbitStack is the live science surface.
- Is the language open source?
- The teaser and the design are public here. The compiler tree is in private lab until it is ready to open.
- Does this need SpaceX cooperation?
- No. Endpoint-only is the default. Hints are optional and fail-closed.