[jsjohnst]

Fiber gets closer to the speed of light, even when accounting for retransmission delays, than just 50%. The index of refraction for glass is commonly quoted as 1.5. That makes it 33%. Also, Starlink isn’t vacuum speeds either (unless you are talking about the inter satellite laser links and not the downlink to CPE, which even they I’m not sure operate at full “speed of light in a vacuum”).

[troymc]

The refractive index of air is very close to that of vacuum (which is 1), so most Starlink transmissions between nodes (ground-satellite, and satellite-satellite) will travel at near light speed.

[jsjohnst]

> The refractive index of air is very close to that of vacuum

Yeah, 1.0003 vs 1.0. Very close, but not the same was my point.

[zamadatix]

The refractive index is only one part of the story, light does not follow a straight path through fiber it bounces around. I'd assume this to be about the newer inter-sattelite links as comparing fiber to fiber plus some seems obvious.

[wallacoloo]

> light does not follow a straight path through fiber it bounces around.

this is only relevant for multimode fiber, right? in singlemode fiber the light must propagate parallel with the fiber. even in multimode fiber, some light is parallel, so it doesn’t necessarily limit latency — it creates dispersion.

[zamadatix]

Singlemode is designed to minimized modal dispersion but it still occurs, especially over the longer distances including from stresses in the core rather than traditional bouncing. First photon isn't as important, it's the signal peak that matters as the receiver will try to decode from that. Typically this latency is hidden by needing repeaters every ~60km anyways due to both dispersion and loss. I'm not sure how far apart Starlink inter-satellite can repeat and if they can "skip" satellites as long as there is a clear direct path to another farther along the path, I haven't been in contact with their engineering folks since I changed jobs last year and am no longer a corporate customer.