Comment by xt00 | Hacker News Reader

Comment by xt00 | Hacker News Reader

FYI: it takes 1.9ms for light to travel 570km. Given a situation where people are just streaming Netflix, then the latency should not be an issue, but if you are browsing a webpage that hits 50 domains downloading random javascript stuff, it will result in a summation of a bunch of 1.9ms*2 events occurring. Plus, if people are at an angle to the satellite, or your data hops between satellites, then you will be seeing additional latency. It would seem like the optimal approach to web browsing for example would be similar to the Amazon Silk browser, where you communicate your "intent" of what you want to do like which webpage you want. Then it renders that webpage in a local instance on the cloud, then transmits the fully rendered webpage to you in a streaming sense. That process would likely cause additional latency before you start to see something show up, but the result is actually a shorter total time to get your page in front of you.

This is literally what happens all the time for all Internet users though wires and fiber on ground-based internet at similar or worse single-hop latencies.

The number of domains is irrelevant since most of those requests would be asynchronous. What matters is the number of hops that the traffic needs to take and the latency for each hop.

1.9ms times any reasonable number of hops would still produce a sub-100ms latency which is the FCC sites as their requirement, and would be a perfectly serviceable latency for nearly all Internet use (and is around what most users experience now).

It is unclear why the FCC would reference "the laws of physics" as a barrier for StarLink achieving this, as it is unclear how they would be causing any problems.

SpaceX does not have cross links. They will need to have several hops up and back to different satellites before the signal arrives in many cases. It's not unlikely that when processing time is included, it's well over 100ms.

If conventional geosynchronous satellite internet can achieve real world latencies between 600ms and 1200ms, can you describe a scenario where having a hop time that is 60x faster could not manage to improve this by a factor of 12x?

Think of it this way: traditional geo satellite has a theoretical latency of 430-500ms in most cases. That means there's AT LEAST another 100ms purely due to processing delay and other hops, but typically much more.

Reply on news.ycombinator.com