Hacker News new | past | comments | ask | show | jobs | submit login

Starlink is currently expected to get $900M of RDOF funds. I'm curious about how FCC is going to enforce performance. Starlink's bid promised 20Mbps upload speeds but the reality for users is less than 10. Download speeds aren't reliably at the 100Mbps the RDOF bid specifies, either. https://starlinktrack.com/speedtests/region/us



Starlink won't get paid until they meet the performance goals, but IIRC they have until 2028 to actually fully meet them. They do have to have a plan of how that would happen, and FCC must accept that plan.

It's public knowledge that they intend to launch a lot more satellites before that, so they are probably counting on that to help them.


>> It's public knowledge that they intend to launch a lot more satellites before that, so they are probably counting on that to help them.

Yeah, but they also want to substantially increase the number of subscribers. It seems a shame to add capacity and not users just to get a cheque from the government for hitting some arbitrary numbers. IIRC they could have bid for a lower speed target that they would probably be hitting today.


> It seems a shame to add capacity and not users

... I think they will be doing both. They currently have ~2500 sats in the ~53° constellation. The V2 system that has been approved and is targeting similar latitudes has 3 times the sats, and each sat is claimed to support 9 times the bandwidth of the current ones. So if they need to triple bandwidth per user, they can still increase subscriber count 9-fold and meet it.


The speed categories were set by the FCC, and SpaceX would have gotten nothing if they weren't in the "fast" category.


Sad to say, but 50% of promised performance is spectacular with regard to American telecom companies' rural promises.


Just visited fast.com, got 55 Mbps down on a whim (Starlink in Michigan). Sure, it doesn't hold a candle to in-home fiber, but it definitely beats the pants off HughesNet and our local WISP. I'd say for Monday afternoon traffic, I'm getting really solid performance. Definitely worth the $100/month just for the unmetered connection alone.

I too am curious with how far they can push performance though. I've got a hunch that it has something to do with balancing satellites, users and regional coverage, but only time will tell if they make good on their promise. Even if things stay the same as they are right now though, there's nothing else that comes close in my area.


How does RDOF test that? Speedtest.net? Fast.com? Rarely have I received 100Mbps via http transfer mechanisms, and that's the majority of what people do all day. Websites, streaming video, etc are all small discrete downloads via lots of multiple requests. They don't have time to get "up to speed". What that 100Mbps is allow ~10 people to use ~10Mbps at the same time. That's the testing I'd be interested in seeing proof of their service being met.


Historically the FCC relied on self-reported numbers from ISPs to verify performance. Unsurprisingly, ISPs lied. Lately FCC has gotten smarter about verifying performance.

The RDOF performance requirements are much more subtle than just average speeds though, there's details about percentiles and time of day and the like. I'm hopeful someone is being thoughtful at the FCC


If they do any testing it will probably be through SamKnows.

I don't think 10 people has anything to do with a single household but maybe I'm just an out-of-touch liberal coastal elite. But seriously, one person can use around 25 Mbps which is why there's discussion of defining broadband to be 100 Mbps per household.


I believe HLS chunk downloads reuses same TCP/QUIC connection.


To be honest, that part of how the internet and browsers work is beyond my knowledge. I just know how to create HLS packages, and then see how the browser makes the multiple requests in the network tab. If browsers are smart and keep connections like that open vs renegotiating per request then it makes sense to me. Maybe this will be my today's TIL for me.




Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: