Comment by majormajor | Hacker News Reader

Comment by majormajor | Hacker News Reader

Parent Full thread

> One trouble with fiber is that most people (who have cable) think the cable service they have is "good enough" in terms of performance, but they'd rather pay less for it.

This is where I am. My local speeds and offerings (cable or a smaller local company with an ethernet jack in the condo's closet) are much better than yours, and I don't really need anything better. The tech to push speed requirements dramatically higher isn't there yet. For one or two people doing some streaming and light downloading, 20Mbps is fine for now, I'd just rather pay $20 than $50. 100Mbps for $50 would probably cover me into the near future for 4K streaming, too, but currently my local providers charge more than that for the faster speeds.

I'm not sure what the next big thing network-usage wise that pushes me to want more than 100Mbps would be. Heck, I was mostly happy at 5-6Mbps cable/DSL speeds for close to a decade before video streaming starting pushing those limits - downloading linux distros or such at those speeds kinda sucked, but I didn't do it that often to justify paying a lot more.

EDIT: the commercial world looks quite a bit different, both in need and demand, so maybe there are more opportunities there, but it also seems pretty well served currently, at least in big cities. I have hardly any direct knowledge, but have been at or heard of quite a few office spaces with direct fiber links to datacenters or such. When I worked in a small town it was much worse (Charter Business Cable wasn't the greatest, there, reliability- or speed-wise, to share across the small office), maybe there's a market there?

Note that the case for fiber in rural areas that don't have cable is that DSL doesn't have a real upgrade path for long distances. (Funny there is a standard for long-range high performance DSL, HDSL, but you've never heard of it because it makes fiber look cheap)

Another story you aren't hearing is that big improvements in communication speeds from current commodity protocols (i.e. 1Gig Ethernet, 4G wireless) aren't practical with Silicon-based technology as we know it, and these things are only likely to get affordable if the industry makes a transition to faster semiconductors such as Indium Phosphide.

It would be a strange turn of events for rural areas to become more desirable to technical people because they are the first to get fiber rollouts due to it being the only upgrade path for those areas.

No one in their right mind will plow the insane money into rural areas required to trench fiber to the curb. Meanwhile a quiet revolution is happening with point-to-point wireless, where $300 gets you a gigabit connection over up to 10 miles through a directional antenna that you point at your local water tower or something, where your ISP leases space on the roof.

As someone who has a 10 mile wireless connection that I built myself all I can say is : Hahahahahahahaha. Where do I send my $300?

https://www.newegg.com/Product/Product.aspx?Item=9SIA1EA45M8...

Well that made me laugh since I actually build and operate wireless networks. Few things you should know:

1. When unlicensed wireless gear vendors talk about throughput, you can typically divide that number by 4 to get a real, achievable full-duplex throughput (because the radios are half-duplex and encoding and packet overhead accounts for roughly another factor of 2). So 750 becomes 150 which is an order of magnitude less than 1GBit.

2. That product has a range of roughly....not very far. You can do the path math yourself from their specs, but let's say it will work at full speed over 1Km. Not 10 miles.

3. To achieve the 750MBits they advertise you need to use 80MHz of contiguous spectrum in the 5GHz unlicensed bands. I will eat my keyboard if you can find a location with a water tower where that much usable spectrum is available.

To push 1GBit FD over 10 miles today needs licensed spectrum and 3' dishes with the associated high cost.

And this is before you have found an ISP willing to give you microwave access to their water tower POP (which they will not do since spectrum is so tight and they are using that tower for their own long-haul P2P links), and willing to sell you 1GBit transit for anything less than a fortune.

Thanks for the thorough reply. I've looked at how unlicensed links are set up in dense urban settings over about 1 km each, and encountered none of the problems you describe. But I can imagine they become issues over longer distances.

It's no more insane than the earlier deployments of electric infrastructure. The only difference is connection has to be made to rest of the internet at some point.

SiGe is plenty fast as mmwave ICs are using it. No need for InP. Regular Si CMOS/BiCMOS is able to do 10 Gbps transceivers.

> 100Mbps for $50 would probably cover me into the near future for 4K streaming, too

Netflix 4K streaming is only about 15mbps (and looks pretty great), so 100mbps is definitely more than enough.

The 15 Mbps is TCP bandwidth, and the 100 Mbps is physical layer bitrate. They are not comparable. There are overheads and latency and packet loss can reduce how much physical bitrate TCP can use. So the 4k stream may or may not do great in 100mpbs.

Also, how about a second 4k screen ? Multiple TVs are the norm in American homes. How about multiple tablets or mobile devices consuming HD streams ?

It seems current access link infrastructure in the US cannot support current TV viewing habits, if all the TVs were switched to 4k digital streams.

> The 15 Mbps is TCP bandwidth, and the 100 Mbps is physical layer bitrate. They are not comparable.

In practice TCP gets very close to the link rate unless you have latency much higher than what you typically have on cable/fiber. Fast.com (which "performs a series of downloads from Netflix servers" according to Netflix) benchmarks at 155 mbps on my 150/150 connection. (Cable/fiber systems are usually provisioned at a higher link rate than what is marketed: http://www.dslreports.com/faq/15643).

I'm pretty sure fast.com, like Netflix, uses either UDP and its own congestion control or multiple TCP streams, depending on client details.

I can get over 150 on single TCP stream tests like dslreports as well.

Netflix won't enable 4K on a 15Mbit connection. Perhaps 50, definitely not 15. Remember they have to allow some headroom and for the possibility of more than one concurrent stream to the same subscriber.

Reply on news.ycombinator.com