Hacker News Reader

Comment by cyphar | Hacker News Reader

Glad to see that GNU/Linux is getting a DTrace-equivalent. Sucks that it's all a bunch of string and staples rather than a single ioctl interface like DTrace -- but hey, that's the Linux way.

View on HN

Reminds me of developing countries that leapfrog industrialized G8 countries in certain technologies, like skipping landline deployments and going straight for wireless networks.

OTOH, industrialized countries leapfrog them in certain other technologies, like wired networks.

I'm not sure if wireless would scale to the traffic capacity of our developed and industrialized telephone and TV cabling, especially in dense cities.

That is not leapfrogging. To have decent internet access that can power economy - you need some form of wire to the home. S/FTP,coaxial, twisted pair or fiber.

At the moment.

Laws of physics tend to change infrequently. Sharing spectrum means low speed for all.

So beams must be directional, to avoid sharing of spectrum with all, but only with some.

The problem is that broadcasts are exactly that: broadcasting. They transmit in all directions. While you have bandwidth constraints in wires, they're all localised to the wire. Even parabolic dishes (which make getting a signal a huge PITA) still aren't as localised as cables. Wired connections get a bad rap, but they're actually a much more preferable (and scalable) medium for transmission than radio waves.

Not to mention that "bidirectional beams" is another way of saying "noise".

A fiber of glass is the ultimate directional antenna.

I'm not suggesting that, but that economics changes things much more rapidly, c.f. 5G, which may roll out multi-gigabit speeds to consumers far before landline.

And how you will provide multi gigabit speeds to a 100 people in close proximity simultaneously? There is the question of throughput of the whole network.

why can't TCP/BATMAN be defined as "decent internet access"?

It's a shared medium and can never reach the qualities of wired networking, though it might get better enough that we don't notice it as much as we do now. This reminds me of efforts like xDSL G.fast. People only talk about approaching the speeds of common fiber configurations, but miss the point of very low latency in FTTH and easy upgrade to, say, 10 or 40 Gbs, whereas that would require a technological enhancement in DSL or CABLE. With fibre it would only require switching out the "modem" and network gear in the local uplink station (hut).

Eh, you're not gonna get 10 or 40Gbps over most FTTH installs. Most run GPON with beam splitters which maxes out at 2.5Gbps/line. To get 10/40/100 you need OS1/2 singlemode or OM3/4 multimode fiber which I've never seen used in residential/small business FTTP setups.

GPON isn't used everywhere, but it's used a lot, yes. Still, getting 2.5Gbps, exclusively, with low latency isn't possible with the other connections (yet?).

No doubt fiber is definitely the way to go if it's an option. Other than fiber, the closest you'll get is DOCSIS 3 theoretically runs at ~1gig downstream. Anecdotally, I've had Fios for a couple years and get faster than advertised speeds consistently and never had an outage, so I'm all about fiber.

I think it's due time that having an actual fiber drop becomes as common place as having a phone line or cable line. With how much people use the internet these days, it makes sense to have it's own connection instead of piggybacking on whatever physical connection people already have.

Yeah the UX is user hostile to say the least.

Shameless plug: I've started working on a more coherent tool. Inspired by awk and DTrace, but it is still in its infancy.

https://wkz.github.io/ply/

Reply on news.ycombinator.com