A better metric would be a measurement of user initiated actions. Now sure how to pull that off, but it would be a more accurate measure of how the net is changing.
Besides, the direction and relative values show clear and growing dominance.
It's very challenging to figure out how to do this in a way that benefits users and improves user experience - we are very careful about being fair to users on mobile connections - but I think that a distributed web platform is the future and the only way we can combat the (lack of) net neutrality.
We already have a distributed internet, just not as distributed as we had hoped for from a censorship perspective.
Bandwidth is expensive and many let youtube host the video for free instead of paying to host it themselves which is why it is no surprise that Youtube tops the list.
This is easily solved by switching to peer to peer protocols, which would work like a charm if our providers did the sensible thing and gave us a symmetric bandwidth to begin with.
By the way, even now, a user have to send requests to the right place to get any data. IP packets still have to go both ways.
I was also talking about the ADSL tragedy. We should have had SDSL from the beginning. (And yes, the corresponding decrease in raw download rate would have been worth it).
If your only connection is through your phone, I have to feel a bit sorry for you.
One can still have an R-Pi at home.
These devices are also thermally constrained.
There's far more to this than upload speed.
However 5G promises much higher bandwidth and more symmetric data rates
First came dial-up modems. Can't have a server with those without monopolising the phone. Then came the DSL. Providers noticed that nobody has a server, and deduced in their immense commercial wisdom that they didn't want a server. So they made the DSL Asymmetric. Now people could have a server, but the upload was so terrible it wasn't worth it. Conclusion: the ISPs were right: nobody wants a server.
There was this P2P fluke, but the copyright lobby kindly explained this was only used for illegal stuff. Everything's mostly back in order now.
Anyway, now we've established nobody needs nor wants a server, we can put NATs everywhere and nobody will notice.
---
DSL was the first avoidable mishap in a long string of centralisation trends.
How would _you_ implement a secure P2P where you (and only you) can rapidly change content (dynamic web pages) and control who has access to authorized pages?
What about credit card info? Passwords?
I imagine there "are" implementations that solve many of those issues. But it's sure-as-heck not common knowledge if there are "easy" answers to all those problems.
I can't help but feel "Reflections on Trusting Trust" by Ken Thompson screaming at me. How do you ensure every bit of the web stack isn't hacked on every computer in the network, as opposed to your own?
[1] [PDF] http://vxer.org/lib/pdf/Reflections%20on%20Trusting%20Trust....
In fact, when dealing with P2P, you're more likely than not, to be full of compromised programs. Many unintentional, but many others intentionally. Torrent users who run clients that don't upload. Every P2P game ever made being hacked. (See the entire purpose behind QuakeWorld moving to a client-server model over P2P, over the original quake.) And every Kazaa uploader had viruses on his PC back-in-the-day.
I'm not saying P2P is impossible. But first, you've got to define what application you want to implement (webpage, vs public video hosting) because the solutions are completely different. And even then, you're still walking into dangerous territory when you can't control the machines your relying on.
What happens if you need more bandwidth than your usernet can give? You can't simply "install more servers" or "buy more bandwidth" when your "server" is every user connected to your site.
And what happens when you roll out new versions, and some users don't upgrade?
The list goes on. And I say all of this as someone who has considered using distributed computing model for game hosting, where machines assist in computations and bandwidth. It's possible to do, but it opens up entire dimensions of additional problems to solve.
---
The trust problem still applies to HTTP. We still download viruses on the web, and we still lose our credit card numbers to con artists. The only reason there's more malware on current P2P network is because those networks are disproportionately used to infringe copyright. The risk doesn't come from the distributed nature of the network, but from the lack of legality of the content.
Securing peer to peer communications to current web levels is trivially easy: just sign the damn data, and have a certificate authority ascertain the identity of the signer. For static content such as YouTube videos, you can also use a content addressable system.
While that would require some level of centralisation, it woulndn't exceed that of DNS, and would definitely solve the bandwidth issue.
> What happens if you need more bandwidth than your usernet can give?
I won't happen, because we enjoy symmetric bandwidth, thanks to our regulators being sensible, competent people. (At least that's the case in my rainbows & unicorns world). Seriously, though, symmetric bandwidth is the ultimate and only solution to many problems: it ensures total upload keeps up with total download, so we get the equilibrium we want.
That would be flawed if individual sites couldn't also host videos -- or if we didn't care whether they do.
a measurement of user initiated actions
how to pull that off
Traffic is traffic, and it’s no one’s business whether I’m watching a video, clicking a button, issuing shell commands, or typing a reply to a message.
No thank you, to the idea of changing protocols, or forcing people onto the application layer, in furtherance of analyzing our already-over-analyzed activities with an electron microscope.