Slashdot Mirror
Creating a Backboneless Internet?
Peter Trepan asks: "The Internet is the best thing to happen to the free exchange of ideas since... well... maybe ever. But it can also be used as a tool for media control and universal surveillance, perhaps turning that benefit into a liability. Imagine, for instance, if Senator McCarthy had been able to steam open every letter in the United States. In the age of ubiquitous e-mail and filtering software, budding McCarthys are able and willing to do so. I Am Not A Network Professional, but it seems like all this potential for abuse depends upon bottlenecks at the level of ISPs and backbone providers. Is it possible to create an internet that relies instead on peer-to-peer connectivity? How would the hardware work? How would the information be passed? What would be the incentive for average people to buy into it if it meant they'd have to host someone else's packets on their hard drive? In short, what would have to be done to ensure that at least one internet remains completely free, anonymous, and democratized?"
You're describing the original design of the internet, which we're still running with essentially.
In practice though, it would be insane to let everyone with a DSL line to two different locations update routing table through the entire internet. The mechanisms to allow this exist (bgp, ospf) but major ISPs that don't want their network to fall apart prevent it because their service would quickly turn to crap. ISPs with missing filters have actually caused internet wide splits, when the entire internet tried to route through someone's T1's connected to two different ISP. BGP with a little better cost system could help that, but anyone could still cause a split anytime they liked. Think of an entire internet that acts more like IRC.
The core of the internet is still just a bunch of peers, but if you want things to stay up, they've got to be a select group that really know what they're doing. You're still free to peer directly with anyone you want, just don't expect everyone else to use your internet connection to get there too. Most people don't want to have to buy two internet connections for marginal gains anyway.
Perhaps a software solution like TOR or Freenet could help you sleep better at night?
If you need something like a terabit of bandwidth between the US east and west coasts, consider how many peer to peer link chains across the country will be saturated carrying it.
One of the major problems right now in the commercial ISP backbone environment is what happens if there's an outage; what's called route flapping, where routes dissapear and reappear, and all the routers affected have to recalculate how to get to various endpoints, can already saturate the router CPU logic for big, industrial grade room-full-of-racksize-router backbone facilities. Going to a more diffuse network at high bandwidth requirements exponentially makes this worse.
P2P across a city? Not ridiculous.
P2P across the world? Baaad idea.
Before and during WWII all mail crossing an international border in or out of the US was steamed open and read. This included all mail, all packages, all telegrams, and all telephone calls. In addition to all mail being steamed open and read, it was censored if the Army deemed it to be necessary to support the goals of the Army. Letters would arrive with portions cut out by scissors. They also censored all international media -- radio, newspapers, and magazines both incoming and outgoing.
It's quite easy to imagine as it's already been done.
Imagine, for instance, if Senator McCarthy had been able to steam open every letter in the United States. In the age of ubiquitous e-mail and filtering software, budding McCarthys are abel and willing to do so.
As an administrator of a few reasonably small domains, my first thought was oh, the fools!
You don't want to read every piece of e-mail that comes into even one site, let alone the whole internet. You don't even want to try to write programs to do it.
You would be better off trying to understand the inner thoughts of a lava lamp then trying to figure out why anyone thinks anyone would buy "farmasuiticals (the 1 U've been lOOking 4!)", let alone ingest them! Or invest in "s+0cks" that are about to "+ake 0ff" based on the say so of a stranger named "Brandice Hornyslut." Or the pointlessly malformed sludge, the server errors from misconfigured machines...if anyone really wanted to hide something they'd be about as well off e-mailing it as flushing it down the toilet--and trying to find it would be about as pleasant.
--MarkusQ
You've described the original implementation of USENET. Participating machines would dial each other up and exchange current traffic. A message injected at one machine would eventually end up in the rec.practicaljokes.hotfoot newsgroup on every participating machine within a day or two, just by this simple machine-to-to-machine exchange.
-The complexity of the routing tables. Although people complain that we are running out of IP address space, this isn't exactly true. The problem is in badly fragmented IP address space. That is to say that the route tables of our core routers that join the backbone providers have grown to be huge. There are a whole pile of class C networks (254 hosts each) that the IANA is trying to claw back so they can be consolidated into larger /16 and /8 CIDR networks.
-BGP AS space. Due to what i can only assume was poor foresight, the AS# used to identify BGP "Autonomous Systems" (Corporations, and entities that use BGP to exchange routing information with the backbone providers) is a 16 bit value. So there are only ~65K numbers that can actually be given out.
-Complexity of configuring these routing protocols. It's rocket science, plain and simple. A misconfigured BGP router will not work, and may even disrupt traffic over the rest of the internet. If anyone was allowed to broadcast any BGP route without the consent of all their peers and a pile of red tape, i could advertise a route to 24.0.0.0 and half the internet would disappear for a good number of cable-broadband users.
-Required bandwidth, and latency problems. The current top-level backbone providers have many millions of dollars worth of equipment and high-speed point to point connections to keep the number of hops for each packet to a minimum. They have the capacity to push more traffic than you'll use in a week down their wan links every second. This is a vast improvement over a pile of 56, 1024 and 3068 kilobit connections that would be meshed together in a distributed model.
What? Me? Worry?
Maybe I'm getting grouchy in my old age - see parent for details. This is how real men connect to the internet:
There are three ISPs in the world - Sprint, UUNet, and [other]. Get on the phone and order a T1 from one of the two real ones. They'll get your payment information and then someone will ask how many IP addresses you need. Tell 'em you want a
Go to ARIN.net's site. Figure out how to get yourself an autonomous system number. Call up the other ISP you didn't originally order from and get a circuit from them. No IP addresses required, we'll just use the block from ISP 1.
Assuming you're using a Cisco box do the following:
router bgp [your AS number]
network [your shiny new
! UUNet
neighbor yadda yadda AS 701
! Sprint
neighbor yadda yadda AS 1239
And *poof*! Your little
Take this little story and abstract it a bit - there is no 'backbone' to be found on the internet, just a web of large carriers with all sorts of peering agreements with each other. This won't happen at the home DSL router monkey level, but the diverse internet the asker speculated about already exists and happens to be pretty resistant to fools trying to monitor it.
I am very easy to get along with, but I don't have time to waste being nice to people who are being stupid. -Theo