[mnet-devel] Grid Of Trust -- pre-design

[Jim Dixon]

On Wed, 10 Dec 2003, [iso-8859-1] Some Guy wrote:

> > You are describing a service which has been available for decades.  This
> > is how leased lines work.  When you get a leased line, this is what you
> > get.  It's not cheap.
>
> Yes, that's been around before internet access.  It has since been
> improved appon.  When I buy a leased line from New York to London,
> they don't lay a seperate trans-Atlantic cable.  In fact they can
> share the same physical medium with lots of other lines.  If I buy 50
> of them going to different offices they can all share the same
> physical line going into the main office along with regular internet
> connectivity.  It's just a question of routing.  The bandwidth along
> the "leased line" connections can be safe from flooding of the
> internet connection.

This is just a common service.  It's expensive, very expensive if you are
actually talking open your own private circuits.  But it has been
available for decades.  This is nothing new at all.  For small pipes, it's
based on TDM, time division multiplexing.  For large pipes, you use WDM,
wave division multiplexing.  WDM is new, it dates from around 1996.  The
service is provided in terms of lambdas.  A lambda is a frequency of
light.  How much bandwidth you can push down a lambda depends upon the
equipment at both ends of the circuit and the quality of the fiber.  I
think that 2.5 Gbps is a common sort of number for shorter distances.

What's new is the cost.  I did a presentation on this for a EuroISPA
conference in 1998 or so.  At that time a 2 Mbps circuit across the
Atlantic or between European capitals cost about $30,000 a month.  Since
then the cost has dropped precipitately.  I haven't looked recently but I
would expect it to be somewhere between say two and ten thousand dollars a
month.

> ATM is almost like what I'm asking for.  You can allocate a connection
> which will get certain quality of service guarantees.  The only
> problem is that I can still be bombarded with connection requests.
> Which means they can flood you so that nobody can connect to you, but
> your current connections will be safe.  The only perfect 100% way to
> solve the problem is to have connections initiated from both sides,
> which requires out of bound communication.

ATM is old technology and no one in his right mind would set up a service
based on it today.

MPLS is more like it -- but how do you pay for this service?

> Here's another fun idea, ISPs could give you a random set of IPs (with
> the new IP).  Each one could get a fixed amount of bandwidth.  If an
> adversary couldn't guess all the IPs, you'd be safe.

The IP addresses have to be routable.  This means that they form part of
a CIDR block, a contiguous block of IP addresses aligned on 2^N
boundaries.  All routing on the Internet is done in terms of CIDR blocks.
Such blocks get advertised in OSPF and BGP4 announcements.  This is how
routing is done.

No ISP is going to provide the service you describe, except insofar as
they already do so.  That is, if you get an IP address from an ISP, either
it's static, in which case it is randomly assigned to you from a CIDR
block, or it's assigned to you dynamically when you connect, in which case
it's randomly assigned to you from a CIDR block.

The CIDR blocks involved are usually fairly small.  Each CIDR block is
routed by some specific router.  Except where there is firewalling, anyone
on the Internet, anywhere in the world, can identify your router by
running a traceroute to your IP address.

You are not "safe" in any sense.  The Internet is a global shooting
gallery.  Everyone knows where you are.  You can buy a certain amount of
protection by acquiring your own IP address space and multiple connections
to the Internet.  That is, you can protect yourself at this level by
setting up your own ISP.  It's not cheap.

>  This is kind of like maintaining multiple mailboxes, so when one gets spammed up you know sold
> your address out and you can just ignore all his mail.
>
> Am I giving you any ideas here?

Not so far.

> > > Each of my quake buddies has an internet connection at home where he
> > > runs his sub-node.  We can all trust each other.  If they know one of
> > > our IPs because one of our neighbors was a spy, they can flood him.
> > > That shouldn't stop us from servicing our other neighbors.
> >
> > I think that you fail to understand how Internet routing works.
> >
> > Your subnode is going to have one (1) IP address.  You can hang up to 64K
> > friends off of it in private IP address space, each on his own port, but
> > they will all be sharing that same IP address in the global IP address
> > space.  When someone decides to zap that IP address, you are all toast.
>
> Jim, I sick of repeating it.  Each of my dozen friends is at a
> different house, maybe in the same town.  We each have our own IPs,
> maybe even different ISPs.  If one gets flooded sure, he'll be cut off
> from the world.  You want to keep thinking of in terms La Resistance;
> we're a cell.  Though some of us may die La Resistance goes on
> (Southpark).

Let's see.  You say that each person runs a subnode.  This appears to be
indistinguishable from a node.  It has one IP address reachable through
one ISP.

Are you suggesting that you set up a VPN (virtual private network) between
these nodes?  You could do this.  In that case you could use some one
person's IP address to route everyone's traffic.  Let's call this guy the
patsy.  The upside is a certain illusory security.  The downside is that
if there are N participants in the VPN, then the patsy loses most of his
bandwidth, say (N-1)/N of it because everyone's traffic is passing through
it.  And also, when someone decides to DOS your VPN, the patsy loses ALL
of his connectivity.

There is some moderate benefit to everyone else, but exactly how do you
find people dumb enough to play the patsy?

> This cell, perhaps you perfer "cluster", can participate as a single
> node in Grid of Trust or other P2P networks.  It has the power to give
> away only a few IPs at a time to minimize damage that can be done to
> it by its neighbors.  This is your style of idea.

Well, I wouldn't say so.

The ideas you are coming up with generally require that you build your own
global ISP.  You make no provision in any way for paying for it.  Go ahead
if you like, but I don't see a viable revenue model.

In the year 2003, it makes a lot more sense to let someone else lose money
provide Internet connectivity.  The problem is to use the extraordinarily
cheap Internet bandwidth that is commonly available to build secure and
reliable overlay networks.  As soon as you start digressing into trying to
change the way the Internet works, your costs skyrocket.  Your solution
must not involve re-engineering the Internet.  You must not try to
constrain the way that ISPs run their businesses.  You have to understand
how they do work and structure your solutions around this.

--
Jim Dixon  jdd at dixons.org   tel +44 117 982 0786  mobile +44 797 373 7881
http://jxcl.sourceforge.net                       Java unit test coverage
http://xlattice.sourceforge.net         p2p communications infrastructure



-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials.
Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
_______________________________________________
mnet-devel mailing list
mnet-devel at lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mnet-devel