[Sinned]

Hello everyone,

There are a lot of documentations about P2P networking with a main server, which keeps track of all clients - and to this server each client connects to get a list of all other clients, in order to connect to them.
This is not what I want, because I heard about server-less P2P networking.

However I don't really understand how the clients could find each-other without the list, the concept of server-less P2P seems interesting to me.
So I was wondering if anyone could explain a little about it (and maybe about implementing in C/C++) or give me a reference to a online documentation.

Thanks,
Sinned

[RudiVisser]

Well of course you can do this but like you say, the client list is going to be the difficult stopping point and there's no real way to do it unless you have some form of discovery service that scans everything.

[modi123_1]

A few things to read:

http://msdn.microsof...y/cc297274.aspx

http://www.cs.duke.e...lessim_urso.pdf

http://www.dreaminco...2p-connections/

http://msdn.microsof...v=vs.85%29.aspx

[no2pencil]

Sinned, on 17 September 2012 - 10:56 AM, said:

because I heard about server-less P2P networking.

However I don't really understand how the clients could find each-other without the list

Even though there is no centralized 'server', each client would also act as a stand alone server binding to a port. So in a mass client to client network, I would think each client would act as a server as well as a client. That's how the P2P networks seed.

[Sinned]

There are two things I don't understand on (server-less) P2P networking.

Each peer would act as a server as well as a client.
Normally when you start running a server - actually waiting for incoming connections on a certain port - your firewall has to allow it as well as your router has to forward/open the ports.
So when I run a normal server (client-server architecture) it won't be accessible if my router is not configured. But I can participate in P2P networking without any problems.
Here I don't understand how this "server" could run. (I understand the client part)

How the other peers could be found.
As I said there won't be a list. And scanning all possible IPs from 1.1.1.1 to 255.255.255.255 - if there is another peer behind - sounds "a bit" devious to me.
I read about peer groups and channels, but that also doesn't look very server-less. Because those groups and channels should also be maintained by a main-server. Doesn't it?

So it is all really unclear to me.

[modi123_1]

It operates like gnutella or any other torrent-esque fashion.. you need to know at least one other client and essentially queries for users get flooded across all the clients until one is found. The starting off point would be having some sort of dummy machine on people can snag into initially and get a more larger list of other clients from there. Example.. you the creator of the software will probably have a machine running it, right? Have people automatically get that address and snag into the network!

http://computer.hows...le-sharing3.htm
http://computer.hows....com/kazaa3.htm

[Sinned]

I think I misunderstood the concept of server-less P2P networking.

For me the "super-node" or "another client" which has to be hard-coded in the program, sounds like some sort of server to me.
Now it looks a lot less impressive
(I will read myself more in the subject)

Anyway everyone many thanks for the help and explanation.

[rogerperkins]

Sinned, on 18 September 2012 - 06:40 AM, said:

I think I misunderstood the concept of server-less P2P networking.

For me the "super-node" or "another client" which has to be hard-coded in the program, sounds like some sort of server to me.
Now it looks a lot less impressive />
(I will read myself more in the subject)

Anyway everyone many thanks for the help and explanation.

Your program could work in conjunction with 'freenet' and store and retrieve the seed file from the freenet. Users would have to install freenet on there computer as well as your program.