No no. This isn't for some stupid mIRC crap. Anyway, if I were to establish a connection to a server and the server was for peer to peer connections kind of like napster. Would it be possible to have clients send data directly to each other through the open port? This is where the spoofing comes in. Basically it will disguise the packet as a server packet, but the client will know it's from the other client. Let me show you a diagram. client--<-->-server-<-->---client        \               /  <- spoof here        &nbsp----->-------/ The server would normally handle the file transfers, but that would require a lot of bandwidth. Now, would this work...for computers protected by a firewall? Is this legal? --------------------------- "Be like china! Use linux!" - TANSTAAFL Edited by - nes8bit on January 26, 2001 4:53:34 AM

I''ve finally got my server running again at
http://therabbithole.redback.inficad.com

That''s one of my security concerns. If a client were to access another client directly and send bogus packets the client could be rendered temporarily unuasable.

I''ll try it out this weekend and let you know if know one responds first.

Ben

@KalvinB: Clients would only be able to do that if you give them the other clients'' IP addresses. You don''t necessarily have to do that (depending on what you want to achieve)

@nes8bit: What you''re talking about isn''t really spoofing. The way you do it is simply:
- create a socket on a given port; dup it (this would work differently depending on the OS/sockets implementation/protocol type you''re using)
- one of the sockets will connect to the server
- the other one will be a listening socket

The server simply relays IP/Port pairs to clients that wish to share files who then connect to that given IP/Port pair. That''s not really spoofing.
This might work behind firewalls depending on the firewall setup. Some firewalls might allow incoming connections on IP/Port-pairs that are associated with outgoing connections.. not sure on that.
There are no issues with legality at all - after all, the user wants the program to do what it does. However, it''s always good to add those annoying "I SHALL NOT BE MADE LIABLE FOR ANYTHING THAT HAPPENS TO YOU WHILE USING OR NOT USING THIS SOFTWARE" clauses in your license

An entirely different topic is the protection against IP spoofing by hackers. If you''re using TCP/IP you don''t have too much to worry about because of the SEQ number the protocol has (you''re never 100% safe, mind you). If you''re using UDP/IP you''re on a connectionless packet, i.e. you have to make sure the incoming packets are valid yourself.

cu,
Prefect

---
Sanity is the trademark of a weak mind.

Um. I was not talking about having the client create another connection to the client. Basically one client will send a packet to the other client through the open port. Both clients will have ports designated for the server only, but I was thinking that I could send out a packet that was from one of the clients, to the client''s port that is connected to the server. Basically this would be like sliding a little "juice" into the punch. Let me try and make a better explanation.

Computer A connects to the Server. Computer B connects to the Server. Computer A requests to download a file from Computer B. the Server notifies Computer B and Computer B sends out packets with a certian ID attatched to the packet to Computer A. Computer A will know that it is from Computer B, but the IP says it''s from the Server to keep the firewalls happy and such.

---------------------------
"Be like china! Use linux!" - TANSTAAFL

Isn''t there a way to NOT bypass the basic networking protocol such as TCP/IP or UDP? (I''m using TCP) I''ve seen ways on how to do it in *real* winsock headers, but I don''t want to waste time on trying it.

Also, I didn''t think of the firewall on Computer B being a pain. That could pose a problem.

---------------------------
"Be like china! Use linux!" - TANSTAAFL

The only reason you''d need to do something like this is if both of the clients are behind firewalls (dynamic NAT).

You may want to look into supporting Socks proxies, it would let you accomplish this - but requires the users to have a functional Socks proxy server... this is how mIRC does it (albeit with limited success - but some success).


There''s a good chance that the firewall would not even recongize the a packet as spoof, but it *will* recongize that it doesn''t belong in that subnet and will likely drop it.

Does anyone have anything on Socks proxies?

Umm...I''m trying to anticipate the worst. Like firewall->firewall connections.

---------------------------
"Be like china! Use linux!" - TANSTAAFL

screw it.

---------------------------
"Be like china! Use linux!" - TANSTAAFL