i was reading this topic
http://ftp.icm.edu.pl/packages/socks/socks4/SOCKS4.protocol
and what im trying to do is:
i have a client/server application, what im trying to do is to use socks 4 BIND request to bind my server to a remote socks server, and make the clients connect to that socks server and the socks server will make them connect to my server (at least thats how i understand socks BIND request)
but i don't fully understand it (my English is kinda bad), what im asking is, is it possible to do so when i dunno any of the remote IPs of the clients? since the server's BIND request package must contain the address of the remote client and i dont really have than since the clients are from unknown users retrieving status info from my server (or can i use 0 for INANY_ADDR) ?
What you are asking for is not possible with SOCKS, nor is it meant for that purpose. Read the spec again more carefully. The BIND command is meant for use with multi-connection protocols (like FTP), where a primary connection is used to communicate between a client and a server, and BIND facilitates situations where the server needs to connect a secondary connection to the client after the client tells the server where to connect. In that situation, the client would issue a BIND command to SOCKS telling it the server's IP/Port so it only accepts that connection, then send the resulting SOCKS listening IP/Port to the server to connect to.
What you are asking for is better served by using a router with Port Forwarding rules defined. Then you can open a listening port on the router that accepts any inbound connection and forwards it to your app's listening IP/Port. Most modern routers support uPNP (Universal Plug-N-Play) so you can configure the forwarding rules programmably instead of requiring admin access to the router's configuration software.
Related
A server, say for example an FTP server is listening on a port 21. Multiple clients can connect to it simultaneously and get ftp services asynchronously without being blocked.
My assumption is that the server has a separate thread to handle each client request. (Question: Does the new thread start negotiating new port with this client or continue using the same port (21) for data exchange ?)
my understanding is that the server uses a single port (21) to serve all the clients, including new connection requests.
if that is the case how does the server using a single port differentiates between clients and how does it delegate relevant data for each relevant thread ?
The same was as any other TCP/IP server.
Each client have a unique IP address and/or a port number.
The combination of client address and port and server address and port creates a virtual channel. The channel is used to exchange data between client and server software running on different machines. This is typically built into OS. The client and server software does not have to handle it specifically.
I'm working on a desktop C++ application which uses OpenSSL sockets (a raw TLS socket, not HTTPS) to communicate with our server.
One of our clients are required to route their traffic through a proxy. The client is using ZScaler in Tunnel with Local Proxy mode.
In theory, it's possible to reconfigure ZScaler to force our traffic through a proxy chosen by ZScaler. However, I want to investigate solutions where our application uses the Windows OS-level proxy settings rather than relying on ZScaler configuration.
I've read this post:
openssl s_client using a proxy
But I'm uncertain whether those answers apply to my situation, because that user didn't mention whether they're using Windows or Linux, and they appear to be talking about an HTTP/HTTPS proxy. Also, that question appears to be asking about the s_client function, rather than simply creating a TLS socket to my server through a "Tunnel with Local Proxy" on Windows.
So, my questions are:
Can OpenSSL be used to open an SSL socket to a server through Tunnel with Local Proxy?
Can we make an OS call to determine the IP/socket for the Tunnel with Local Proxy configuration?
If this is possible, then I have another question: suppose we have a single proxy at 10.100.10.0:5000.
If one user in our client's office opens a socket to our server via their proxy, will a 2nd user be unable to connect from their office because they're bottlenecked at single proxy socket?
Put another way: what is the standard way of implementing proxy-awareness for a Windows application using OpenSSL?
Note: This question was originally posted to Network Engineering stack exchange, but it was closed because it refers to an issue above OSI layer 4.
Note: I'm looking for a solution that does not require administrator permissions on the user PC. I would prefer for our application to discover and use OS-level proxy settings without making any administrative changes to the machine, i.e. by calling netsh.
Can OpenSSL be used to open an SSL socket to a server through Tunnel with Local Proxy?
OpenSSL doesn't do it for you but OpenSSL does not prevent it either. The tunnel has to be established before you do the TLS handshake to the endpoint. Depending on what kind of proxy this is you might need to use a HTTP CONNECT method for this or might need to use the SOCKS protocol or whatever your proxy requires. In case of ZScaler this is likely the HTTP CONNECT method but you need to make sure that the connection to the target IP and port is actually allowed by the security policy.
Once you've established the tunnel to the endpoint using the proxy you can just build the SSL socket on top of the TCP socket for the tunnel. Just do the usual SSL setup (i.e. SSL_new etc) and then associate the SSL object with the existing socket using SSL_set_fd. Then proceed as usual with the handshake, i.e. SSL_connect or similar.
Can we make an OS call to determine the IP/socket for the Tunnel with Local Proxy configuration?
I don't know but Winsock use system proxy settings might answer this part.
If one user in our client's office opens a socket to our server via their proxy, will a 2nd user be unable to connect from their office because they're bottlenecked at single proxy socket?
This should not be a problem. It is perfectly normal to have multiple connections through the proxy.
In short we have two separate servers for our web app. The first one is the main server that uses Websockets for handling "chat rooms", and the second server only handles WebRTC audio chat rooms via Websocket. Both servers use Express to create a HTTPS server, use secure Websocket and the port 443.
I recently encountered a problem where a corporate client's firewall blocked the wss-connection to only the WebRTC server. The error logged in the user's browser was "ERR_CONNECTION_TIMED_OUT", which means the user never connects via Websocket. This has not happened with any other clients.
The Websocket connection works normally between the user and the main server, and no rules have been added to their firewall to use our app.
Has anyone encountered something similar? What kind of a firewall setting might cause this? Could this be a cors problem, since the servers are on their own sub-domains?
The main server could be restricting the type of data sent on port 443, which will use SSL to secure that transmitted data.
Refer to this page for information on the "Well-know port numbers".
The WebRTC audio data may need to be transmitted on its own dedicated port number that has been configured on the main server for this.
The problem was that the main server WebSocket used TCP and the WebRTC server used UDP, and UDP was blocked by corporate firewall on default.
WebRTC should use TCP as a backup, but I'm assuming UDP is still needed for the handshake.
I know already about the web-sockets, and they are great, the problem with them is that they have to keep the connection open in order to be able to communicate.
I have a small system where from time to time the server has to update the status and notify the clients about that, and keeping the connection open from every client is not so optimal. At same time is very important that the update on the client side to be made just in time.
So my question is, if the server has a unique address does the client have a public temporary address where the server can send request? So when the client will connect to the server it will provide it's unique address and the server will cache it, and when there will be an update the server will send the request to that address?
I understand that there many problems as the address will constantly change, but this is already other question.
If client does not have a dedicated IP-address then it is not available from WAN unless it has an open connection with any node in it.
When client from local network sends request to a server it's (client's) router remembers client's local IP-address and port and translates it using NAT protocol to one of router's free ports and then sends data further with router's own 'IP-address of the sender' in IP protocol header and 'Sender's port' in TCP header. When router get's server's response it uses NAT table from it's memory to translate addresses back and deliver data to the client. Addresses are normally kept in NAT table while connection between server and client is open. So if there are no opened connections between server and local network client then server will not be able to connect with client because server does not know how to reach it.
You say you have a small system. Why then do you think that you will not have enough free ports at your server to work with websockets? If you just want to get updates from the server (not to both send and get data through a persistently opened connection) you'll probably find long polling or SSE more suitable. It is definitely easier to implement than websockets.
my question is simple
When you send data through TCP/IP protocol with EX:firefox you can receive reply on some random port that the browser listen on, while when i try to use a port for another task like CS Gaming or anything else it don't work unless i use kind of VPN ?
PS: there r no firewall blocking connection and port forwarding from my router didn't work as well.
Browsers are client apps that make outbound connections to web servers. When connecting to a server through a router’s NAT, the NAT takes note of the source and destination IP/port pairs so messages sent back from the server on the same connection are automatically routed to the correct client IP/port.
Browsers also support the websocket protocol. This feature makes it seem like the browser is listening on a specific port. However, in reality, it is initiated on a new connection to the server, a connection which remains open all throughout the websocket communication.
What matters is which peer is behind the NAT — the server or the client. For an outbound connection from a client, it can usually use any random port that is available at the time. For an inbound connection to a server, the server's IP/port must be known ahead of time and be routable. If the server is behind a NAT, the router(s) must be configured to make the server reachable from the other side of the NAT.
The server software can make a UPnP request to ask a router to forward inbound packets to the correct IP/Port. The router, depending on its configuration, may or may not honor such a request. If not, the router has to be configured manually by a network administrator.