HTTP and TCP/IP

[L Goundalkar]

Hi all,
HTTP works on TCP/IP connection. Does this mean that its a from of "Tunnelling"?
Thanks.

------------------
L Goundalkar
[e-mail]
Sun Certified Programmer for Java 2 Platform

[nan sh]

No, Tunnelling is under Application level(HTTP) and above the TCP/IP,
Tunnelling needs TCP/IP's.

[L Goundalkar]

Hi,
As per the Servlet Tunneling definition, i.e.
" Tunneling is a way to use a existing road of communication and create a subprotocol within it to perform specific tasks. The subprotocol will contain all the necessary information to create an object on the web server, invoke methods on that object and return results back to the client."
So doesn't we use the communication link of TCP/IP through HTTP. This means that HTTP is the subprotocol which is used. So isn't it a example of tunneling?
Thanks.

------------------
L Goundalkar
[e-mail]
Sun Certified Programmer for Java 2 Platform

[nan sh]

Http is a standard protocal, when we say tunnelling, normally means your private protocal, well I think this question is too academic that I can answer.

[L Goundalkar]

Thanks Nan.

------------------
L Goundalkar
[e-mail]
Sun Certified Programmer for Java 2 Platform

[Peter den Haan]

HTTP does not tunnel through TCP/IP, but is sitting on top of it. Why? Because it is making heavy use of TCP/IP facilities. When you request a URL, you are using a "connection" to a "socket" on a host represented by an "IP address" - all these quoted words are infrastructure coming from the TCP and IP layers. You could conceivably use some other infrastructure if you wanted, but it would need to offer facilities similar to TCP/IP for HTTP to work.
Tunneling works differently. In tunneling, you route a protocol over another, completely unrelated protocol. You don't need or even want that other protocol, but have to use it because nothing else gets through. Tunneling is usually (but not necessarily) done through HTTP because that's how you can get through firewalls most easily.
Take, for example, tunneling the RMI protocol through HTTP. RMI, in itself, does not make use of HTTP at all. The infrastructure of HTTP (GET and POST requests, headers, response codes, etc) is completely foreign to it, and you wouldn't dream of using HTTP for RMI if you can avoid it. Yet HTTP tunneling packages the RMI call inside a valid HTTP request; the receiving servlet (or cgi binary or...) unwraps the RMI call and performs it; the result is again wrapped in a valid HTTP response. That way, RMI remote procedure calls are smuggled through any firewall that allows HTTP traffic.
- Peter
[This message has been edited by Peter den Haan (edited November 07, 2001).]