interconnected routers. Through this functionality, the
Internet layer makes possible internetworking and the
connection of different IP networks, essentially
establishing the Internet. The packet forwarding at the
Internet layer is best effort and unreliable, and any error
correction and retransmission mechanisms required
have to be implemented at higher layers. The Internet
layer is agnostic to the data structures and the operations
that the layer above it is implementing. If we go back to
the postal mail analogy, the Internet layer acts as the
postal service infrastructure and the mail carriers who
deliver the letters and packages without knowing what
they actually contain. In the same way, the Internet layer
receives data segments from the transport layer and
carries them all the way to the destination without
knowing what the actual data is.
The transport layer is at the core of the TCP/IP reference
model. It establishes end-to-end data channels over
which applications exchange data, regardless of the
structure of the user data and the underlying network.
Since the Internet layer is best effort, the protocols at the
transport layer provide flow control, congestion control,
segmentation, and error control. Differentiating between
different applications running on the same host occurs at
the transport layer through port numbers. The transport
layer port is a 16-bit logical construct allocated for each
of the communications channels an application requires.
The ports in the range 0 to 1023, called well-known
ports or system ports, are used for system processes that
provide widely used types of network services. For
example, web servers that implement HTTP are by
default listening for connections on port 80 for regular
HTTP traffic or on port 443 for secure HTTP traffic via
HTTP Secure (HTTPS). End-to-end connections at the
transport layer can be categorized as either connection
oriented (usually implemented using TCP) or
connectionless (using UDP).