The Transmission Control Protocol has no competitor in sending data reliably over a network. The protocol is
implemented by operating system subroutines (API or system calls) at each endpoint. The endpoints represent
a client and server (or source and destination nodes) within a network. These endpoints transmit data through
links (routers, switches, etc.) that follow the Internet Protocol. Therefore, TCP/IP are the two layers responsible
for reliable data transmission. The TCP layer establishes a connection and validates data between endpoints. The
IP layer divides data into packets and transfers them through connecting links to an endpoint.
The two layers provide mechanisms to counteract problems that may arise during data traffic. The TCP layer adapts
to data congestion through an algorithm, and responds to data corruption by including a packet checksum. The IP layer
incorporates a header-only checksum and reports error messages, but responds to congestion and corruption by dropping data
packets. The combination of both protocols isolates an endpoint application from responsibility except to create a
connection and send (or receive) data.
The TCP layer attempts to avoid congestion, but also responds when it deems appropriate. The response is to reduce data
transmitted from a source. Any congestion that exists within connecting links of a network may be correctly detected
by the TCP algorithm. However, the IP layer responds to congestion by dropping data packets, which need to be retransmitted.
The isolated responses of TCP and IP layers may cause further congestion and data corruption due to repeated dropping of
packets. The result is a circular situation in which data packets may not advance until dropped by any link. This problem
reflects a prime example of the disjoint between TCP and IP that warrants a resolution.