Question

Request for Comments 1122 states (of TCP): A host MAY implement a "half-duplex" TCP close sequence, so that an application that has called CLOSE cannot continue to read data from the connection. If such a host issues a CLOSE call while received data is still pending in TCP, or if new data is received after CLOSE is called, its TCP SHOULD send an RST to show that data was lost. Sketch a scenario involving the above in which data sent by (not to!) the closing host is lost. You may assume that the remote host, upon receiving an RST, discards all received data still unread in buffers.

Short answer

Data sent by Host A is lost if it calls CLOSE and sends an RST while Host B has unread data in its buffers.

Step-by-step solution

Understand the Context

TCP is a connection-oriented protocol that supports reliable data transfer. The 'half-duplex' TCP close sequence means one side of the connection can be closed for sending, but it may still be receiving data.

Identify Key Points from RFC 1122

RFC 1122 allows a host to implement a 'half-duplex' close sequence. If the host calls CLOSE and data is still pending, or if new data arrives, TCP will send an RST (reset) signal, signifying that any unread data at the remote host is lost.

Scenario Outline

Let's outline a scenario where data sent by the closing host gets lost: Host A and Host B are communicating over TCP. Host A initiates the sending of data to Host B and then calls CLOSE, but Host B still has unread data in buffers when it receives an RST from Host A.

Step-by-Step Scenario

1. Host A sends data to Host B. 2. Host A calls CLOSE, initiating a 'half-duplex' close sequence. 3. Data is still pending in Host B's buffer, but Host A sends an RST. 4. Upon receiving the RST, Host B discards all pending unread data because it believes the connection is lost.

Conclusion

In this scenario, because Host A initiated CLOSE and sent an RST during the half-duplex close sequence, any unread data in Host B's buffers is lost. Thus, data sent by Host A before the RST is lost if it has not been read by Host B.

Key concepts

TCP RST signal

The TCP RST (Reset) signal is a crucial part of the Transmission Control Protocol (TCP). When a TCP connection ends abruptly due to an error or other unexpected event, an RST signal is sent to signify this sudden termination. An endpoint can send an RST signal for several reasons, such as receiving unexpected packets, protocol errors, or situations described in RFC 1122.
When an RST signal is received, the connection is immediately terminated, and any data that was in the process of being transmitted is discarded. This mechanism serves to quickly free up resources and inform the communicating party that the connection has been lost.
In the context of a half-duplex TCP close sequence, if Host A sends an RST signal to Host B, all data that has not yet been read by Host B is discarded, signaling an abrupt termination of data transfer.

RFC 1122

RFC 1122 is a standard from the Internet Engineering Task Force (IETF) that defines requirements for Internet hosts, particularly at the communication layers (link layer, IP layer, and transport layer). This document includes guidelines and mandatory rules for TCP implementations.
Specifically, RFC 1122 addresses how hosts should handle different closing scenarios in TCP. It allows 'half-duplex' close sequences, which means one side of a connection can stop sending data while still possibly receiving data. This is useful for certain types of applications that need to shut down sending operations without completely closing the connection.
However, if the host calls the CLOSE function and data is still pending or new data arrives, the RFC states that a TCP RST signal SHOULD be sent. This RST indicates that the connection is abruptly closed, and any unread data must be discarded.

Reliable data transfer

TCP is known for providing reliable data transfer. It ensures that data sent from one host to another arrives in the correct order and without errors.
Reliability in TCP is achieved through mechanisms such as:

• Sequence Numbers: Keeping track of data chunks to ensure they arrive in order.
• Acknowledgments (ACKs): Confirming receipt of data packets.
• Retransmissions: Re-sending data if acknowledgments are not received.

However, in a half-duplex TCP close scenario, if an RST signal is sent, this mechanism of reliable data transfer is bypassed for any data still in transit or unread. The connection's abrupt termination means that any pending data is lost, and no further acknowledgments or retransmissions occur.

Host data loss

Host data loss occurs when data that has been transmitted is not successfully received or acknowledged by the other host, or when it is discarded due to connection issues.
In a half-duplex TCP close sequence as described in RFC 1122, host data loss can occur under the following conditions:

• Host A sends data to Host B.
• Host A then calls CLOSE, but Host B has not read all the received data.
• Host A sends an RST signal, prompting Host B to discard any unread data in its buffer.

This sequence ensures that any unread data on Host B is immediately discarded, effectively causing data loss from Host A’s perspective. This loss is particularly detrimental if crucial information was in transit, as recovery mechanisms such as retransmissions are not invoked after an RST signal is received.