Slide 1 [Introduction]
This lecture will cover the fundamentals of the transport layer, and how the most commonly available transport layer protocol: TCP. As we will find, it is TCP that builds-in the robustness of the Internet, as IP does not provide for acknowledge-ments, and there is no guarantee of delivery. TCP, of course, is not the only transport layer protocol, others include: SPX (as used with IPX in Novell NetWare), AppleTalk (on Apple networks) and NetBEUI (as used in Windows). The window below shows some of the protocol options on a network connection.

At one time only TCP/IP could be transported over the Internet, but this is changing a protocols such as PPP allow other protocols to be transported from one network to another over any transport mechanism (see additional material).

Slide 2 [Bits ,frames, and so on]
As we have previously seen we refer to a protocol data units in different ways, depending on where they fit in the OSI model. At the network layer we use the term data packets, and at the transport layer they are referred to as data segments. The transport layer basically creates a data stream from application to another. It is the function of the layered architecture to hide the complexity of the network from the data stream. As far as the applications actually know they are connecting to each other, as if they were running on the same host.

Slide 3 [Functions of the Transport Layer]
We have seen that the network layer is only responsible for two main things: routing the data, and network addressing. The transport layer has many responsibilities, such as:

Synchronization and acknowledgement. Initially, when the transmitter makes contact with the receiver it makes a unique connection. The transmitter thus knows that the receiver is on-line, and willing to receive data.

Acknowledgements and retransmissions. This allows the receiver to send back acknowledge-ments which tell the transmitter that the data segments have been received correctly. If no acknowledgements have been received, the transmitter can either resend the data, or can assume that the receiver has crashed and that the connection is to be terminated.

Flow control. This allows the receiver to tell the transmitter that it cannot receive any more data at present. This typically happens when the receiver has filled-up its receiving buffer.

Windowing. This is where the transmitter and the receiver agree on a window size when the connection is initially made. The window then defines the number of data segments that can be sent before the transmitter must wait for an acknowledgement from the receiver.

Multiple connections onto a single data stream. The transport layer takes data from one or more applications; it then marks them with a unique connection number and segment number. At the receiver these can be demultiplexed to the correct application program.

Reordering of data segments. All the data segments that are transmitted are marked with a sequence number. Thus if any are delivered in the incorrect order, or if any of them are missing, the receiver can easily reorder them or discard segments if one or more are missing.

For the rest of this lecture, see the PDF version.