The TCP/IP Protocol

THE TCP/IP PROTOCOL

The TCP/IP is a group of protocols developed to interconnect nets of calculators. These protocols have a stratified architecture. We'll try to explain it with an example. Let's suppose to manage E-Mail traffic. First level is the one of the protocol needed to send the message, which identifies the sender, the receiver and the text of the message. Above this level there is TCP/IP, the protocol that controls that data forwarded will reach effectively the target. For this purpose the message is split up into datagrams, and TCP/IP cares for the datagrams to reach correctly the target. Above both of these levels there is still another protocol, IP, which cares for data routing on the Internet. Last, we have a fourth level: the interface (Ethernet, serial port, etc.)

THE TCP LEVEL

TCP or Transfer Control Protocol splits data to be sent into datagrams, and has to reassemble received datagrams, or to ask for re-sending of lost data packets. IP proceeds for routing of datagrams. Each protocol level ignores completely the other, and as a matter of fact each protocol adds headings to datagrams, so that they are unique and impossible to be lost. An example, that we will repeat later, will allow us to understand the mechanism. Let's suppose to send a message to another computer through the Internet. We represent the bytes of this file with bullets:

The protocol controlling the transfer (TCP) splits the file into smaller pieces, by mutual consent with the receiver about the maximum size to be used, and depending also on the net they share. Thus file becomes:

Now, as we said, at the beginning of each datagram (the three bullets) is attached a header. Present technology uses 20 bytes (160 bits) for each header, that we can represent as follows:

BIT0	8\|9	_	16\|17	_
SOURCE PORT			\|	DESTINATION PORT
SEQUENCE NUMBER
ACKNOWLEDGMENT NUMBER
DATA OFFSET	\|	RESERVED	\|	WINDOW
CHECKSUM			\|	URGENT POINTER

Let's explain this list:

SOURCE PORT is an interface that manages the transit of the sent file.
DESTINATION PORT is, obviously, the same on remote computer.
The reason for these two partitions is that each system can have multiple ports, to connect simultaneously more users. Source Port and Destination Port interexchange when receiver sends data back.
SEQUENCE NUMBER specific for each datagram, allows the receiver to reconstruct perfectly the sequence, and to receive a correct file.
How do you realize that the target has received the datagram effectively, and with the correct sequence? The target sends back to the sender an AKNOWLEDGMENT NUMBER, which will form the sequence number of data received from the correspondent.
WINDOW is the field that controls how many data can be forwarded. As the correspondent must retrieve arriving datagrams (which vary depending on many factors), he sends back to the sender, using the field Window, information on how many datagrams he can still handle. As data arrive, field changes continuously. This field is extremely important to inform the sender when to stop forwarding data and when to start again.
CHECKSUM sums up all bytes contained in the datagram, and it must be identical in both forwarded and arrived datagrams. This is another control information: if during tranfer some data are lost, checksums will not match and wrong datagram will be re- transmitted.
URGENT allows one system to ask the other to transmit a given octet. With such a system you can, for example, force the disconnection.

The file, after being split up into smaller pieces, and with the add of a header for each piece, becomes:

LEVEL IP

Now TCP has made up, out of the starting file, a series of datagrams; now these datagrams are passed to the IP (Internet Protocol) instructions. These instructions will be used by datagrams to find their way to the target. For this reason even IP adds a 20 bytes header to each datagram. We show a list of this datagram:

BIT 0			\|	9	16\|17
VERSION	\|	IHL	\|	TYPE OF SERVICE	\|	TOTAL LENGHT
IDENTIFICATION					\|	FLAGS	\|	FRAGMENT OFFSET
TIME OF LIVE			\|	PROTOCOL	\|	HEADER CHECKSUM
SOURCE INTERNET ADDRESS
DESTINATION INTERNET ADDRESS
TCP HEADER, DATA, ......

Let's explain the content.

SOURCE INTERNET ADDRESS is the sender address; it is a 32 bit number (i.e. 130.192.59.20). In the same way as the sender address on an envelope, it informs the target of a file about who is the sender.
DESTINATION INTERNET ADDRESS is another 32 bit number, and is the univocal address of the target. Thus both sender and target are correctly and unmistakably recognized by any computer forwarding datagrams.
PROTOCOL NUMBER identifies the kind of used protocol and determines the correct datagram processing of the receiver.
CHECKSUM allows the receiver of the datagram to control the presence of any corruption of data during forwarding. It is independent from TCP checksum, and is related only to IP header.
FLAGS and FRAGMENT OFFSET fields are used to reassemble datagrams in case they break. It may happen that they are too big for a certain net, and that they crumble trying to pass through.
TIME TO LIVE it's a safety to avoid the looping of a datagram: every time a datagram passes through a system this number decreases progressively to zero, hence datagram is cancelled.

Now, adding IP header, our file becomes:

etc...

INTERFACE LEVEL

Now comes another level, since superimposed to datagrams are control characters, typical to the electric interface (Ethernet, RS232, a TNC in case of radio communication). These are low hardware level characters, and they are processed as such. A graphic description of the group of packets obtained is......

etc. etc.

You can easily see that electric interface control characters are at the beginning and at the end of the datagram.