Reading Notes on "using TCP/IP for Internet connection"

1. TCP/IP uses a 32-bit binary address as a general machine identifier, known as an IP address. The identifier can be divided into two parts: the prefix identifies the network to which the computer is connected, the rest provides a unique identifier for the computer on the network. The original IP address addressing method is called the classification method. Each prefix is assigned as one of the three basic classes. The leading bit defines the class to which it belongs, and the capacity of each type is different. The classification method provides 127 networks connected to millions of hosts, thousands connected to thousands of hosts, and millions of networks with a maximum of 254 hosts. For ease of understanding, the address is read and written in dotted-decimal notation. The values in four octal groups are written in decimal numbers, separated by decimal points.

2. By broadcasting ARP requests, one machine can use ARP to find the hardware address of the other machine. This request contains the IP address of a host. You need to find the hardware address of the host. All machines on the network will receive ARP requests. If the request matches the IP address of a machine, the machine sends a response containing the required hardware address. The response is sent directly to another machine rather than broadcast. Because the request sent by A to B also contains the binding of A's IP address to the physical address.

3. To successfully use RARP, the network must contain at least one RARP server. The PARP server responds to the request by entering the target Protocol address field, changing the packet type from "request" to "response", and then sending the response back to the requesting machine. The requesting machine receives responses from all RARP servers, but it only needs the first one.

4. When will the data part be reorganized? After a network is used or after the part arrives at the target host? In TCP/IP Internet, once the datagram is sliced, each piece is transmitted on the Internet as an independent datagram until it reaches the target host.

5. Partition control. The target host identifies the datagram by the ID field of the datagram chip and the source site address. The computer sending IP datagram must generate a unique value for each datagram as the value of the identification field. One method used by the IP software is to report a Global Counter in the primary storage. Each time a new datagram is generated, the counter is incremented by 1 and the value is assigned to the ID field of the datagram.

6. TCP uses the connection instead of the protocol port as the basic abstract concept. The connection is identified by a pair of endpoints. Because TCP uses two endpoints to identify connections, a TCP port number on a machine can be shared by multiple connections. From the programmer's point of view, connection abstraction is very important. This means that programmers can design programs that provide multiple connection services at the same time without setting up their own local port numbers for each connection. TCP is a connection-oriented protocol, which is different from UDP. It requires two endpoints to agree to participate in communication. This means that before TCP starts Internet communication, the applications at both ends of the connection must establish a connection.