1. Domain Name System

• Servers communicate using ip addresses.

• However, its easier for us to remember the host names instead.

• This is why DNS server helps translating the host names to ip addresses.

• Two ways to identify the host are:

  • Hostname : url
  • IP address : 32 bits segemented to 4 parts.

• Domain Name System helps translating the host name into an ip address.

• Sometimes the ip address can be multiple of them.

• Canonical names are the official names of the host name.

  • Accessing through
    

DNS: Resource Records (RR)

• type A: hostname to IP address
• type NS: domain to hostname
• type CNAME: alias name to domain (real, canonical) name
• type MX: for mail server

Distributed, Hierarchical Database

• Goes down each level.
  
• Root Server:
• Top Level Domains: .com, .org, .net, .edu, .sg, .uk, .jp...
• Authoritative Server: Organization's own DNS server(s).

Local DNS Server

• When host makes a DNS query, the query is sent to its local DNS server.

DNS Caching

• Local DNS server caches the query (name-to-address) information in cache.
• has expiracy date (Time To Leave (TTL)), in case ip addresses change, or other infos change.
• DNS runs over UDP.
  • Packet loss does not happen usually, because the request is sent to local DNS server.
  • If it is lost, can query again.
  • Browser sends multiple DNS query back-to-back, in case one query is lost, other query can get in.
  • TCP is slower, because has to make a TCP connection first before the packets are sent.

DNS Name Resolution

• The above is an iterative query.
• local DNS server makes query to root, TLD, and authoritative DNS server, then stores the information.
  
• Above is a recursive query.
• iterative query is used more commonly than recursive query.

2. Socket Programming

Process

• program running within a host.
• Within the same host, two processes communicate using Inter-Process Communication (IPC).
• Processes in different hosts communicate by exchanging messages.

Addressing Processes

• IP address is 32-bit integer (e.g. 999.999.99.99)
• IP address is not sufficient to identify a process running in a host.
• IP address can identify a host, but does not know which process to pass the packets to, because there could be multiple processes.
• A process is defined by (IP address and Port Number).
• Port number is 16-bit integer.

Socket

• Application interface between application processes and transport layer. (WHY?)
• Could think of it as a set of APIs.

Socket Programming

• TCP socket and UDP socket.

UDP Socket

• Has no connection between the server and the client.
• Client sends each packet that contains the IP addresses and port number of the destination.
  

UDP Socket Process

• Datagram is a name of packet for UDP.

Example: UDP Socket Server

• AF_INET is a specific version if IP address. Will be covered in future lectures.
• Binding to '' means to bind to the localhost.
• recvfrom() inside the paranthesis is the byte size of the buffer.

Example: UDP Socket Client

• mesasge.encode() is to change string to bytes.

TCP Socket

• TCP connection set when client creates a socket.
• Server TCP creates a new socket for server process to communicate with the client.
  • This allows the server to communicate with multiple clients individually.
    
    

Example: TCP Server Socket

• serverSocket.list() waits for the client to request for TCP connection.
• accept() accepts the TCP connection, returning new socket to communicate with client socket.

Example: TCP Client Socket

• connect() to send TCP connection request to the server socket.
• Then is welcome socket necessary to create additional sockets for individual TCP connection requests?

TCP vs UDP Socket Programming

• For UDP, we need to attach both IP address of the server and the port number for every request sent, because there is no persistent connection.
  • When a client application wants to send a UDP datagram, it always needs to specify the destination IP address of the server and the destination port number to the operating system's networking stack.
  • This is because each UDP datagram is an independent unit of data, and the OS needs to know where to send it. Since there's no pre-established connection, every single datagram needs its full addressing information to reach the intended recipient and application process.
• For TCP however, we do not need to add the server name and port number because it uses persistent connection.
  • When a client application initiates a TCP connection (e.g., using connect() in a socket programming API), it provides the destination IP address of the server and the destination port number once to the operating system.