TCP UDP and IP

TCP/IP

TCP/IP is essential if you want to use the Internet. TCP/IP stands for ‘Transmission Control Protocol / Internet Protocol’. TCP/IP (usually called TCP) is the standard method of sending data on the Internet. It is based on data packets that have a set format, including to and from addresses, similar to a letter. If you want to use the Internet or WinGate, it needs to be installed on every machine on your LAN. Actually TCP and IP are different protocols, but they are so tied up that they are usually referred to in this way.

Packet

A data packet is a like a ‘mail parcel’. Think of a package that gets sent in the post. There are a few things that you have to have, requirements. There has to be a name and address for the recipient, a return address, there have to be stamps, and of course the envelope or wrapping paper. But, what you put in the parcel is up to you. You can send (with in reason) anything that will be accepted by the post office. A data packet is very similar to this. You have to supply certain ‘Wrappers’ such as ‘to’ and ‘from’ fields, but what is sent as the payload is up to you. There are different types of packets used on the internet and other networks, but all of them use this idea of a parcel of data.

IP

IP stands for Internet Protocol. This is the method used on the internet (and on many LANs) to communicate. IP is a system of datagram packets. IP is not usually dealt with directly, this is the job of TCP. IP gets datagrams from point A to point B. TCP sends IP a datagram, and a destination. It assembles and sends a packet with information from the source (eg TCP) and a checksum that indicates the integrity of the packet. IP doesn’t care about what is in the datagram. In fact it does not care if the packet it sends even gets there, and when IP receives a packet, if it has be garbled, IP throws it away! It is up to the protocol using IP to arrange for the packet to be resent if required.

IP Number / IP Address

An IP number is the way IP distinguishes computers (or more specifically Interfaces) that exist on the same network. On the Internet you simply can not have two computers sharing an IP, as this creates havoc when trying to send data to the correct location. All computers that are ‘on’ the Internet (or LAN) need discrete IPs. There are different types of IP.

You have probably seen addresses in the form 128.211.23.45. This is a 32-bit number separated in to 4 8 bit parts. The four parts are similar to a mailing address, except the detail is the other way round. The first number of the IP is the most general, the last is the most specific. Since each computer on the Internet needs a different IP, there has to be some way of dishing out the IPs so that large companies and organizations have one for all their machines, while smaller organizations have some to go around as well. Since there are a small number of Large organizations and a large number of small organizations, ranges of IPs can be allocated accordingly.

In an IP number there are 2 parts, the network and the host identifiers. There are three ways the IPs can be split in to 2 parts.

Class A nnn.hhh.hhh.hhh

Class B nnn.nnn.hhh.hhh

Class C nnn.nnn.nnn.hhh

where n’s=network identifier, h’s=host identifier

A huge company with very complex internal networks may be allocated a class A address range such as 105.*.*.* . Only the range 1.*.*.* to 126.*.*.* are available for A class addresses. There are very few A class addresses, and no more are to be allocated, mainly because no-one has 16 million computers on their network!

B class addresses however are common for Large companies, allowing a range of around 65000 IPs. Microsoft and IBM probably have several each. When an B class IP address is allocated, (say 165.103.*.*), the first two numbers identify that companies network. The company can decide what to do with the next two (*’s in this case mean any number), and give any IP in that range to any computer on their network. B class networks addresses have 128 - 191 as the first number in the IP.

Class C addresses, giving 254 possible addresses (0 and 255 are reserved) are the third type. Here, the first 3 8 bit fields are specified, and the remaining field is allocated by the owner of the address. C class licenses are in the range 192.*.*.* to 223.*.*.*

Networks that are directly connected to the internet are connected to an ISP via some fulltime connection (such as a cable or leased line) and the ISP will inform the network administrator of which IP’s can be used on the network. A router is used to ‘tell computers how to get to a particular IP’. (You may wish to read about Routing )

ISP’s typically have 1-2 C class licenses, providing 250 to 500 IP’s. When you dial up an ISP with a modem, you are Dynamically allocated an IP address. This will be in the range of the C class license that they own.

Private IPs

Private IP numbers are ranges of IP numbers that are ‘Known not to exist’ on the Internet. What this means is that no computer on the Internet will be assigned these addresses. These can safely be used in internal LANs, as they have no direct connection to the Internet. One example of a Private IP range is the 192.168.0.* range that this manual commonly refers to.

The private IP ranges that will not be allocated on the Internet are

10.0.0.0 to 10.255.255.255 Class A

172.16.0.0 to 172.31.255.255 Class B

192.168.0.0 to 192.168.255.255 Class C

Do not choose an IP range that is not on this list. Also note that 0 and 255 are reserved in any class.

Netmask

Network masks are IP filters. They are used in directing or ‘routing’ network traffic. The mask is related to whether you are on an A B or C class network. See Routing for an extended explanation.

localhost

localhost is a special term in TCP/IP. 127.0.0.1 is the localhost (loopback interface) this is a software only interface internal to the stack itself, and is not accessible over any interface. It doesn’t matter what your LAN card IP really is, 127.0.0.1 will always refer to the local machine. This means that this interface can only be accessed from the machine itself. It is like saying “ME” or “I” in reference to yourself. Any one can refer to themselves like this. “I lost my shoe!”, meaning quite clearly that who ever said this is in need of another shoe. However, you couldn’t say that your friend lost his shoe, simply by repeating his phrase, people would think you had lost your shoe. TCP/IP uses localhost in this same way. If a machine wants to talk to itself on a different port it can say “localhost:<port#>”. The TCP stack looks at this, realises it refers to itself, and directs to the correct port, with out sending anything on the network. An instance of this is when you Log on to WinGate with Gatekeeper for the first time (Or in LITE), you connect to localhost:808. That means “This machine, port 808”. Programmers familiar with Objects will realise that this is like referring to the object reference: ‘this’ (C++/Java) or ‘self’ (Pascal).

UDP

User Datagram Protocol is a ‘Connectionless’ protocol. It uses IP to send datagrams in a similar way to TCP, except that like IP, and unlike TCP, UDP does not ensure the packets reach their destination. One important difference between TCP and UDP packets is that UDP packets do not include the sender address. UDP is used in applications where it is not essential for 100% of the packets to arrive. This may sound strange, but often you don’t need all the packets. Think of an image. If you cover the top half of the image then it is hard to understand the whole picture, but if you cover lots of tiny parts of the picture, say with dots from a pen, you have to put a lot of dots before you loose the overall picture. Think of a television. On a windy day, your antenna gets blown around and you have static all over the screen. It doesn’t stop you understanding the story line. It is the same with radio transmission. You can have really bad interference before you cant understand someone talking. Programs that use Video and Audio on the Internet don’t need to worry about every single packet. But you wouldn’t use UDP to transmit a program, because if one single bit was wrong (let alone loosing a whole packet) the file would be useless. It is up to program designers to choose what method is most suitable. While TCP is safer, UDP is often faster and is becoming more common. It especially favored for ‘Streaming’ or Real-time applications. These tend to be A/V programs, allowing conferencing or similar. More recently, internet applications have used both UDP and TCP. TCP is used for the essential or Control data, while UDP is used for data for which losses are acceptable.