IP characteristics and binary conversion

IP characteristics

IP lives at OSI Layer 3, the TCP/IP internet layer. A packet is self-contained: data plus enough info to route from source to destination without depending on other packets.

• Connectionless: one-way send, no advance notice to the destination, no status returned to the sender.
• Packets treated independently: each can take a different path.
• Hierarchical addressing: network ID is the street, host ID is the house on that street.
• Best effort: no delivery guarantee. Packets can be misdirected, duplicated, or lost.
• No recovery: IP doesn’t fix corrupted packets; the end systems handle that (upper layers).
• Media-independent: doesn’t care what’s carrying it.

Two flavors exist: IPv4 and IPv6.

The postal analogy: mail three letters and the postal service makes its best attempt, but doesn’t guarantee delivery, the same carrier or route, or arrival order. Letter = data, your return address = source IP, recipient = destination IP, and each letter handled independently = each packet handled independently.

How a company gets a public IP

The allocation chain, top down:

1. ICANN assigns regional blocks to the RIRs (Regional Internet Registries).
2. Each RIR assigns smaller blocks to ISPs in its region.
3. A company requests a public IP from its ISP, which assigns one from its block.
4. The company configures it on the router interface (or the router uses DHCP and the ISP assigns it automatically).

Positional numbering systems

The base matters, but it’s the position of a digit that gives it value. In any base system, each column to the left raises the exponent by one.

Decimal (base 10) has digits 0-9: run out at 9, roll to 10; run out at 99, roll to 100. Columns are powers of 10, so 27398 = (2 x 10⁴) + (7 x 10³) + (3 x 10²) + (9 x 10¹) + (8 x 10⁰).

Binary (base 2) has digits 0 and 1 only, counting 0, 1, 10, 11, 100, 101, 110, 111, 1000… Columns are powers of 2, so 10011 = (1 x 2⁴) + (0 x 2³) + (0 x 2²) + (1 x 2¹) + (1 x 2⁰) = 16 + 2 + 1 = 19.

Contrast with Roman numerals, which use symbol values (I=1, V=5, X=10, L=50, C=100, D=500, M=1000) with add and subtract rules (smaller before larger subtracts, so MCMXCIV = 1000 + 900 + 90 + 4 = 1994). Position plays a role there, but not with base-system logic.

Binary to decimal and back

Computers and network gear work in binary; we type decimal. An IPv4 address is 32 bits, four octets (8 bits each) in dotted-decimal notation: 192.168.10.22 is 11000000.10101000.00001010.00010110. Always convert each octet using all 8 bits, leading zeros included.

Each octet has 8 place values. Memorize them:

128   64   32   16    8    4    2    1
2^7  2^6  2^5  2^4  2^3  2^2  2^1  2^0

The trick: start at 1 on the right and keep doubling leftward.

Binary to decimal

Drop the bits into the chart, add up the columns with a 1.

10111001 is 128 + 32 + 16 + 8 + 1 = 185.

Decimal to binary

Work left to right, subtracting as you go: does the place value fit into what’s left? For 147:

• 128 fits: write 1, remainder 19
• 64 and 32 too big: 0, 0
• 16 fits: 1, remainder 3
• 8 and 4 too big: 0, 0
• 2 fits: 1, remainder 1
• 1 fits: 1, remainder 0

So 147 = 10010011.

This chart matters beyond conversion: the place values are also the block sizes networks increase by when subnetting outside classful boundaries. Memorizing 128-64-32-16-8-4-2-1 cold pays off later.