Network Addresses (URL, Port, IP, MAC)

Context: FIT1047_MOC · one address type per layer — the table that organises half the networks exam · resolution between them in Address Resolution (DNS and ARP)

Quick Revision

• 🎯 Objective: URL (application) ➔ port (transport) ➔ IP (network) ➔ MAC (data link) — identity at each layer; an address identifies but does NOT route by itself.
• ⚡ Key Constraint: IPv4 structure — $130.194.66.43/26$ means first 26 bits = network+subnet, last 6 = host ⟹ same-subnet test by prefix comparison.

⚖️ Core Decision Matrix — one address per layer

Layer	Address	Example	Identifies
Application	URL	http://www.stanford.edu	a resource
Transport	port	$80$ (HTTP), $25$ (SMTP)	the application on a host
Network	IP	$\mathrm{130.194.66.43}$	a device across networks
Data Link	MAC	00:23:ae:e7:52:85	an interface within a LAN

📝 Core

IPv4

• 32 bits, dotted decimal ➔ each dot-group = one byte: $\mathrm{130.194.66.43}=10000010110000100100001000101011$.
• Hierarchy ➔ network (e.g. Monash) | subnet (a LAN) | host — the split point given by the subnet mask: $/26$ ≡ $\mathrm{255.255.255.192}$ (26 prefix bits, $2^6-2$-ish hosts).
• Why hierarchy ➔ a device can see immediately whether a destination shares its subnet (deliver directly) or not (send to the router).
• Exhaustion ➔ $2^{32}\approx 4.3$ billion, roughly half practically usable — new IPv4 blocks are gone.

IPv6

• 128 bits ➔ $2^{128}\approx 3.4\times 10^{38}$ addresses — not $4\times$ IPv4 but astronomically more.
• Deliberately “excessive” ➔ the space buys clean hierarchy: IANA+RIR $/23$–$/32$ → ISP (9 bits) → organisation (16 bits, $/48$) → subnet (16 bits, $/64$) → interface ID (64 bits) ⟹ every organisation gets $>65,000$ LANs, assignment stays simple.

Ports (application addressing)

• Why ➔ several browser windows share one IP; the port pairs each response with its connection.
• Convention ➔ server ports fixed ($80$ HTTP, $25$ SMTP), client ports random, one per connection ($57017,57018,\dots$) — a connection is the 4-tuple (src IP:port, dst IP:port).

⚠️ Common Mistakes

• 💡 Address ≠ route ➔ knowing $\mathrm{216.58.220.99}$ says nothing about how to reach it — that’s routing/resolution’s job.
• 💡 $/26$ arithmetic ➔ mask counts PREFIX bits; hosts get $32-26=6$ bits — reading it backwards wrecks subnet questions.
• 💡 MAC is LAN-scoped ➔ frames use MACs only within one subnet; crossing a router swaps frames (and MACs), never the IP.

🧠 Active Recall

For $130.194.66.43/26$: give the binary, the subnet mask in dotted decimal, and how a peer decides if it's in the same subnet.

Answer

• Short answer: Mask $=\mathrm{255.255.255.192}$; compare the first 26 bits of both addresses — equal prefixes ⟹ same subnet ⟹ direct delivery, else ⟹ router.
• Why: Prefix routing ➔ hierarchical addresses make the local/remote decision a single masked comparison.

Two browser tabs both fetch from port 80 of the same server — why don't the responses mix?

Answer

• Short answer: Each tab’s connection uses a distinct random client port; the 4-tuple disambiguates.
• Why: Transport-layer identity ➔ ports address applications/connections, IP only addresses the machine.