Question 2 of 6: Fundamental Concepts and Mechanisms

Question 2: Fundamental Concepts and Mechanisms

Question text not reproduced: the examination questions are © Engineers and Geoscientists BC. Open the official past paper (linked at the top of this page) to read the question, then follow the worked solution below.

(a) Middleware limitations, and CORBA vs. Java RMI vs. Web services. Middleware's convenience comes at a cost: it adds an extra abstraction/translation layer (marshalling, stub/skeleton generation) that costs latency and CPU relative to hand-written networking code; it can hide the reality of partial failure so thoroughly that programmers write remote calls as if they were local, then are surprised by the different failure modes (network partitions, partial completion) that local calls never have; heavyweight middleware (classical CORBA) has significant learning curve and deployment complexity; and different middleware products are frequently not interoperable with one another, creating platform/vendor lock-in.

(b) Three limitations of IPv4. 1. Address exhaustion. IPv4's 32-bit address space provides only about 4.3 billion addresses, long since insufficient for the number of Internet-connected devices, forcing widespread use of NAT and address-conservation techniques as stop-gaps rather than every device having a unique global address. 2. No built-in security. The base IPv4 header carries no authentication or encryption; confidentiality and integrity must be added by a separate layer (IPsec, or higher-layer TLS), so a plain IPv4 packet is trivially spoofable/eavesdroppable. 3. Weak support for QoS and modern traffic patterns. IPv4's header has no adequate flow-labelling for real-time/multimedia traffic, and its variable-length options field is rarely hardware-accelerated, making fine-grained quality-of-service and fast router processing harder than in IPv6's fixed, streamlined header.

(c) Network Address Translation. A NAT-capable border router/gateway maintains a translation table mapping each internal (private, RFC 1918) address and source port to a single public IP address and a (usually router-chosen) public port. As a packet leaves the private network, the router rewrites the source address/port to the public mapping and records the mapping; on the return packet it looks up the destination address/port in the table and rewrites it back to the original internal address/port before forwarding it inward. This lets many internal hosts share one (or a small pool of) public IPv4 address, directly mitigating the address-exhaustion problem from part (b), and as a side effect hides the internal network's topology and addressing from the outside (a modest security benefit), at the cost of breaking pure end-to-end connectivity for protocols that expect inbound connections to be initiated from outside (requiring port forwarding, STUN/TURN, or application-layer gateways for VoIP/P2P/some FTP modes).

(d) The five TCP/IP layers.