GCE A/L General Information Technology common mistakes
GCE A/L General Information Technology · Understand the traps and how to avoid them.
When converting between binary, octal, decimal and hexadecimal, the candidate converts digit groups left-to-right and pads or groups from the wrong end, producing answers that are reversed or off by a factor of 8 or 16.
Grouping rules (three bits per octal digit, four bits per hexadecimal digit) are memorised as phrases but not tied to place value, so the candidate does not check whether the least significant group is on the right. Working quickly under time pressure, they start grouping at the leftmost bit because that is how they read the number.
What to do instead
Before every conversion, write the place-value headings (for example 16^2 16^1 16^0) above the digits and always pad with leading zeros on the left so the final group is complete. Practise a fixed drill of ten mixed conversions per session and always verify by converting the result back to the original base; only move on when the round trip is correct without notes.
The candidate calculates storage or transmission sizes by mixing decimal (kB, MB) and binary (KiB, MiB) prefixes, or forgets to divide by 8 when converting between bits and bytes, so answers differ from the expected value by a factor of 8 or about 2.4 percent per prefix step.
Everyday usage (advertised disk sizes, broadband speeds) uses decimal prefixes while operating systems often report binary ones, and the candidate has absorbed both conventions without noticing the conflict. Bit-versus-byte conversion is treated as a definition to recall rather than an operation to apply.
What to do instead
Adopt one working convention for the exam and state it in the answer, for example 'using 1 KB = 1024 bytes'. For every calculation, write the unit chain explicitly (Mbps to bits per second, divide by 8 for bytes per second, multiply by time in seconds) and cancel units on paper before computing. Do timed practice with both conventions so the difference is visible.
The candidate answers questions about the OSI or TCP/IP model by listing layer names in order but cannot place a device, protocol or address type at the correct layer, for example putting a switch at layer 3, a router at layer 2, or MAC addressing above the data link layer.
Layer names are learned as a sequence to recite rather than as a set of responsibilities, so the candidate has no criterion for deciding where a new item belongs. Encapsulation is understood as a diagram, not as a process of adding headers at each layer.
What to do instead
For each layer, write one sentence describing its job and one example each of a device, protocol and address or data unit. Then classify unfamiliar items by asking which job they perform, not by memorising lists. Draw the encapsulation of a single HTTP request down to the frame and back up, labelling where each header is added and removed.
The candidate confuses the roles of IP addresses, MAC addresses, ports and domain names, and gives answers in which a MAC address is said to be assigned by DHCP, a port number identifies a host, or DNS resolves a MAC address.
These identifiers appear together in networking diagrams and configuration screens, so the candidate treats them as interchangeable labels for 'an address'. The distinction between logical, physical and service-level addressing is never made explicit in their notes.
What to do instead
Build a four-row comparison table (identifier, what it names, who assigns it, whether it changes when the device moves networks, which layer it belongs to) and fill it from memory, then check it. For each identifier, write a one-line scenario showing it in use, such as a laptop keeping its MAC but receiving a new IP on a different network.
The candidate writes SQL or relational-algebra answers that select the right columns but ignore the join condition or the grouping rule, producing duplicated rows, a Cartesian product, or an aggregate computed over the wrong set of rows.
Simple single-table SELECT statements are practised until fluent, and multi-table queries are then attempted by adding tables to the FROM clause without reasoning about how rows should be matched. GROUP BY is treated as a formatting instruction rather than a change to the unit of output.
What to do instead
Before writing any query, state in words what one output row should represent, then list the tables needed and the condition that matches their rows. For aggregates, name the group first and check that every non-aggregated selected column appears in the GROUP BY. Run small hand-worked examples on paper with three or four rows to confirm the result before trusting the query.
The candidate describes normalisation by repeating the definition of 1NF, 2NF and 3NF but cannot take a given table and decompose it, or produces a decomposition that loses data or cannot be recombined without spurious rows.
The normal forms are memorised as verbal definitions ('no repeating groups', 'no partial dependency') without linking them to functional dependencies, which are the actual working tool. Practice is spent reading about normalisation rather than applying it to sample tables.
What to do instead
For each practice table, first list all functional dependencies in the form X to Y, then check each one against the normal-form rules in order. Decompose one dependency at a time and after each step verify that a natural join of the resulting tables reproduces the original rows exactly. Keep a worked example of a table taken from unnormalised to 3NF in your own handwriting.
The candidate traces an algorithm or writes pseudocode that is logically close but uses the wrong loop boundary or initial value, so it is off by one at the first or last element, or a search fails when the target is the final item.
The candidate reads code for its intent rather than executing it step by step, and assumes that a loop 'goes through the array' without checking whether the index reaches the last valid position. Zero-based indexing is confused with one-based counting from mathematics.
What to do instead
Trace every algorithm on paper with a small example, writing the value of each variable at each iteration in a table, and include the boundary cases: empty input, single element, target at the first position, target at the last position, target absent. State the loop invariant in one sentence before writing the loop, and check the initial value and the termination condition against it.
The candidate answers questions about the processor by naming components (ALU, control unit, registers, cache) but cannot explain the sequence of events in the fetch-decode-execute cycle, or states that cache is larger and slower than main memory, or that RAM is non-volatile.
Components are learned as a labelled list from a diagram, and the relationships between them (which unit controls which, what is stored where, relative speed and volatility) are never tested. The candidate recognises the vocabulary and assumes understanding.
What to do instead
Write the fetch-decode-execute cycle as a numbered sequence naming the registers involved at each step (program counter, memory address register, memory data register, instruction register, accumulator) and the role of the control unit. Then build a memory hierarchy table with capacity, speed, volatility and cost per bit for registers, cache, RAM, SSD and magnetic disk, and reproduce it from memory.
The candidate answers ethics and security scenarios with the name of a law or the name of an attack, without connecting the specific facts in the scenario to the definition, so a phishing scenario is labelled as hacking, or a confidentiality breach is described as an integrity failure.
Definitions are memorised as short labels and matched to scenario keywords rather than to the underlying properties being violated. The candidate does not separate the three security goals (confidentiality, integrity, availability) and so cannot classify a scenario by which goal is harmed.
What to do instead
For each scenario, first identify the asset, the actor, the action and the harm, then decide which security goal is affected and only then name the category. Write a one-line definition for each attack type that includes the mechanism (for example deception of the user into revealing credentials) rather than just an example. Practise by rewriting given scenarios so that each illustrates a different category.
The candidate treats e-government and e-commerce models as a list of abbreviations (G2C, G2B, B2C, C2C) and classifies a scenario by the technology used rather than by who is transacting with whom, so a government portal serving businesses is called G2C because a citizen logs in.
The abbreviations are learned without the rule that the letters denote the two parties involved, and real services often involve intermediaries, so the candidate falls back on surface features such as the device or the login method.
What to do instead
For every model, write the two parties explicitly and one genuine local example, then classify unfamiliar scenarios by naming the two parties first and only afterwards choosing the label. Include borderline cases, such as a government service delivered through a bank, and decide which party is the primary provider and which is the recipient.