📚 Year 12 OCR Computer Science: High-Frequency Topics and Common Mistakes Analysis | Year 12 OCR 计算机科学：高频考点与易错题分析
Mastering Year 12 OCR Computer Science requires more than just memorising definitions — it demands a deep understanding of how the theoretical concepts from Paper 1 (Computer Systems) connect with the practical problem-solving and programming skills assessed in Paper 2 (Algorithms and Programming). In this revision guide, we will walk through the most frequently examined topics, spotlight the common pitfalls that cost students marks, and illustrate each point with precise, bilingual explanations. Whether you are revising CPU architecture, data structures, Boolean algebra or network protocols, the following sections will help you sharpen your exam technique and build lasting confidence.
掌握 Year 12 OCR 计算机科学需要的不仅仅是背诵定义——它要求你深刻理解试卷一（计算机系统）中的理论概念如何与试卷二（算法与编程）所考查的实际问题解决和编程技能联系起来。在这篇复习指南中，我们将梳理考查频率最高的主题，点明令学生失分的常见陷阱，并用精确的中英双语解释对每个要点加以说明。无论你正在复习 CPU 架构、数据结构、布尔代数还是网络协议，以下各节都将帮助你打磨应考技巧，建立持久的自信。
1. CPU Architecture and the Fetch–Decode–Execute Cycle | CPU 架构与取指–译码–执行周期
In almost every OCR AS paper, a question appears requiring you to describe the role of the program counter (PC), memory address register (MAR), memory data register (MDR), current instruction register (CIR) and accumulator (ACC) during the fetch–decode–execute cycle. A frequent mistake is confusing the MAR (which holds the address of the memory location to be accessed) with the MDR (which holds the actual data or instruction that has been read from or is to be written to memory). Remember: MAR points to where the data lives; MDR temporarily stores that data. Another high-risk area is failing to specify that the PC is incremented before the instruction is fully executed, so that it already points to the next instruction while the current one is being decoded.
几乎在每一份 OCR AS 试卷中，都会出现要求你描述取指–译码–执行周期中程序计数器（PC）、存储器地址寄存器（MAR）、存储器数据寄存器（MDR）、当前指令寄存器（CIR）和累加器（ACC）的作用。一个常见的错误是将 MAR（存放要访问的内存单元地址）与 MDR（存放从内存读出或即将写入内存的实际数据或指令）混淆。记住：MAR 指向数据所在的位置；MDR 暂存该数据。另一个易踩坑点是未能说明 PC 是在指令完全执行之前递增的，这样在执行当前指令时 PC 已经指向下一条指令。
Also, students often lose marks by not linking the registers to the buses. The address bus carries the value from the MAR to main memory; the data bus transmits the contents of the MDR in either direction; and the control bus sends read/write signals. When asked to trace a simple instruction like LDA #5, explicitly state which bus is used and at which stage. In the decode step, the CIR splits the instruction into opcode and operand — many candidates forget to name the opcode and operand fields, simply writing 'the instruction is decoded' without detail, which does not gain full marks on a 4–6 mark question.
此外，学生常常因为没有将寄存器与总线联系起来而丢分。地址总线将 MAR 中的值传送到主存；数据总线双向传输 MDR 的内容；控制总线则发送读/写信号。当被要求追踪一条类似 LDA #5 的简单指令时，一定要明确说明在哪个阶段使用了哪条总线。在译码步骤中，CIR 将指令拆分为操作码和操作数——许多考生会忘记说出操作码和操作数字段，只是写“指令被译码”而不提供细节，这在4至6分的题目中无法获得满分。
2. Pipelining and the Three Classic Hazards | 流水线及其三种经典冒险
Pipelining is a favourite topic for higher-tier questions. You must be able to explain how overlapping the fetch, decode and execute stages improves throughput, and you must be able to identify and describe the three hazards: data, structural and control (branch) hazards. A typical error is to describe a data hazard as 'two instructions trying to access the same register at the same time' without clarifying that it is specifically a Read-After-Write (RAW) dependency where a later instruction needs a value that has not yet been written back by an earlier instruction. Similarly, when explaining a control hazard after a branch instruction, many candidates say 'the pipeline is stalled' but omit the crucial detail that the fetched instruction may need to be flushed because the branch outcome is unknown until the execute stage.
流水线是高分题目常考的主题。你必须能够解释如何通过重叠取指、译码和执行阶段来提高吞吐量，同时必须能够识别并描述三种冒险：数据冒险、结构冒险和控制（分支）冒险。一个典型错误是将数据冒险描述为“两条指令同时尝试访问同一个寄存器”，却没有阐明这具体是一种读后写（RAW）依赖关系，即后一条指令需要的值还未被前一条指令写回。类似地，在解释分支指令后的控制冒险时，许多考生只会说“流水线停顿”，却遗漏了关键细节：由于在译码阶段尚不知道分支结果，取出的指令可能需要被清除（flush）。
Be precise with terminology. Structural hazards occur when the hardware cannot support all the instructions that want to execute simultaneously — for example, a single memory port being needed for both instruction fetch and data access in the same clock cycle. Students often mislabel a structural hazard as a data hazard. Using a diagram or timing table in your answer can help, but even in prose, distinguishing 'resource conflict' from 'data dependency' is essential to secure the marks.
术语要准确。结构冒险发生在硬件无法支持所有想要同时执行的指令时——例如，单个内存端口在同一时钟周期内既需要用于取指又需要用于数据访问。学生常常将结构冒险错误地标注为数据冒险。在答案中使用图示或时序表会有所帮助，但即使在纯文字中，区分“资源冲突”和“数据依赖”对于得分至关重要。
3. Memory Management: Paging, Segmentation and Virtual Memory | 内存管理：分页、分段与虚拟内存
Virtual memory is examined almost every year, and the most common mistake is stating that it 'increases the amount of RAM' or 'uses hard disk space as extra RAM' without the caveat that it allows the execution of programs that are larger than physical memory by moving inactive pages to the hard disk. You must mention pages, page tables, and the role of the memory management unit (MMU) in translating virtual addresses to physical addresses. When a required page is not in RAM, a page fault occurs, triggering a swap from secondary storage — a process that causes disk thrashing if it happens too frequently.
虚拟内存几乎每年都考，最常见的错误是说它“增加了RAM的数量”或“把硬盘空间当作额外的RAM来用”，而没有说明通过将不活动页面移到硬盘，它能够运行比物理内存更大的程序。你必须提及页面、页表以及内存管理单元（MMU）在将虚拟地址转换为物理地址中的作用。当所需的页面不在RAM中时，会发生缺页中断，触发从辅助存储器的换入操作——如果这一过程过于频繁，就会导致系统颠簸。
Differentiating paging and segmentation is another challenge. Paging divides memory into fixed-size blocks, while segmentation uses variable-sized blocks that reflect the logical structure of a program (code segment, data segment, stack segment). A common error is to imply that segmentation eliminates external fragmentation; actually it can suffer from external fragmentation, whereas paging suffers from internal fragmentation. Candidates often mix up the two, so be ready to compare them with a small table in your own mind: paging — fixed size, no external fragmentation, invisible to programmer; segmentation — variable size, possible external fragmentation, visible to programmer. That clarity will earn you the top marks in comparison questions.
区分分页和分段是另一个难点。分页将内存划分为固定大小的块，而分段则使用反映程序逻辑结构的可变大小块（代码段、数据段、栈段）。一个常见错误是暗示分段可以消除外部碎片；实际上，分段可能遭受外部碎片，而分页则遭受内部碎片。考生常常将两者混淆，因此在脑海中准备一张小型对比表将大有裨益：分页——固定大小、无外部碎片、对程序员不可见；分段——可变大小、可能有外部碎片、对程序员可见。这种清晰度会让你在对比题中拿到最高分。
4. Interrupts and the Interrupt Service Routine | 中断与中断服务程序
Interrupt questions often look straightforward but carry subtlety. Many students fail to describe the complete sequence: an interrupt signal is raised, the current fetch–decode–execute cycle completes, the processor saves the contents of its registers (including PC) onto the stack, identifies the source of the interrupt by polling or using a vectored interrupt mechanism, and then loads the address of the appropriate Interrupt Service Routine (ISR) into the PC. The error here is to say that the processor 'immediately' jumps to the ISR — in reality, the current FDE cycle finishes first to avoid corrupting the state. Furthermore, after the ISR executes, the saved register contents are restored from the stack so that the original program can resume as if nothing happened.
中断的相关题目看似简单，实则暗含细节。许多学生未能描述完整的顺序：中断信号被提出，当前取指–译码–执行周期完成，处理器将寄存器（包括PC）的内容保存到栈中，通过轮询或向量中断机制识别中断源，然后将相应中断服务程序（ISR）的地址加载到PC。这里的错误在于说处理器“立即”跳转到ISR——实际上，当前FDE周期会先完成，以免破坏状态。此外，在 ISR 执行完毕后，保存的寄存器内容会从栈中恢复，这样原程序就能无缝继续。
Another nuance: distinguishing between hardware interrupts (generated by devices like a printer or keyboard) and software interrupts (exceptions such as division by zero). Candidates sometimes treat them identically, but software interrupts are synchronous (caused by an instruction), whereas hardware interrupts are asynchronous. A precise description of the role of the interrupt priority register can also give your answer an edge, especially if you link it to the idea of nesting interrupts when a higher-priority interrupt pre-empts a lower-priority one.
另一个细微之处：区分硬件中断（由打印机、键盘等设备产生）和软件中断（如除以零等异常）。考生有时将它们等同对待，但软件中断是同步的（由指令引起），而硬件中断是异步的。若在答案中精确描述中断优先级寄存器的作用，还能让你的回答更胜一筹，尤其是当你能将其与高优先级中断抢占低优先级中断时的中断嵌套概念联系起来时。
5. Scheduling Algorithms and Operating System Roles | 调度算法与操作系统功能
When asked to compare scheduling algorithms — Round Robin, First Come First Served, Shortest Job First, Shortest Remaining Time, Multi-level Feedback Queues — do not just list definitions. Examiners want evaluation: throughput, waiting time, response time, and the risk of starvation. A classic mistake is stating that Round Robin guarantees low waiting time; it actually provides low response time but can increase average waiting time if the time quantum is poorly chosen. Always tie your explanation to the concept of a time quantum and pre-emption. For instance, a very small quantum causes frequent context switches, wasting CPU cycles, while a too large quantum degenerates into FCFS.
当被要求比较调度算法时——轮转调度、先来先服务、最短作业优先、最短剩余时间优先、多级反馈队列——不要只列出定义。考官希望看到评估：吞吐量、等待时间、响应时间以及饥饿的风险。一个经典错误是说轮转调度能保证低等待时间；它实际上提供的是低响应时间，但如果时间量选择不当，反而会增加平均等待时间。解释时一定要联系时间量和抢占的概念。例如，时间量过小会导致频繁的上下文切换，浪费CPU周期；而时间量过大则会退化为先来先服务。
Beyond scheduling, the operating system's roles — memory management, resource management, file management, I/O management, user interface, security — appear in many 6‑mark essays. The common pitfall is to describe each role in isolation without connecting them. For example, explain how the memory manager and scheduler work together when a process is swapped out: the scheduler moves the process to the blocked/suspend state, and the memory manager frees its frames. Demonstrating such interaction signals higher-level understanding.
除了调度，操作系统的角色——内存管理、资源管理、文件管理、I/O管理、用户界面、安全——经常出现在6分的论述题中。常见的陷阱是孤立地描述每个角色而不建立它们之间的联系。例如，解释当进程被换出时内存管理器和调度器如何协同工作：调度器将进程移至阻塞/挂起状态，内存管理器释放其帧。展示这种交互能体现出更深层次的理解。
6. Data Structures: Arrays and Linked Lists Time Complexity | 数据结构：数组与链表的操作时间复杂度
OCR Paper 2 frequently asks you to justify the choice of an array over a linked list for a given scenario, and vice versa, using Big O notation. A high-frequency error is to claim that inserting at the end of a dynamic array is always O(1). While it is amortised O(1), occasional resizing makes the worst case O(n). For a linked list, insertion or deletion at the head is O(1), but accessing the k-th element is O(n) because you must traverse from the head. With arrays, random access is O(1) thanks to index arithmetic. Students often confuse the cost of searching in a sorted array (which is O(log n) with binary search) with the cost of accessing by index (O(1)). Be explicit about which operation is being evaluated.
OCR试卷二经常要求你针对给定场景，使用大O表示法论证选择数组而非链表的理由，反之亦然。一个高频错误是声称在动态数组末尾插入元素总是O(1)。尽管其均摊复杂度是O(1)，但偶尔的扩容会使最坏情况变为O(n)。对于链表，在头部插入或删除是O(1)，但访问第k个元素需要O(n)，因为必须从头遍历。对于数组，借助索引运算的随机访问是O(1)。学生常常混淆在已排序数组中进行搜索的代价（通过二分搜索为O(log n)）与按索引访问的代价（O(1)）。务必明确说明评估的是哪一种操作。
When drawing memory representations, another trap is forgetting to show the 'next' pointer field in a linked list node, or drawing an array with elements of inconsistent data types. Lists, arrays and records (structures) each have distinct representations. A record stores a fixed collection of fields of possibly different types, accessed by field name, not index. Don't blur these lines. In pseudocode, updating a pointer incorrectly (e.g., losing the reference to the head when inserting into a linked list) is a frequent algorithm-writing mistake. Always sketch a diagram to guide your code.
在绘制内存表示图时，另一个陷阱是忘记画出链表节点中的“next”指针字段，或者画出具有不一致数据类型的数组。列表、数组和记录（结构）各有不同的表示方式。记录存储一组固定的、类型可能各异的字段，通过字段名而非索引访问。不要混淆这些概念。在伪代码中，错误地更新指针（例如，向链表插入时丢失头节点的引用）是常见的算法书写错误。始终画一张草图来指导你的代码。
7. Stacks and Queues: Operations and Applications | 栈与队列：操作及应用
Stacks (LIFO) and queues (FIFO) are not only examined as abstract data types but also as underlying mechanisms for recursion, expression evaluation, and process scheduling. A common mistake is to implement a queue using two stacks without realising that this changes the amortised cost. More perilous is misinterpreting stack frames in recursion. Students often state that a recursive subroutine 'uses a stack to store return addresses' but forget to mention that local variables and parameters for each call are also stored in the stack frame. When tracing a recursive factorial function, missing that each call's return address must be popped successively to resume execution can lead to incorrect ordering of operations.
栈（后进先出）和队列（先进先出）不仅作为抽象数据类型被考查，还作为递归、表达式求值和进程调度的底层机制被考查。一个常见错误是使用两个栈来实现队列而没有意识到这会改变操作的均摊成本。更危险的是在递归中错误解读栈帧。学生经常说递归子程序“使用栈来存储返回地址”，却忘记提及每次调用的局部变量和参数也存储在栈帧中。在追踪递归阶乘函数时，若不理解每次调用的返回地址必须依次弹出才能恢复执行，就可能导致操作顺序的错误。
ut checking for underflow, or to leave the stack non-empty at the end for a balanced bracket algorithm, forgetting to check the final isEmpty condition. Always include error states in your pseudocode. For queues, circular array implementation is a favourite — 1) MOD array_size is essential, and omitting the MOD operation is a frequent coding slip.
在试卷二的算法题中，你可能会被要求使用栈来反转字符串或检查括号匹配。一个典型错误是从空栈中弹出而未检查下溢，或者在括号匹配算法结束时栈中仍有剩余元素，却忘记检查最终的 isEmpty 条件。伪代码中一定要包含错误状态。对于队列，循环数组实现是常考点——使用 (front + 1) MOD array_size 计算下一位置至关重要，遗漏 MOD 操作是常见的编码疏忽。
8. Searching and Sorting Algorithms: Comparison and Big O | 搜索与排序算法：比较与大O表示法
Linear search, binary search, bubble sort, insertion sort, merge sort, and quick sort are the core algorithms. When comparing them, don't just quote the average time complexity — describe the best and worst cases and the reason behind each. For example, quicksort's worst case O(n²) occurs when the pivot is the smallest or largest element repeatedly, while its average O(n log n) rests on balanced partitioning. Merge sort guarantees O(n log n) but requires O(n) extra space. A common error is to claim binary search works on any list; it requires a sorted list and random access, so it cannot be applied directly to a linked list. Also, many students confuse the number of comparisons with the number of swaps or moves.
线性搜索、二分搜索、冒泡排序、插入排序、归并排序和快速排序是核心算法。在比较它们时，不要只引用平均时间复杂度——要描述最好和最坏情况及其背后的原理。例如，快速排序的最坏情况 O(n²) 发生在每次选择的枢轴都是最小或最大元素时，而其平均 O(n log n) 则依赖于平衡划分。归并排序能保证 O(n log n)，但需要 O(n) 额外空间。一个常见错误是声称二分搜索适用于任何列表；它要求列表有序且支持随机访问，因此不能直接应用于链表。此外，许多学生将比较次数与交换或移动次数搞混。
In OCR's pseudocode style, you must also be comfortable tracing these algorithms and identifying an algorithm from a given snippet. A specific pitfall lies in the insertion sort inner loop: students often forget to shift elements right before inserting, or they misplace the index boundaries. When tracing a merge sort on an array of odd length, incorrectly handling the midpoint can lead to off-by-one errors. Practice drawing splitting trees and merge steps. Lastly, be precise with stability: bubble, insertion and merge sorts are stable; quicksort and heapsort (though heapsort is not always required in AS) are not. Verifying stability can earn marks in a comparison table.
在OCR风格的伪代码中，你还必须能熟练追踪这些算法，并根据给定代码片段识别算法。一个具体的陷阱在于插入排序的内循环：学生常常忘记在插入之前先将元素右移，或者搞错索引边界。在追踪对奇数长度数组的归并排序时，错误处理中点可能导致差一错误。多练习画出分割树和归并步骤。最后，关于稳定性要精确：冒泡排序、插入排序和归并排序是稳定的；快速排序和堆排序（尽管堆排序在AS中并不总是要求）则不是。在比较表中验证稳定性能为你赚分的。
9. Recursion: Base Case, Stack Usage and Converting to Iteration | 递归：基案、栈的使用及转化为迭代
towards it. Many students write a recursive function that works in theory but lacks a well‑ ion factorial(n) with base case n == 1 will never stop if called with n = 0. Always test base cases with boundary values. Furthermore, understanding stack depth is vital: for each recursive call, a new stack frame is pushed. When asked to convert a recursive algorithm to an iterative one, a common mistake is to try to replace it with a simple loop when an explicit stack data structure is needed, especially for tree traversals or Tower of Hanoi.
递归需要仔细思考。两个必不可少的组成部分是基案（停止条件）和逐步逼近基案的递归步骤。许多学生写出的递归函数在理论上可行，但缺乏定义良好的基案，或者由于参数错误导致基案永远无法到达，从而引发栈溢出。例如，一个阶乘函数 factorial(n) 以 n == 1 为基案，当 n = 0 被调用时将无法停止。始终要用边界值测试基案。此外，理解栈深度至关重要：每次递归调用都会压入一个新的栈帧。当被要求将递归算法转化为迭代算法时，常见错误是试图用一个简单循环来代替，但实际上有些情况需要一个显式的栈数据结构，尤其是在树遍历或汉诺塔问题中。
Another nuance: tail recursion. OCR may not explicitly test tail‑call optimisation, but understanding that a tail‑recursive function can be mechanically converted into a loop without a stack can demonstrate advanced insight. In any recursion question, mention that recursion is elegant but carries overhead due to function calls and stack usage, while iteration is more memory‑efficient. However, never claim that recursion is 'always slower' — that oversimplification can be penalised. Provide a balanced evaluation, comparing time and space complexity with a concrete example.
另一个细微之处：尾递归。OCR 可能不会直接考查尾调用优化，但能认识到尾递归函数可以机械地转换为无需栈的循环，可以体现出更深的洞察力。在任何递归问题中，都要提到递归虽然优雅，但由于函数调用和栈的使用会产生额外开销，而迭代则更节省内存。但是，绝不要断言递归“总是更慢”——这种过度简化可能会被扣分。要提供平衡的评价，用具体的例子来比较时间和空间复杂度。
10. Data Representation: Floating‑Point Normalisation and Two's Complement | 数据表示：浮点数规格化与补码
Floating‑point representation, especially normalisation, is a notorious minefield. In a normalised binary floating‑point number, the mantissa should have its first two bits different (i.e., 01 or 10) to ensure maximum precision and a unique representation. A typical mistake is to normalise by simply moving the binary point and adjusting the exponent without checking the sign‑bit constraint; for a positive number, the normalised mantissa starts with 0.1, and for a negative number it starts with 1.0 in two's complement form. When adding two floating‑point numbers, students often align the exponents but forget to denormalise and re‑normalise the result. Always work through the steps: equalise exponents, add mantissas, normalise, round.
浮点数表示，尤其是规格化，是众所周知的雷区。在规格化的二进制浮点数中，尾数的前两位应不同（即 01 或 10），以确保最大精度和唯一表示。一个典型的错误是简单地移动二进制点并调整指数来进行规格化，却没有检查符号位约束；对于正数，规格化尾数以 0.1 开头；对于负数，则以补码形式的 1.0 开头。当两个浮点数相加时，学生经常对齐了指数，却忘记对结果进行反规格化和再规格化。务必按照步骤进行：使指数相等、相加尾数、规格化、舍入。
Two's complement arithmetic is another high‑frequency topic. Adding two negative numbers and obtaining a positive result signals overflow. Students often check only the most significant bit (MSB) of the result, but overflow detection in two's complement requires comparing carry‑in and carry‑out of the MSB: if they differ, overflow has occurred. When extending the bit length of a two's complement number, you must perform sign extension — copying the sign bit into the new higher‑order bits. Mistakenly padding with zeroes for a negative number changes its value.
补码运算是另一个高频考点。两个负数相加却得到正数结果，表明发生了溢出。学生往往只检查结果的最高有效位（MSB），但在补码中检测溢出需要比较 MSB 的进位输入与进位输出：如果两者不同，则发生了溢出。当扩展补码数字的位长时，必须进行符号扩展——将符号位复制到新的高位中。对负数错误地用零填充会改变它的值。
11. Boolean Algebra and Logic Circuit Simplification | 布尔代数与逻辑电路化简
Boolean identities, De Morgan's laws, and Karnaugh maps (K‑ expression algebraically, a common slip is to misapply De Morgan: forgetting that (A + B)' = A' � B' and (A � B)' = A' + B', or failing to keep the overbars precise. Students sometimes try to 'cancel' terms without justification, for example going from A + A � are using (absorption, distribution, complement, idempotent, etc.) to demonstrate method.
布尔恒等式、德摩根定律和卡诺图（K‑maps）经常被考查。在用代数方法化简布尔表达式时，一个常见失误是误用德摩根定律：忘记 (A + B)' = A' � B' 以及 (A � B)' = A' + B'，或者无法保持上划线的精确。学生有时试图在没有依据的情况下“消除”项，例如从 A + A � B 得出 B，这是不正确的；正确的化简是通过吸收律得到 A + B。始终注明所使用的定律（吸收律、分配律、互补律、等幂律等），以展示解题步骤。
When using a 2‑variable or 4‑variable K‑map, do not group incompletely. A high‑ two in size, or to extend a group across the map edges incorrectly. Remember that a 4×2 K‑map wraps both horizontally and vertically; missing that wrap‑round grouping leads to a less‑minimal expression. Also, don't forget to handle 'don't care' conditions if provided — ) briefly to show you can map algebra to hardware. A simple table comparing the reduced expression with the original truth table can verify correctness.
在使用2变量或4变量卡诺图时，不要进行不完整的画圈。一个高频错误是圈起一组大小不是2的幂的1，或者错误地跨越地图边缘进行分组。记住，4×2 的卡诺图在水平和垂直方向上都循环连接；遗漏这种环绕分组会导致表达式未达最简。此外，若题目提供了“无关”条件，不要忘记使用它们——它们能显著简化你的电路。在导出最小积之和后，简要指明相应的逻辑门（与、或、非），以展示你能将代数映射到硬件。用一个小表格对比化简后的表达式与原始真值表也是验证正确性的好办法。
12. Network Models and Protocols: TCP/IP Stack and Application Layer | 网络模型与协议：TCP/IP协议栈与应用层
Network questions range from the functions of the four TCP/IP layers (Application, Transport, Internet, Link) to the specific roles of protocols like HTTP, FTP, SMTP, POP3, DNS, DHCP, TCP, UDP, IP, and Ethernet. A frequent error is to place protocols in the wrong layer. DHCP, for example, resides in the Application layer, not the Link layer, although it uses UDP and IP. Similarly, DNS is an Application layer protocol. Students often confuse TCP (connection‑oriented, reliable, with flow control and sequencing) with UDP (connectionless, lightweight, no guaranteed delivery). When explaining why a particular service uses UDP instead of TCP — such as VoIP or live streaming — mention that low latency is more important than occasional packet loss, and that TCP's retransmissions would introduce unacceptable delays.
网络题目涵盖从TCP/IP四层的功能（应用层、传输层、网际层、链路层）到HTTP、FTP、SMTP、POP3、DNS、DHCP、TCP、UDP、IP、Ethernet等协议的具体作用。一个常见错误是将协议放在错误的层级。例如，DHCP位于应用层而非链路层，尽管它使用UDP和IP。类似地，DNS是应用层协议。学生经常混淆TCP（面向连接、可靠、具有流量控制和序列号）和UDP（无连接、轻量级、无传输保证）。在解释为何某项服务使用UDP而非TCP时——比如VoIP或直播——要提到低延迟比偶发的丢包更为重要，而TCP的重传会引入不可接受的延迟。
When addressing an IP address and subnet mask question, many candidates can state the network ID and host ID but stumble on calculating the number of usable hosts (2number of zero host bits – 2) and often forget to subtract the network and broadcast addresses. In the context of NAT (Network Address Translation), a frequent misconception is that NAT increases security; while it provides a level of obscurity, it is not a firewall and should not be described as a comprehensive security solution. Finally, always relate the model to real‑world scenarios: when you open a browser, HTTP uses TCP port 80, TCP segments are encapsulated in IP packets, IP packets are framed by Ethernet, and the physical signal travels over twisted‑pair or fibre. Explaining this encapsulation flow precisely demonstrates thorough understanding.
在回答IP地址和子网掩码问题时，许多考生能说出网络ID和主机ID，但在计算可用主机数（2主机位零的个数 – 2）时会出错，而且常常忘记减去网络地址和广播地址。在NAT（网络地址转换）的语境中，一个常见的误解是认为NAT能增强安全性；尽管它提供了某种程度的隐蔽性，但它并非防火墙，不应被描述为一种全面的安全解决方案。最后，始终将模型与现实场景联系起来：当你打开浏览器时，HTTP使用TCP端口80，TCP段被封装在IP数据包中，IP数据包被以太网帧封装，物理信号通过双绞线或光纤传输。精确描述这种封装流程能展现你透彻的理解。
Published by TutorHao | OCR Computer Science Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
或截图后用「扫一扫 → 相册」选择本图