OCR GCSE计算机 J277 二进制 布尔逻辑 SQL
J277 是 OCR 的计算机科学 GCSE，2022 年首次考试，两卷各 80 分、各占一半。题量不算大，但每一分都要写清依据，只背定义拿不到高分。这篇文章按卷面顺序走一遍：两张卷子的考核目标，几个每年都出现的计算点，网络与安全题的配对要求，以及 Section B 的程序写法。文中所有卷面结构与分值都可以在 OCR 官方考纲里逐条核对。
OCR's J277 sits behind two written papers of ninety minutes and eighty marks each, with no coursework element. What separates a grade 7 script from a grade 4 script is rarely factual recall; it is whether the answer explains cause and effect in the terms the mark scheme expects. This piece works through the paper structure, the calculations that reappear every series, and the programming section where candidates leak marks most easily.
两卷各 80 分与三个考核目标的分工
J277 只有两张笔试卷，考完就结束，没有课程作业。卷一考计算机系统，六个主题从系统结构、存储、网络一直排到法律与伦理。卷二考计算思维、算法与编程，卷面分成两部分，Section A 五十分，Section B 三十分专门考编程。两卷都是九十分钟、八十分、各占一半，都不允许使用计算器。按 OCR 官方考纲的权重，AO1 知识与理解占三成，AO2 应用占四成，AO3 分析与评价占三成。卷二同时覆盖三个目标，卷一不考 AO3。很多学生以为把定义背完就够了，实际七成分数落在应用与分析上。
Both papers run for ninety minutes, carry eighty marks each and forbid calculators, so arithmetic has to be done by hand and shown in the space provided. Paper 1 rewards knowledge and application across the six systems topics, while paper 2 adds analysis because Section B asks candidates to design, write, test and refine code. The weighting is blunt: thirty per cent knowledge, forty per cent application, thirty per cent analysis, which caps a purely memorising candidate well below the top grades.
数据单位、二进制加法与十六进制换算
数据量的单位从 bit 一路排到 PB，四位二进制叫 nibble，八位二进制叫 byte，这两个名称常在选择题里出现。二进制加法只记四条规则，零加零得零、零加一得一、一加一写零进位、一加一再加进位得一。左移一位相当于乘二，右移一位相当于除二，题目常用移位来问数值变化。十六进制把四位二进制并成一位，A 到 F 分别对应十到十五，转换时从右往左补足四位再写。常见的错误是丢掉前导零，把八位答案写成七位甚至六位。单位换算题还要看清题干默认的进位，一千还是 1024 由题目决定，别凭习惯写。
Units run from the bit up to the petabyte, and nibble and byte definitions are expected to be exact rather than approximate. Binary addition has only four cases, with one plus one producing a zero and a carry into the next column; left shifts double a value and right shifts halve it. Converting to hexadecimal means grouping bits in fours from the right and padding with leading zeros, a step candidates skip far more often than they should.
图像与声音的文件大小计算
图像文件大小等于宽乘高再乘色深，算出来的是位数，除以八才是字节。一张 1920 乘 1080、色深 24 位的图片，位数是四千九百七十六万六千四百，折算下来约 5.9 MB。声音的文件大小按采样率、位深、时长与声道数相乘，采样率用赫兹，时长要换成秒。提高分辨率或色深都会推高文件大小，题目常问「提高一倍会变成多少」。答题时把算式、单位、结果三行写出来，过程分是单独给的。常见的错误是只写答案不写单位，或者把位数直接当字节。
Image size comes from width times height times colour depth in bits, then divided by eight to reach bytes, so a 1920 by 1080 image at 24 bits lands near 5.9 megabytes. Sound follows the same logic with sample rate, bit depth, duration in seconds and channel count multiplied together. Marks here are awarded for the working, and writing only the final figure costs candidates the method marks that made the question easy in the first place.
RAM、ROM 与虚拟内存
RAM 是易失的，断电就清空，存放的是正在运行的程序与当前数据。ROM 的内容出厂写入、平时只能读，存放的是开机自检与启动所需的指令。虚拟内存把硬盘上的一块空间当作内存用，物理内存不足时页面被换进换出，机器会明显变慢。闪存没有活动部件、读写快、体积小，固态硬盘、手机与 U 盘都用它；磁性存储单位成本低、容量大，数据中心仍在大量使用；光学存储适合长期归档。很多学生以为 ROM 可以随意改写，这个误解会让定义题直接失分。答这类题写清「存什么」与「是否易失」两点，比堆术语更有效。
Volatile memory clears when power is cut, and that single property separates RAM from ROM in most mark schemes. Virtual memory borrows disk space when physical RAM runs short, which explains why a machine slows down once pages start swapping in and out. Flash storage has no moving parts and suits portable devices, magnetic disks stay cheaper per gigabyte, and optical media still serve archiving.
网络类型、拓扑与协议分层
局域网覆盖一栋楼或一个校园，广域网跨城市乃至跨国，互联网是规模最大的广域网。客户端服务器模型把资源集中管理，对等网络让每台机器既是客户又是服务端，安全性弱但成本低。星型拓扑以交换机为中心，布线简单，单点故障会拖垮整个网络；网状拓扑冗余高，代价是线路数量与成本。无线接入点把无线设备接入有线网络，路由器在不同网络之间转发数据包，交换机在局域网内部按 MAC 地址转发。协议是设备之间通信的约定，HTTP、HTTPS、FTP、SMTP 与 IMAP 各管一段。分层的好处是每层可以独立演进，不必重写其他层，这句话在评价题里能直接拿分。
A LAN covers a single site while a WAN spans cities, and the internet is simply the largest WAN in use. Client-server networks centralise control and backup, whereas peer-to-peer networks trade security for cost and simplicity. Star topologies depend on one central switch, mesh topologies buy resilience with extra cabling, and the layered model lets one protocol change without breaking the others.
网络威胁与防护措施要配对
钓鱼靠伪造邮件或网站骗走账号密码，暴力破解靠反复尝试穷举密码，拒绝服务用海量请求占满带宽与资源。SQL 注入把恶意语句塞进输入框，根源是程序没有净化用户输入，这一点和卷二的文件读写题连在一起。数据拦截发生在传输途中，加密能让截获的内容读不出来。防护措施要跟威胁对上号：防火墙过滤流量，访问级别限制权限，强密码与双因素挡住猜密码，渗透测试提前找出漏洞，反恶意软件处理已知病毒。阅卷时常见学生把措施全列一遍，却不对应题目问的那个威胁，写了六条只拿到两分。
Each threat has a matching defence, and the mark scheme rewards the pairing rather than a catalogue of security products. Phishing is answered by staff training and verification, brute force by lockouts and two-factor authentication, denial of service by filtering and bandwidth management, and SQL injection by sanitising input before it reaches the query. Physical controls and encryption cover interception, which is why a firewall on its own never earns full marks on this question.
线性查找与二分查找的比较次数
线性查找从第一个元素开始逐个比较，数据无序也能用，最坏情况要走完整张表。二分查找要求数据先排好序，每次比较中间元素，把候选范围砍掉一半。一千零二十四个有序数据用二分十次左右就能定位，逐个比较最坏要走完一千零二十四条。中间位置由 low 加 high 除以二取整得到，边界更新写错是这道题最常见的失分点。题目有时给一段伪代码要你补全循环条件，看清 while 里写的是小于等于还是小于。很多学生以为二分更快就在任何列表上都用二分，遇到无序数据这一步就错在前面。
Linear search compares items one by one and works on unsorted data, which is why it still appears in questions about short lists. Binary search needs sorted data and halves the remaining range at each comparison, so a thousand items take about ten checks instead of a thousand. Candidates lose marks by forgetting to update low or high correctly, or by applying binary search to a list that was never sorted.
冒泡排序的趟数与归并排序的空间代价
冒泡排序每一趟把当前最大的元素推到末尾，n 个元素最多需要 n 减一趟。考试常给出一列数字问第二趟之后的序列，这时只看还没排好的前几项就够。归并排序把列表对半分开、分别排好再合并，比较次数在 n log n 量级，代价是需要额外的存储空间。插入排序像整理手里的扑克牌，数据接近有序时非常快。三种算法的名字要能和过程描述对上，题目给流程让你选算法时不要只看「快」这一个词。常见的错误是冒泡排序只写一趟就交卷，没注意问的是第几趟。
Bubble sort pushes the largest remaining value to the end of each pass, so a list of n items settles in at most n minus one passes. Merge sort splits, sorts and merges, which keeps comparisons near n log n but demands extra memory for the merging step. Insertion sort behaves like sorting a hand of cards and is fast when the data already sits close to order.
数据类型、字符串处理与数组使用
整数、实数、布尔、字符与字符串是题目里出现的五种数据类型，选错类型会被判成逻辑问题。输入一律先当字符串读进来，需要计算时再转成整数或实数，题目提示输入可能不合法时这一步最容易出问题。字符串的长度、截取、大小写转换与 ASCII 码是常考项，字符与编码值的互转要写清方向。数组存放同类型的多个值，用下标访问，循环的终止条件用数组长度而不是写死的数字，换一组数据仍然成立。变量与常量的区别在于运行中是否允许改变，给常量命名时要能看出含义。很多学生以为变量名随便取不扣分，Section B 里读不懂的命名会被当成逻辑不清。
Choosing integer instead of real, or string instead of character, counts as a design error rather than a slip of the pen. Input arrives as text, so conversion has to be explicit and checked before any arithmetic is attempted. Loops should be bounded by the array length rather than a hard-coded number, and names should describe what a variable holds, because a marker has to follow the logic without guessing.
SQL 查询语句与文件读写的动作
SQL 的查询骨架是 SELECT 选字段、FROM 指定表、WHERE 加条件，星号表示取全部字段。字符串条件要加单引号，数字条件不加，写反了整条语句就不成立。排序用 ORDER BY，默认升序，要降序得写明 DESC。文件操作要写全打开、读或写、关闭这几个动作，写模式会覆盖原有内容，想保留旧数据就得用追加模式。题目给出数据文件的结构时，先把字段与类型抄在草稿上，再写读入循环。这一个动作能省下大量返工，因为字段顺序错位是文件题最主要的扣分原因。
A query needs the fields, the table and the condition, and quoting a number or leaving a string bare makes the whole statement invalid. Sorting is handled by ORDER BY, which defaults to ascending order unless descending is stated explicitly. File handling follows open, read or write, then close, and write mode overwrites existing content, so appending is a separate decision that has to be made on purpose.
健壮程序：验证、净化与三类测试数据
验证检查输入是否落在允许范围内，范围、类型、长度、是否为空与格式五项各有对应手法。净化处理的是内容本身，把可能被当成代码的字符去掉或转义，SQL 注入与脚本注入都靠这一步挡住。测试数据分正常、边界与无效三类，边界包括零、上限、上限加一与空字符串，错得最多的是漏掉上限加一。语法错误让程序跑不起来，逻辑错误程序能跑但结果不对，两者的排查方向完全不同。迭代测试在开发过程中做一段测一段，最终测试在交付前按整个程序走一遍。题目问「为什么还要做最终测试」时，答案要落到模块衔接与整体数据流上。
Validation checks that input falls inside the permitted range, type, length and format, while sanitisation removes or escapes characters that a program could otherwise read as code. Test data should cover normal, boundary and invalid cases, and the boundary set is usually where candidates stop too early. Syntax errors stop a program running at all, logic errors let it run and produce the wrong answer, and final testing exists because modules that work alone can still fail once they are joined.
布尔逻辑真值表与 AND OR NOT 的组合
三个基本逻辑门是 AND、OR 与 NOT，真值表的行数由变量个数决定，两个变量四行、三个变量八行。AND 只有输入全为真才输出真，OR 只要有一个输入为真就输出真，NOT 把真假反过来。在门前面加一个小圆圈得到 NAND 与 NOR，它们的输出与对应的门完全相反。判断顺序是先括号、再 NOT、然后 AND、最后 OR，跳步计算是失分的主要原因。考试有两种问法，给表达式画真值表，或者给真值表写表达式，后一种先找出输出为真的行再合并。常见的错误是把 A AND B OR C 当成 A AND 括号 B OR C，加一层括号意思就变了。
Three gates cover the whole specification, with AND requiring every input to be true, OR requiring at least one, and NOT inverting a single value. Truth tables grow as two to the power of the number of variables, so two inputs need four rows and three inputs need eight. Evaluation runs from brackets to NOT, then AND, then OR, and candidates who skip that order produce tables that contradict their own expression.
戴睿（dere）Neo优选系列 创意双触摸屏高性能笔记本电脑大学生AI学习商务办公设计炒股娱乐手提轻薄本32/1TB¥4586.0📱 去京东购买京东精选
需要一对一辅导？Need One-on-One Tutoring?
A-Level / GCSE / IB / KS2-KS3 全科辅导
16621398022 同微信
Visit aleveler.com
或截图后用「扫一扫 → 相册」选择本图