Loading…

Worked solutions

IB Computer Science HL · Paper 2

Computer Science, Higher Level

Session: May 2019
Marks: 260
Duration: 1 hour 20 min
Calculator: Not allowed
Questions: 17

Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Q9 Q10 Q11 Q12 Q13 Q14 Q15 Q16 Q17

Keep your copy of the paper open alongside. We don't reproduce the questions: each part below is labelled to match the May 2019 Global paper, which your school or the IB can provide.

Practice mode

Hide our working until you've had a go.

Option A: Databases · Option

Questions 1–4 · 65 marks

Question 1

15 marks

Marble Reading Book Stores (MRBS) relational schema, queries, and ACID properties

Databases

(a)

2 marks

Identify the three core entities in the bookstore scenario: PUBLISHER, AUTHOR, and BOOK.
Establish the cardinality from PUBLISHER to AUTHOR: one publisher publishes works from one or more authors (1-to-many relationship).
Establish the cardinality from AUTHOR to BOOK: one author writes one or more books (1-to-many relationship).
Construct the Entity-Relationship Diagram (ERD):
PUBLISHER (1) ----< (publishes books from) >---- (M) AUTHOR (1) ----< (writes) >---- (M) BOOK

Mentor tip

Always label both the entities and the directional relationship verbs along with 1:M crow's foot or cardinality notation.

(b)

2 marks

Recognize that Book_Title is stored as an arbitrary free-text string (VARCHAR / String data type).
Because book titles can legitimately consist of almost any combination of words, numbers, or punctuation of varying lengths, standard automated rules like range checks, lookup lists, or strict format masks cannot verify whether a title is sensible.

Mentor tip

Contrast unrestricted string fields like titles with structured attributes like ISBN or Year_Pub, which easily support format, length, or range validation.

(c)

1 mark

Locate the row in the BOOK table matching both conditions: Genre = "Non-fiction" AND ISBN = '0-98124-612-2'.
Extract the projected Book_Title attribute:
Seeking the truth

Mentor tip

Output only the exact attribute requested in the SELECT clause rather than the entire tuple.

(d)

4 marks

Identify the required tables: BOOK (contains Book_Title) and AUTHOR (contains Publisher_Name and links to BOOK via Author_Num), optionally joining PUBLISHER.
Join BOOK and AUTHOR on their common key: AUTHOR.Author_Num = BOOK.Author_Num.
Filter the records where Publisher_Name = "Orlando Crux".
Write the complete SQL statement:
SELECT BOOK.Book_Title FROM BOOK INNER JOIN AUTHOR ON AUTHOR.Author_Num = BOOK.Author_Num WHERE AUTHOR.Publisher_Name = 'Orlando Crux';

Mentor tip

Since Publisher_Name is already present as a foreign key inside the AUTHOR table, a single INNER JOIN between BOOK and AUTHOR is sufficient, though joining PUBLISHER as well is also valid.

(e)

2 marks

In transaction processing, atomicity ('all-or-nothing' execution) ensures that every operation within a database transaction completes successfully as a single indivisible unit.
If any part of the transaction fails, the DBMS triggers a rollback to undo partial changes, preventing corrupted or half-updated bookstore records (note: at the attribute level, atomic fields also speed up searching and sorting by keeping values indivisible).

Mentor tip

Link transaction atomicity directly to full completion or complete rollback upon failure.

(f)

2 marks

Provide each user with a consistent, isolated view of the data during active transactions so concurrent edits do not conflict.
Enforce integrity constraints and validation rules before committing changes; if any constraint is violated during a transaction, reverse the uncommitted modifications via rollback to restore the database to its prior valid state.

Mentor tip

Mention how rollback mechanisms and integrity constraints work together to keep the database in a valid state before and after every transaction.

(g)

2 marks

Redundant data duplicates the same attribute values across multiple records (for example, storing publisher details repeatedly).
When a value changes (such as a publisher relocating to a new city), every occurrence must be updated; missing even one record causes an update anomaly and results in data inconsistency across the database.

Mentor tip

Focus on update anomalies and data inconsistency rather than minor disk storage concerns.

Question 2

12 marks

School server database concurrency, security, recovery, and student privacy

Databases

(a)

3 marks

When a teacher opens a shared teaching resource to edit it, the DBMS applies row-level or record-level locking on that specific resource.
Locking enforces transaction isolation: other teachers can concurrently read or edit different resources simultaneously, but competing write access to the locked resource is blocked until the first teacher's transaction commits and releases the lock.
This prevents lost updates where one teacher's modifications accidentally overwrite another teacher's concurrent changes.

Mentor tip

Clearly connect record/row locking to transaction isolation and preventing conflicting simultaneous edits.

(b)

4 marks

Method 1 — Role-based access rights (minimizing permissions): Assign granular privileges based on user roles so teachers can only view, insert, or modify data necessary for their duties, restricting access to sensitive administrative records.
Method 2 — Database views and table isolation: Create restricted database views or stored routines so staff interact only with filtered subsets of student data rather than having direct access to underlying base tables.
Method 3 — Audit trails: Maintain automated audit logs recording user logins, queries, and permission changes to establish accountability and trace unauthorized access attempts.

Mentor tip

For 4 marks, state two distinct administrative security controls and explain how each protects the school's records.

(c)(i)

2 marks

Strategy 1 — Regular off-site/mirror backups: Maintain scheduled full and incremental backup copies (or mirrored servers) so the database can be restored to the latest clean state following corruption.
Strategy 2 — Deferred update or shadow paging with transaction logs: Hold changes in transaction logs or shadow pages until a commit point is reached, allowing the system to discard incomplete writes or replay committed log entries during recovery.

Mentor tip

State a concrete recovery mechanism (such as mirror backups, shadow paging, or deferred updates) and explain how it restores uncorrupted data.

(c)(ii)

3 marks

Select a privacy protection technique such as data encryption or data masking (obfuscation).
Explain the mechanism: Encryption scrambles sensitive plaintext into unreadable ciphertext using a cryptographic algorithm, while data masking replaces confidential characters with symbols or pseudonyms.
Apply to the school context: Scrambling fields like Student_ID, medical notes, or home addresses ensures that even if unauthorized users inspect the storage media or view a screen, the personal details cannot be read without the decryption key.

Mentor tip

Include three elements for full credit: name the technique, explain how it transforms data, and give a specific student attribute example.

Question 3

18 marks

Armour Hardware Company database normalization to 3NF and key constraints

Databases

(a)

4 marks

Reason 1 — Eliminating data redundancy and anomalies: Normalization removes duplicated data across rows, preventing insertion, deletion, and update anomalies while improving storage and update efficiency.
Reason 2 — Enforcing strict key dependency and structural stability: Every non-key attribute becomes fully dependent on the primary key, allowing the schema to be extended with new data types while minimally affecting existing applications.

Mentor tip

Structure your answer as two distinct points, each pairing a core principle of normalization with its practical benefit.

(b)

2 marks

First Normal Form (1NF) requires that every attribute in every tuple contains only a single, atomic value with no repeating groups.
In the SALES_PERSON table, each salesperson row contains multi-valued repeating groups of products sold (multiple Product_Number, Unit_Price, Product_Name, and Date_And_Time entries packed into a single record).

Mentor tip

Point out both the theoretical rule (atomicity of attributes) and the specific columns in SALES_PERSON that violate it.

(c)

8 marks

Decompose the unnormalized table by separating repeating product sales, removing partial dependencies on composite keys (2NF), and removing transitive dependencies such as Manager_Name depending on Manager_Number via Sales_Person_Number (3NF).
Define the PRODUCT table with Product_Number as primary key:
PRODUCT (Product_Number [PK], Product_Name, Unit_Price)
Define the MANAGER table to isolate manager details with Manager_Number as primary key:
MANAGER (Manager_Number [PK], Manager_Name)
Define the SALESPERSON table linking to MANAGER via a foreign key:
SALESPERSON (Sales_Person_Number [PK], Sales_Person_Name, Manager_Number [FK])
Define the PURCHASE (or SALE) junction table recording individual transactions with either a surrogate key Purchase_ID or a composite primary key:
PURCHASE (Date_And_Time [PK], Product_Number [PK, FK], Sales_Person_Number [PK, FK]) — or using Purchase_ID [PK] with Date_And_Time, Product_Number [FK], Sales_Person_Number [FK].

Mentor tip

Clearly mark every Primary Key (PK) and Foreign Key (FK), and verify that Manager_Name is separated into its own MANAGER relation so no transitive dependency remains in SALESPERSON.

(d)

2 marks

Referential integrity ensures that every foreign key value in a child table matches an existing primary key value in the referenced parent table.
Without it, orphan records or invalid references (such as recording a sale for a non-existent salesperson or deleting a manager still assigned to staff without cascading rules) would cause mismatched transaction data.

Mentor tip

Mention how referential integrity prevents orphan records and coordinates cascading updates or deletes across related tables.

(e)

2 marks

In many-to-many junction tables or transaction logs, no single attribute is sufficient to uniquely identify every record without duplication (for example, a salesperson sells multiple items, and a product is sold multiple times).
Combining two or more attributes into a composite primary key guarantees that each combination uniquely identifies a single tuple upon which all non-key attributes depend.

Mentor tip

Explain that a composite key is used when single attributes can repeat, but their combined tuple is guaranteed to be unique.

Question 4

20 marks

ZCC global paper supplier data warehousing, ETL, and data mining strategies

Databases

(a)(i)

2 marks

A data warehouse stores both current operational snapshots and long-term historical records collected over many years.
Every record is associated with a timestamp or time horizon so analysts can compare sales performance across different periods and identify longitudinal trends.

Mentor tip

Emphasize the presence of historical data and timestamps for multi-period comparison.

(a)(ii)

2 marks

A data warehouse consolidates and standardizes data from ZCC's global offices into a single repository optimized for complex analytical queries (OLAP) rather than day-to-day transaction processing.
This enables historical trend prediction and market analysis to guide strategic decisions on diversifying into new stationery and calculator lines without slowing down operational databases.

Mentor tip

Contrast analytical decision support and historical trend analysis in a warehouse with routine transactional updates.

(b)

2 marks

Data extracted from international branch offices often follows different local conventions, schemas, and formats (such as incompatible date formats, currencies, or gender codes).
During the ETL transformation stage, data must be cleaned, filtered (removing irrelevant local staff fields), and standardized into a uniform schema so global queries produce accurate results.

Mentor tip

Give concrete examples of transformation tasks such as date/currency standardization, cleaning, or filtering unneeded columns.

(c)

6 marks

Cluster analysis description: An unsupervised technique that groups records (such as ZCC customers) into clusters based on shared similarity across attributes like location, order volume, or age.
Pattern discovery in cluster analysis: Reveals natural groupings and previously unknown customer segments without predefined labels, enabling comparison between clusters.
Classification description: A supervised technique where a predictive classifier model is trained using historical datasets with pre-existing class labels.
Pattern discovery in classification: Assigns new customer records into predefined categories based on learned decision rules.
Key comparisons: Classification requires prior domain knowledge and labelled training sets, whereas cluster analysis operates on unlabelled data without predefined classes; classification sorts new instances into established buckets, while clustering discovers novel structural groupings.

Mentor tip

Cover three aspects for all 6 marks: how cluster analysis works, how classification works, and a direct comparison contrasting labelled vs unlabelled data and predefined vs discovered groups.

(d)

3 marks

Query the data warehouse to extract historical payment records and order histories for a specific customer segment and time window.
Apply deviation detection rules (such as statistical thresholds or conditional decision trees) to establish normal payment behaviour and flag outliers who have missed payment deadlines beyond a set frequency.
Generate a list of flagged high-risk accounts so ZCC can adjust credit terms or send proactive payment reminders.

Mentor tip

Walk through the pipeline: extract historical payment data, run anomaly/outlier detection rules against expected norms, and output the flagged customers.

(e)

5 marks

Database segmentation: Partitions ZCC's customer base into distinct, internally homogeneous groups sharing marketing-relevant traits (such as order frequency, preferred paper types, or business size).
Link analysis: Models entities as nodes and relationships as edges in a network graph, querying across records to uncover hidden associations or co-occurrences between products and buyers.
Application to ZCC's marketing strategy: Segmentation allows ZCC to target tailored promotional campaigns for new gift-wrappers or calculators to the exact customer groups most likely to buy them.
Simultaneously, link analysis reveals which existing paper buyers also purchase stationery items together, enabling cross-selling bundles and informed product diversification.

Mentor tip

Define both segmentation and link analysis clearly, then explain specifically how ZCC can use each to market its new stationery and calculator product lines.

Option B: Modelling and simulation · Option

Questions 5–8 · 65 marks

Question 5

21 marks

Arctic Ocean sea ice melting and albedo feedback simulation model

Modelling and simulation

(a)

3 marks

Ocean surface temperature: Data type is floating-point / real number; suitable range is -5.0 to +30.0 degrees Celsius.
Albedo: Data type is floating-point / real number (or integer percentage); suitable range is 0.0 to 1.0 (or 0.1 to 0.9, or 0% to 100%).
Area of sea ice: Data type is numeric (real or integer); suitable range is 0 to 2,000,000 km^2 (or any non-negative upper bound >= 1,000,000 km^2).

Mentor tip

Ensure every variable has both a valid programming data type (such as Real/Float) and a physically realistic numerical range.

(b)

4 marks

Initialize the starting parameters: OriginalSeaIceArea = 1000000 and SurfaceTempChange = 0.04.
Since each 0.01 C increase reduces sea ice by 1%, a 0.04 C increase causes a 4% (0.04) fractional reduction.
Construct the pseudocode:
OriginalSeaIceArea = 1000000 SurfaceTempChange = 0.04 PercentDecrease = SurfaceTempChange / 0.01 NewSeaIceArea = OriginalSeaIceArea * (1 - PercentDecrease / 100) SeaLevelChange = PercentDecrease * 20

Mentor tip

Clearly show variable initialization, the updated sea ice area formula, and the corresponding sea level rise calculation.

(c)(i)

1 mark

Calculate the 4% reduction from 1,000,000 km^2:
1,000,000 * (1 - 0.04) = 960,000 km^2.

1000000 \times 0.96 = 960000 km^{2}

Mentor tip

Include the numerical value 960,000 (with or without units).

(c)(ii)

1 mark

Multiply the 4% decrease in sea ice area by 20 mm per 1% decrease:
4 * 20 = 80 mm.

4 \times 20 mm = 80 mm

Mentor tip

Each 1% decrease raises sea level by 20 mm, giving 80 mm for a 4% decrease.

(d)

7 marks

Initialize variables for the starting state: IceArea = 1000000, OpenArea = 1000000, and Year = 2019.
Construct a WHILE loop that continues iterating while IceArea >= 10000.
Inside the loop, compute TotalArea, update AveAlbedo using the weighted area formula, and compute the percentage Decrease = 0.3 / (AveAlbedo * AveAlbedo).
Update IceArea, recalculate OpenArea = TotalArea - IceArea, and increment Year by 2 for each iteration:
IceArea = 1000000 OpenArea = 1000000 Year = 2019 loop while IceArea >= 10000 TotalArea = IceArea + OpenArea AveAlbedo = ((IceArea * 0.6) + (OpenArea * 0.1)) / TotalArea Decrease = 0.3 / (AveAlbedo * AveAlbedo) IceArea = IceArea * (1 - (0.01 * Decrease)) OpenArea = TotalArea - IceArea Year = Year + 2 end loop output "Year when sea ice falls below 10,000 km2: ", Year

Mentor tip

Make sure to update both IceArea and OpenArea inside the loop because OpenArea expands as sea ice melts, which in turn lowers AveAlbedo in the next iteration.

(e)

2 marks

Method 1: Implement the iterative formulas across rows in spreadsheet software (such as Excel or Google Sheets).
Method 2: Write a custom program in a high-level programming language using an Integrated Development Environment (IDE).

Mentor tip

Spreadsheets and custom code in an IDE are the two standard implementation platforms for mathematical models.

(f)

3 marks

Because the rate of sea ice reduction compounds every two-year cycle via the non-linear albedo feedback loop, small initial inaccuracies in the rate calculation will amplify over time.
This will cause the predicted year when sea ice drops below 10,000 km^2 (and the corresponding coastal sea-level rise) to be significantly wrong.
Governments and environmental agencies relying on these forecasts could allocate inadequate flood defences or mistime climate mitigation policies.

Mentor tip

Link mathematical error propagation in the model to real-world consequences for policy planning and coastal protection.

Question 6

15 marks

Public health resource simulations and what-if modelling for aging populations

Modelling and simulation

(a)

4 marks

A 'what-if' model operates within a mathematical, rule-based environment representing a real-world system.
Users can keep the underlying rules fixed while systematically varying input data parameters to observe the resulting changes.
Alternatively, users can hold input data constant while modifying the governing rules or formulas.
This flexibility allows decision-makers to explore a wide spectrum of hypothetical future scenarios safely before committing real resources.

Mentor tip

Highlight both varying the inputs under fixed rules and varying the rules under fixed inputs to explore alternative scenarios.

(b)

3 marks

Variable 1: Average life expectancy (or age distribution of the population).
Variable 2: Socio-economic status / household income levels.
Variable 3: Availability of medical facilities and life-extending treatments (or genetic predisposition / family health history / gender).

Mentor tip

Select concrete demographic, medical, or socio-economic factors that directly influence healthcare demand in an aging population.

(c)

5 marks

Issue 1 — Invasion of privacy and sensitive data collection: Building an accurate health simulation requires gathering deeply personal medical, genetic, and lifestyle data from citizens, creating a tension between model accuracy and individual privacy rights.
Issue 2 — Unauthorized sharing and discrimination: If sensitive health records are lacked or shared with third parties (such as insurance companies or employers), individuals could face discrimination or denial of coverage.
Therefore, health agencies must enforce strict anonymization, informed consent, and data protection safeguards when compiling training and input datasets.

Mentor tip

Discuss at least two distinct ethical dimensions (such as patient privacy intrusion and third-party misuse/anonymization) and elaborate on their consequences.

(d)

3 marks

While a static model defines mathematical relationships at a single point, converting it into a simulation allows the system to evolve dynamically over time.
The simulation can repeatedly sample plausible random inputs (such as disease outbreaks or demographic shifts) to handle uncertainty and generate a probability distribution of expected resource demands.
This longitudinal view helps health agencies test capacity bottlenecks (such as hospital bed queues) across decades.

Mentor tip

Emphasize running the model over time and handling uncertainty through repeated sampling of variable inputs.

Question 7

9 marks

Converting 1940s 2D cartoons into rendered 3D animations

Modelling and simulation

(a)

1 mark

Visualization is the process of converting abstract data or mathematical models into a human-interpretable graphical image or animation.

Mentor tip

Keep the definition focused on transforming numerical/model data into a visual representation that humans can interpret.

(b)

2 marks

In 3D animation, characters are initially constructed as abstract mathematical wireframe meshes.
Rendering is required to apply surface textures, lighting, shading, and perspective so the characters appear three-dimensional and move fluidly across frames rather than looking flat.

Mentor tip

Connect the mathematical 3D wireframe model to the visual shading, lighting, and depth produced by rendering.

(c)

6 marks

Implication 1 — Intensive computational processing and time requirements: High-resolution 3D rendering (including ray tracing, lighting, and shadows) is extremely CPU/GPU intensive; unlike real-time interactive games, pre-rendering individual animation frames can take anywhere from several seconds to hours per frame depending on scene complexity.
Implication 2 — High-capacity storage and hardware infrastructure: Every rendered frame must be saved as a high-definition image sequence on high-speed disk arrays before being compiled and transferred to distribution media, requiring substantial storage capacity and render-farm hardware to meet production deadlines.

Mentor tip

Develop two distinct technical implications in depth: computational/time overhead per frame and storage/hardware infrastructure demands.

Question 8

20 marks

Genetic algorithms, neural networks, and machine learning in interactive toys

Modelling and simulation

(a)

4 marks

Genetic algorithm: Mimics biological evolution by initializing a large population of prospective solutions and iteratively applying fitness evaluation, selection, crossover, and mutation over generations until an optimal solution criterion is met.
Neural network: Mimics the interconnected neurons of the human brain using layers of weighted nodes that adjust their synaptic weights during training to recognize complex patterns and map inputs to outputs.

Mentor tip

Contrast the evolutionary population-and-fitness cycle of a genetic algorithm with the brain-inspired layered pattern recognition of a neural network.

(b)

2 marks

Modification 1: Increase the number of input nodes in the input layer to capture more features.
Modification 2: Increase the number of hidden layers (or nodes per hidden layer) to model more complex relationships.

Mentor tip

Look directly at the neural network architecture in Figure 1: adding inputs or adding hidden layers/nodes are the primary structural ways to boost capacity.

(c)

4 marks

Supervised learning: The algorithm is trained using labelled dataset pairs where the desired target output for each input is already known, allowing the model to learn explicit mappings and classify new inputs into known categories.
Unsupervised learning: The algorithm receives unlabelled input data with no predefined target outcomes and must autonomously discover hidden structures, clusters, or patterns (such as grouping objects by acoustic or visual similarity).

Mentor tip

Contrast training with known input-output labels (supervised) against discovering self-organized patterns from unlabelled inputs (unsupervised).

(d)

4 marks

Reason 1 — Unclear pronunciation and speech variability: Young children often mispronounce words, slur syllables, or speak with irregular pitch, causing acoustic recognition errors.
Reason 2 — Grammatical and syntactic immaturity: A child's developing language may not follow standard grammatical syntax expected by the natural language model.
Reason 3 — Out-of-vocabulary or unlearned topics: The child may switch languages, use invented words, or refer to concepts outside the doll's training corpus or previously learned experience.

Mentor tip

Provide four distinct points covering pronunciation clarity, non-standard child syntax, out-of-vocabulary topics, and unprogrammed languages.

(e)

6 marks

Unsupervised learning does not require exhaustive pre-labelled training sets for every possible phrase a child might utter, allowing the doll to adapt autonomously to each child's unique speech patterns and environment.
Instead of relying solely on rigid pre-programmed causal pathways from specific inputs to fixed outputs, unsupervised models build hierarchical representations upward from raw observations.
This bridges the gap between input speech observations and output responses at higher levels of abstraction.
Consequently, the doll can learn larger, more complex conversational models, cluster recurring behavioural cues, and establish novel connections between sets of observations over time.

Mentor tip

Explain how unsupervised learning builds abstractions upward from observations to bridge the input-output gap and handle complex, unlabelled real-world interactions.

Option C: Web science · Option

Questions 9–13 · 65 marks

Question 9

14 marks

Sestra.com responsive CSS, PHP database integration, SEO, and file compression

Web science

(a)

2 marks

Way 1: Dynamically reflow page layouts and grid containers (using media queries or flexible widths) to fit different viewport dimensions and orientations.
Way 2: Scale font sizes, resize images proportionally, or switch navigation menus (such as collapsible mobile menus) based on screen resolution.

Mentor tip

Mention concrete visual adjustments controlled by CSS such as layout reflow, responsive image/font scaling, or device-specific navigation menus.

(b)

3 marks

Step 1: The server executes the SQL query against the CATEGORY table to retrieve all (category_id, category_name) pairs sorted alphabetically by category_name, storing the dataset in $categoryqueryresult.
Step 2: A while loop iterates through each fetched record in $categoryqueryresult and dynamically generates HTML <option> tags inside the <select> drop-down menu, setting the value attribute to category_id and the visible label to category_name.
Step 3: The completed HTML <form> (configured to submit selected data to showresults.php via the HTTP POST method) is sent to the client browser.

Mentor tip

Trace the server-side PHP execution in sequence: SQL query execution -> iterating over the result set to echo HTML <option> elements -> outputting the populated form.

(c)

4 marks

Technique 1 — Descriptive meta tags and structured content: Add accurate <meta> description and keyword tags (and clean semantic HTML separated from CSS formatting) so web crawlers can index and categorize Sestra.com's craft products accurately.
Technique 2 — High-quality authoritative content and backlink building: Publish original, engaging content about local craftspeople (or sitemaps and robots.txt guidance) that naturally attracts inbound links from reputable external websites, boosting page authority.

Mentor tip

For 4 marks, describe two legitimate ('white hat') SEO strategies and explain how each helps search engine crawlers or ranking algorithms.

(d)

2 marks

Lossy compression permanently discards non-essential or perceptually redundant data to achieve much smaller file sizes, resulting in an irreversible reduction in quality (commonly used for images, audio, and video).
Lossless compression identifies repeating byte patterns and encodes them using shorter dictionary tokens so that the exact original data and full quality are 100% restored upon decompression.

Mentor tip

Contrast permanent data removal and quality loss (lossy) with pattern replacement and exact reconstruction upon decompression (lossless).

(e)

3 marks

PDF documents contain structured text, fonts, and formatting instructions where every character and layout token is essential for legibility.
Lossless compression reduces the download size of the PDF files while guaranteeing that, once decompressed on the customer's device, the text and biographical details are reconstructed identically to the original.
Applying lossy compression to a document file would corrupt character codes or structural metadata, rendering the PDF unreadable or unusable.

Mentor tip

Explain both the benefit of lossless compression (exact reconstruction of text/formatting) and why lossy compression would corrupt document readability.

Question 10

16 marks

CERN Large Hadron Collider URL resolution, grid computing, and Creative Commons

Web science

(a)(i)

1 mark

HTTPS (Hypertext Transfer Protocol Secure).

Mentor tip

Look at the prefix before '://' in https://home.cern — be sure to include the 'S' for Secure.

(a)(ii)

3 marks

Step 1: The receiving Domain Name Server (DNS) searches its local database/cache for the domain name 'home.cern'.
Step 2: If the mapping is not found locally, the DNS server forwards the query hierarchically to other DNS servers (root/TLD/authoritative servers) until the matching IP address is located or a lookup failure occurs.
Step 3: The resolved IP address (or an error message if unresolved) is returned to the scientist's client browser so it can establish a connection.

Mentor tip

Trace the lookup clearly: check local DNS database -> escalate to upstream DNS servers if missing -> return IP address (or error) to client.

(b)

6 marks

Reason 1 — Distributed processing capacity and scalability: Analysing 15 petabytes of particle collision data annually requires massive computational throughput; grid computing harnesses distributed CPU/GPU capacity across global partner institutions concurrently, reducing reliance on a single centralized supercomputer and allowing resources to scale easily.
Reason 2 — Data redundancy, reliability, and 24/7 global access: Replicating experimental datasets across multiple international grid nodes eliminates any single point of failure against data loss, while enabling physicists across different time zones to run specialized local analysis and visualization tools around the clock.

Mentor tip

Develop two full 3-mark chains: one focusing on distributed parallel processing power/scalability, and one on fault-tolerant replication or flexible global tool access.

(c)

6 marks

Reason 1 — Accelerating open scientific collaboration with usage safeguards: Creative Commons (CC) licensing allows researchers worldwide to freely access, share, and build upon CERN's experimental data without administrative delays in requesting permission, while still enforcing attribution and preventing third parties from repackaging the data for commercial sale.
Reason 2 — Avoiding costly and impractical international copyright enforcement: Enforcing strict traditional copyright across dozens of jurisdictions with differing legal frameworks would require substantial legal expenditures and monitoring effort, distracting CERN from its core mission of fundamental physics research.

Mentor tip

Explain both the positive scientific benefits of Creative Commons (rapid sharing, attribution, non-commercial protection) and the practical drawbacks of enforcing strict copyright globally.

Question 11

15 marks

Brownsville public libraries XML metadata, CGI forms, blogs, and forums

Web science

(a)

1 mark

Extensibility is the capability of a language, file format, or software system to be expanded beyond its initial specification by defining custom, user-created elements or tags (such as custom XML tags or plug-ins).

Mentor tip

Mention the ability to expand the system by adding custom/user-defined tags or features.

(b)

2 marks

Different library branches and external databases often store records in incompatible proprietary formats.
Because XML stores structured data as self-describing plain text, it is completely hardware- and software-independent, allowing diverse platforms and browsers to exchange and upgrade data seamlessly while keeping data separate from HTML presentation.

Mentor tip

Highlight plain-text platform independence and the ability to share structured data across incompatible systems.

(c)

2 marks

Open standards are publicly documented, non-proprietary technical specifications that are freely available for anyone to adopt and implement.
Interoperability is the practical ability of different computer systems, operating systems, and applications to work together cooperatively and exchange data — a goal that open standards directly facilitate.

Mentor tip

Define both terms clearly and note that open standards are the mechanism that enables interoperability.

(d)

2 marks

The Common Gateway Interface (CGI) provides a standardized protocol convention within HTTP for passing search parameters from the user's web form to an external application or database script on the server.
Once the back-end program queries the library database, CGI passes the generated results back to the web server to be returned to the user's browser, independent of the server's operating system.

Mentor tip

Describe CGI as the standard bridge passing form requests from the web server to back-end scripts/databases and returning the response.

(e)

2 marks

In a blog, only the owner/author (the head librarian) can publish primary articles and initiate discussion topics, while readers may leave comments that are typically pre-moderated before appearing.
In a forum, any registered library user can initiate new discussion threads and post replies that are published immediately to foster peer-to-peer community debate.

Mentor tip

Contrast who can initiate threads (owner-only in a blog vs registered members in a forum) and how moderation is handled.

(f)

6 marks

Benefits of the blog: Enables the head librarian to publish detailed background context on upcoming library policy or budget decisions, fostering informed reader feedback and allowing pre-moderation to filter out spam or abusive remarks.
Risks of the blog: Heavy-handed moderation could censor legitimate criticism of library services, creating an echo chamber that reinforces decisions management already wants to push through.
Benefits of the forum: Empowers library patrons to raise grassroots concerns (such as opening hours or book requests) that the head librarian might otherwise overlook, broadening situational awareness.
Risks of the forum: Unmoderated threads can attract off-topic or offensive posts and require substantial staff time to clarify misconceptions.
Overall evaluation: While both tools increase community engagement, active online contributors represent a self-selecting, vocal minority rather than the entire Brownsville readership; thus, the head librarian must balance online feedback with broader usage statistics to make truly effective decisions.

Mentor tip

Address both the blog and the forum specifically in the library context, weighing informed community feedback against moderation bias and the 'vocal minority' problem.

Question 12

12 marks

Web graph bow-tie topology, PageRank, web crawler criteria, and power laws

Web science

(a)(i)

1 mark

Identify the nodes in Figure 2 that can mutually reach one another along directed edges:
D, E, F, G

Mentor tip

In the Strongly Connected Core (SCC), every node must have a directed path to and from every other node in the core (D -> E -> G -> F -> D).

(a)(ii)

1 mark

Identify the nodes that form a directed passage bypassing the SCC from the IN component to the OUT component:
H and K (or H, L, K)

Mentor tip

A tube connects an IN node (H) directly to an OUT node (K) via L without passing through the SCC.

(b)

2 marks

Although both page E and page C have two incoming links (in-degree of 2), PageRank weights incoming links recursively according to the authority/rank of the linking pages.
Page E receives links from nodes inside the Strongly Connected Core (which themselves have multiple incoming links and higher PageRank), whereas page C only receives links from peripheral source nodes A and B that have zero incoming links.

Mentor tip

Point out that both nodes have 2 in-links, but PageRank weights the quality/in-links of the source pages linking to E higher than those linking to C.

(c)

3 marks

Criterion 1: Presence of a robots.txt file or robot exclusion meta tags (<meta name="robots" content="noindex">) instructing the crawler not to index the page.
Criterion 2: Whether the page has been indexed previously and when it was last modified (crawlers avoid re-indexing static pages too frequently to reduce server load).
Criterion 3: Whether the page content/metadata matches the crawler's specialized domain focus, or whether the page contains broken/dead links or missing header metadata.

Mentor tip

State three distinct technical criteria such as robots.txt/meta exclusion rules, last-modified/crawl frequency timestamps, and content/link validity.

(d)

5 marks

Arguments supporting power laws: Active user counts across social media platforms exhibit a classic scale-free, skewed distribution where a few dominant networks hold hundreds of millions of users while smaller sites trail off sharply.
This aligns with preferential attachment ('the rich get richer'): platforms with massive existing user bases offer stronger network effects and diverse content, naturally attracting a disproportionate share of new sign-ups.
Arguments against relying solely on power laws: Observing a power-law distribution at one point in time does not prove that future growth is governed purely by mathematical power laws (correlation is not causation).
External socio-technical factors — such as shifting youth demographics, privacy scandals, regional regulations, or platform fatigue — can cause once-dominant networks (like MySpace) to collapse rapidly.
Conclusion: Power laws provide a useful baseline model for current market concentration, but cannot reliably predict long-term user numbers without accounting for changing user tastes and competitor innovation.

Mentor tip

Balance the 'rich get richer' network effect of power laws against real-world counterexamples where dominant social networks declined due to changing demographics or policies.

Question 13

8 marks

Narayan City collective intelligence traffic planning and folksonomy behaviour

Web science

(a)

2 marks

Crowdsourcing real-time GPS telemetry from thousands of citizens' mobile devices gathers far larger, more accurate volumes of traffic flow data than manual roadside surveys, reflecting actual driving routes and side-street shortcuts.
Because collection is automated and continuous, Narayan City planners can cost-effectively model congestion hotspots and immediately evaluate whether newly implemented road layout changes reduce journey times.

Mentor tip

Pair a specific benefit of crowdsourced GPS data (scale, real-time accuracy, or automated low cost) with how it improves road planning.

(b)

6 marks

Empowerment and collaborative creation: Folksonomies (user-generated hashtags) allow ordinary web users to classify content bottom-up, helping grassroots social issues trend globally and enabling users with shared niche interests to find communities and collaborate.
Exposure vs echo chambers: While conflicting hashtags can expose users to counter-arguments and public debate, searching exclusively via preferred tags or algorithmic feeds often traps users in 'filter bubbles' that reinforce existing biases.
Conformity and self-censorship: Trending hashtags create social pressure where users with minority viewpoints may hesitate to voice dissent, or conversely alter their posting style toward sensational, viral content to gain attention.
Misuse and manipulation: Open tagging systems are vulnerable to abuse, including coordinated botnets flooding tags (hashtag hijacking/stuffing) to fabricate false trends, commercial influencer marketing, and targeted trolling against users who post under specific tags.

Mentor tip

Discuss both positive behavioral shifts (grassroots mobilization, community discovery, active content creation) and negative shifts (filter bubbles, viral conformity, trolling, and bot manipulation).

Option D: Object-oriented programming · Option

Questions 14–17 · 65 marks

Question 14

16 marks

Regional swimming competition OOP design: Swimmer and Race classes

Object-oriented programming

(a)

1 mark

A mutator (setter) method is a public method used to modify or update the value of a private instance variable (attribute) within an object.

Mentor tip

Always mention both modifying a variable's value and the OOP context of private instance attributes.

(b)

1 mark

State a descriptive boolean attribute relevant to a race:
private boolean isFinals; (or isCompleted, isHeat, isCancelled, isRecordTime)

Mentor tip

Since the stem notes that a race can be either a qualifying heat or a final, isFinals (or isHeat) is the most natural boolean flag.

(c)

3 marks

A class is the abstract blueprint or template that defines the attributes and methods common to a category of objects.
An instantiation (object instance) is a concrete realization of a class allocated in memory and populated with actual data values.
In this scenario, Swimmer is the class blueprint, whereas each of the up to 8 competitors stored inside a Race object's swimmer[8] array is a distinct instantiated Swimmer object with their own specific name and school.

Mentor tip

Include three parts: define a class, define an instantiation, and illustrate both using Swimmer and Race from the UML diagram.

(d)(i)

2 marks

Advantage: Aggregation promotes modular organization and code reuse because the existing Swimmer class encapsulates all swimmer-specific attributes and methods, and the same Swimmer instances can exist independently and participate in multiple Race objects without duplicating data.

Mentor tip

Highlight how aggregation reuses the Swimmer class and keeps swimmer data modular and independent of a single race.

(d)(ii)

2 marks

Disadvantage: Aggregating Swimmer objects inside Race arrays introduces coupling and increases traversal complexity — for example, modifying the Swimmer class interface could impact Race, and finding all heats a specific swimmer competed in requires searching across multiple Race objects.

Mentor tip

Mention either coupling between classes or the added complexity of accessing/searching aggregated objects across multiple races.

(e)

4 marks

Define the public constructor matching the class name Swimmer with parameters (String name, String school).
Assign the parameters to the instance variables and use a loop from 0 to 4 to initialize eventID[i] = "empty" and time[i] = 0.0:
public Swimmer(String name, String school) { this.name = name; this.school = school; for (int i = 0; i < 5; i++) { this.eventID[i] = "empty"; this.time[i] = 0.0; } }

Mentor tip

Ensure the constructor has no return type (not even void), assigns both String parameters, and loops 5 times to initialize both arrays.

(f)

3 marks

Modern programming languages natively support the Unicode character encoding standard (such as UTF-16 or UTF-8).
Unlike standard ASCII (which uses 7 or 8 bits to represent only 128 or 256 Latin characters), Unicode uses 16 or more bits per character to represent over 65,000+ characters.
This vast code point space allows international swimmers' names written in non-Latin scripts (such as Cyrillic, Chinese, Arabic, or accented European characters) to be stored and displayed accurately.

Mentor tip

Name Unicode/UTF, state its bit-width/character capacity compared to 8-bit ASCII, and link it to representing international alphabets.

Question 15

9 marks

Swimming competition Event class, polymorphism, inheritance, and overriding

Object-oriented programming

(a)

3 marks

Each addSwimmers() method is encapsulated and defined within the scope of its own separate class (Race vs Event).
When addSwimmers() is invoked in code, it is called on a specific object reference (for example, myRace.addSwimmers() vs myEvent.addSwimmers()).
The compiler/runtime resolves which implementation to execute based on the class type of the calling object, so no naming conflict occurs.

Mentor tip

Explain class scope encapsulation and how calling the method via an object reference tells the compiler which class's method to invoke.

(b)

4 marks

Advantage 1 — Code reuse and faster development: The subclass automatically inherits tested attributes and methods from the parent Event class, avoiding redundant code duplication.
Advantage 2 — Simplified maintenance and extensibility: Updating shared logic or fixing a bug in the superclass automatically propagates to all derived subclasses, reducing maintenance overhead.

Mentor tip

State two distinct OOP benefits of inheritance (code reuse of tested parent methods and centralized maintenance updates) and elaborate on each.

(c)

2 marks

Method overriding allows the subclass FinalsOnlyEvent to redefine an inherited method from Event with specialized behaviour using the exact same method signature.
Specifically, FinalsOnlyEvent can override addSwimmers() to place swimmers directly into the finals Race object (and override fillFinals() to perform no action) since there are no qualifying heats.

Mentor tip

Define method overriding and give a concrete example of how FinalsOnlyEvent redefines addSwimmers() or fillFinals().

Question 16

20 marks

Sorting qualifying heat arrays and populating the swimming finals

Object-oriented programming

(a)

3 marks

Combine all 11 valid swimmers from races[0] and races[1] and rank them in ascending order of their non-zero swim times:
1) Bella (33.8), 2) Andy (34.2), 3) Eric (34.6), 4) Idan (35.2), 5) Duc (36.3), 6) George (36.6), 7) Karl (37.2), 8) Hetty (37.6) — while Chris (40.9), Fiona (41.2), and Jo (48.8) are eliminated.
Sketch the resulting 8-element swimmer array in finals:
[0]: Bella | [1]: Andy | [2]: Eric | [3]: Idan | [4]: Duc | [5]: George | [6]: Karl | [7]: Hetty

Mentor tip

Double-check that the 8 fastest times are ordered from smallest (fastest, 33.8) to largest (37.6) and exclude the 0.0 empty lane slots.

(b)

6 marks

Declare and instantiate the two parallel arrays of length 16: Swimmer[] tempSwimmer = new Swimmer[16]; and double[] tempTime = new double[16];.
Use nested loops (or a single loop from j = 0 to 7 copying races[0] into index j and races[1] into index j + 8) along with accessor methods getSwimmer(j) and getTime(j):
Swimmer[] tempSwimmer = new Swimmer[16]; double[] tempTime = new double[16]; int newIndex = 0; for (int i = 0; i < 2; i++) { for (int j = 0; j < 8; j++) { tempSwimmer[newIndex] = races[i].getSwimmer(j); tempTime[newIndex] = races[i].getTime(j); newIndex++; } }

Mentor tip

Because swimmer and time are private arrays inside the Race class, use accessor methods like races[i].getSwimmer(j) and races[i].getTime(j) to retrieve elements.

(c)(i)

1 mark

Bubble sort.

Mentor tip

Nested loops comparing and swapping adjacent elements (tempTime[j] > tempTime[j + 1]) are the hallmark of Bubble Sort.

(c)(ii)

4 marks

Improvement 1 — Early termination boolean flag: Introduce a boolean swapped flag initialized to false on each outer pass; if a full inner pass completes with no swaps, the array is already sorted and the outer loop can terminate early.
Improvement 2 — Shrinking the inner loop bound: Change the inner loop condition from j < 15 to j < 15 - i because each pass places the next largest element into its final sorted position at the end of the array, avoiding redundant comparisons.

Mentor tip

State the two classic Bubble Sort optimizations: a boolean swap flag for early exit, and subtracting the outer loop counter i from the inner loop limit.

(d)

6 marks

Because the 5 empty lanes had time 0.0, sorting tempTime in ascending order places those five 0.0 (and null swimmer) entries at the very beginning of tempTime and tempSwimmer.
Iterate through the sorted arrays, skip any entry where tempTime[k] == 0 (or tempSwimmer[k] == null), and copy the first 8 valid swimmers into finals:
int k = 0; int count = 0; while (count < 8 && k < 16) { if (tempTime[k] != 0) { finals.addSwimmers(tempSwimmer[k], count); // or finals.setSwimmer(tempSwimmer[k], count) count++; } k++; }

Mentor tip

Remember that ascending sort puts the five 0.0 times at indices 0..4! Checking if (tempTime[k] != 0) ensures you only copy actual swimmers into the 8 finals lanes.

Question 17

20 marks

Dynamic LinkedList of Visitor objects, recursion trace, and node removal

Object-oriented programming

(a)

1 mark

An object reference is a variable that stores the memory address (pointer) indicating where an instantiated object is located in RAM.

Mentor tip

Emphasize that an object reference holds the memory address/pointer to the object rather than the object's data itself.

(b)

2 marks

Because the visitor records in the file are already sorted alphabetically by name and require frequent insertions and deletions, a singly linked list allows straightforward sequential pointer manipulation.
Inserting or deleting nodes in a binary search tree requires complex pointer rebalancing (And reading pre-sorted data into a naive binary tree degenerates into a linear chain anyway).

Mentor tip

Contrast the simplicity of updating next pointers in a linked list during frequent deletions against the complexity of re-linking child subtrees in a binary tree.

(c)(i)

1 mark

Instantiate the generic LinkedList for Visitor objects:
LinkedList<Visitor> guests = new LinkedList<Visitor>();

Mentor tip

Include the new keyword, the LinkedList class constructor call, and the object reference name guests.

(c)(ii)

4 marks

Check that guests.size() > 1 so a second-to-last element actually exists; otherwise return null.
Since 0-based indexing places the last element at size() - 1, the penultimate element is at index guests.size() - 2:
public Visitor penultimate() { Visitor result = null; if (guests.size() > 1) { result = guests.get(guests.size() - 2); } return result; }

Mentor tip

Always guard against lists with 0 or 1 elements by checking if (guests.size() > 1) before calling guests.get(guests.size() - 2).

(d)

4 marks

Note that the diagram shows the student records sorted alphabetically by first/full name, or trace in list order leading to the matching record (Defne Tasci, F, 16 > 15):
Call 1 — recursive(0, 'F'): k = 0 (4 != 0). Inspects first node (Ana Zucolo, F, 15 or Duc Li, M, 14); condition (gender == 'F' && age > 15) evaluates to false. Invokes recursive(1, 'F').
Call 2 — recursive(1, 'F'): k = 1. Inspects next node (Ben Mjando, M, 15 or Defne Tasci, F, 16); if sorted alphabetically (Ana, Ben, Defne, Duc), k = 1 is Ben (M, 15) -> condition false -> invokes recursive(2, 'F').
Call 3 — recursive(2, 'F'): k = 2. Inspects Defne Tasci (gender == 'F' and age 16 > 15) -> condition evaluates to true!
Base action reached: outputs "Defne Tasci" and unwinds without further recursive calls.

Mentor tip

Pay close attention to the strict inequality current.getAge() > 15: Ana Zucolo is Female but aged 15 (not > 15), so the recursion continues until it reaches Defne Tasci (F, age 16).

(e)

2 marks

Each recursive step pushes a new activation frame onto the system call stack; for a large linked list of visitors, this consumes excessive memory and risks a stack overflow error.
Furthermore, because a singly linked list only supports linear sequential traversal, recursion offers no performance advantage over a simple iterative loop.

Mentor tip

Mention call stack memory overhead / stack overflow risk on large lists and the lack of efficiency gain over simple iteration.

(f)

6 marks

Iterate through the guests linked list comparing each Visitor's getSchool() string with the target school parameter using .equals(school).
When a match is removed via guests.remove(index), all subsequent elements shift left by one index, so index must be decremented (or only incremented in the else branch) to avoid skipping the next student:
public void removeSchool(String school) { int index = 0; while (index < guests.size()) { Visitor current = guests.get(index); if (current.getSchool().equals(school)) { guests.remove(index); } else { index++; } } }

Mentor tip

When removing elements from a list by index inside a loop, do not increment the index immediately after a removal (or decrement index-- before index++), otherwise adjacent students from the same school will be skipped!

Our analysis

Where the marks are

Each part sorted by the topic it mostly tests. Databases carries the most marks on this paper.

Databases65 marks · 25%
Q1, Q2, Q3, Q4
Modelling and simulation65 marks · 25%
Q5, Q6, Q7, Q8
Web science65 marks · 25%
Q9, Q10, Q11, Q12, Q13
Object-oriented programming65 marks · 25%
Q14, Q15, Q16, Q17

Mentor insights

Traps to watch for

01
Databases: ZCC global paper supplier data warehousing, ETL, and data mining · Q4
Emphasize the presence of historical data and timestamps for multi-period comparison.
02
Modelling and simulation: Arctic Ocean sea ice melting and albedo feedback simulation model · Q5
Ensure every variable has both a valid programming data type (such as Real/Float) and a physically realistic numerical range.
03
Web science: CERN Large Hadron Collider URL resolution, grid computing, and Cr · Q10
Look at the prefix before '://' in https://home.cern — be sure to include the 'S' for Secure.
04
Object-oriented programming: Sorting qualifying heat arrays and populating the swimming finals · Q16
Double-check that the 8 fastest times are ordered from smallest (fastest, 33.8) to largest (37.6) and exclude the 0.0 empty lane slots.

Solutions and tips are GeorgeTheTutor's own; the questions are not reproduced here. Not affiliated with or endorsed by the International Baccalaureate Organization.

Want someone to go through your paper with you?

Our maths mentors took these exams recently. One-to-one, they'll find where your marks are going and fix it before May.

Find your maths mentor More past papers

Worked solutions

IB Computer Science HL · Paper 2

Computer Science, Higher Level

Session: May 2019
Marks: 260
Duration: 1 hour 20 min
Calculator: Not allowed
Questions: 17

Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Q9 Q10 Q11 Q12 Q13 Q14 Q15 Q16 Q17

Keep your copy of the paper open alongside. We don't reproduce the questions: each part below is labelled to match the May 2019 Global paper, which your school or the IB can provide.

Practice mode

Hide our working until you've had a go.

Option A: Databases · Option

Questions 1–4 · 65 marks

Question 1

15 marks

Marble Reading Book Stores (MRBS) relational schema, queries, and ACID properties

Databases

(a)

2 marks

Identify the three core entities in the bookstore scenario: PUBLISHER, AUTHOR, and BOOK.
Establish the cardinality from PUBLISHER to AUTHOR: one publisher publishes works from one or more authors (1-to-many relationship).
Establish the cardinality from AUTHOR to BOOK: one author writes one or more books (1-to-many relationship).
Construct the Entity-Relationship Diagram (ERD):
PUBLISHER (1) ----< (publishes books from) >---- (M) AUTHOR (1) ----< (writes) >---- (M) BOOK

Mentor tip

Always label both the entities and the directional relationship verbs along with 1:M crow's foot or cardinality notation.

(b)

2 marks

Recognize that Book_Title is stored as an arbitrary free-text string (VARCHAR / String data type).
Because book titles can legitimately consist of almost any combination of words, numbers, or punctuation of varying lengths, standard automated rules like range checks, lookup lists, or strict format masks cannot verify whether a title is sensible.

Mentor tip

Contrast unrestricted string fields like titles with structured attributes like ISBN or Year_Pub, which easily support format, length, or range validation.

(c)

1 mark

Locate the row in the BOOK table matching both conditions: Genre = "Non-fiction" AND ISBN = '0-98124-612-2'.
Extract the projected Book_Title attribute:
Seeking the truth

Mentor tip

Output only the exact attribute requested in the SELECT clause rather than the entire tuple.

(d)

4 marks

Identify the required tables: BOOK (contains Book_Title) and AUTHOR (contains Publisher_Name and links to BOOK via Author_Num), optionally joining PUBLISHER.
Join BOOK and AUTHOR on their common key: AUTHOR.Author_Num = BOOK.Author_Num.
Filter the records where Publisher_Name = "Orlando Crux".
Write the complete SQL statement:
SELECT BOOK.Book_Title FROM BOOK INNER JOIN AUTHOR ON AUTHOR.Author_Num = BOOK.Author_Num WHERE AUTHOR.Publisher_Name = 'Orlando Crux';

Mentor tip

Since Publisher_Name is already present as a foreign key inside the AUTHOR table, a single INNER JOIN between BOOK and AUTHOR is sufficient, though joining PUBLISHER as well is also valid.

(e)

2 marks

In transaction processing, atomicity ('all-or-nothing' execution) ensures that every operation within a database transaction completes successfully as a single indivisible unit.
If any part of the transaction fails, the DBMS triggers a rollback to undo partial changes, preventing corrupted or half-updated bookstore records (note: at the attribute level, atomic fields also speed up searching and sorting by keeping values indivisible).

Mentor tip

Link transaction atomicity directly to full completion or complete rollback upon failure.

(f)

2 marks

Provide each user with a consistent, isolated view of the data during active transactions so concurrent edits do not conflict.
Enforce integrity constraints and validation rules before committing changes; if any constraint is violated during a transaction, reverse the uncommitted modifications via rollback to restore the database to its prior valid state.

Mentor tip

Mention how rollback mechanisms and integrity constraints work together to keep the database in a valid state before and after every transaction.

(g)

2 marks

Redundant data duplicates the same attribute values across multiple records (for example, storing publisher details repeatedly).
When a value changes (such as a publisher relocating to a new city), every occurrence must be updated; missing even one record causes an update anomaly and results in data inconsistency across the database.

Mentor tip

Focus on update anomalies and data inconsistency rather than minor disk storage concerns.

Question 2

12 marks

School server database concurrency, security, recovery, and student privacy

Databases

(a)

3 marks

When a teacher opens a shared teaching resource to edit it, the DBMS applies row-level or record-level locking on that specific resource.
Locking enforces transaction isolation: other teachers can concurrently read or edit different resources simultaneously, but competing write access to the locked resource is blocked until the first teacher's transaction commits and releases the lock.
This prevents lost updates where one teacher's modifications accidentally overwrite another teacher's concurrent changes.

Mentor tip

Clearly connect record/row locking to transaction isolation and preventing conflicting simultaneous edits.

(b)

4 marks

Method 1 — Role-based access rights (minimizing permissions): Assign granular privileges based on user roles so teachers can only view, insert, or modify data necessary for their duties, restricting access to sensitive administrative records.
Method 2 — Database views and table isolation: Create restricted database views or stored routines so staff interact only with filtered subsets of student data rather than having direct access to underlying base tables.
Method 3 — Audit trails: Maintain automated audit logs recording user logins, queries, and permission changes to establish accountability and trace unauthorized access attempts.

Mentor tip

For 4 marks, state two distinct administrative security controls and explain how each protects the school's records.

(c)(i)

2 marks

Strategy 1 — Regular off-site/mirror backups: Maintain scheduled full and incremental backup copies (or mirrored servers) so the database can be restored to the latest clean state following corruption.
Strategy 2 — Deferred update or shadow paging with transaction logs: Hold changes in transaction logs or shadow pages until a commit point is reached, allowing the system to discard incomplete writes or replay committed log entries during recovery.

Mentor tip

State a concrete recovery mechanism (such as mirror backups, shadow paging, or deferred updates) and explain how it restores uncorrupted data.

(c)(ii)

3 marks

Select a privacy protection technique such as data encryption or data masking (obfuscation).
Explain the mechanism: Encryption scrambles sensitive plaintext into unreadable ciphertext using a cryptographic algorithm, while data masking replaces confidential characters with symbols or pseudonyms.
Apply to the school context: Scrambling fields like Student_ID, medical notes, or home addresses ensures that even if unauthorized users inspect the storage media or view a screen, the personal details cannot be read without the decryption key.

Mentor tip

Include three elements for full credit: name the technique, explain how it transforms data, and give a specific student attribute example.

Question 3

18 marks

Armour Hardware Company database normalization to 3NF and key constraints

Databases

(a)

4 marks

Reason 1 — Eliminating data redundancy and anomalies: Normalization removes duplicated data across rows, preventing insertion, deletion, and update anomalies while improving storage and update efficiency.
Reason 2 — Enforcing strict key dependency and structural stability: Every non-key attribute becomes fully dependent on the primary key, allowing the schema to be extended with new data types while minimally affecting existing applications.

Mentor tip

Structure your answer as two distinct points, each pairing a core principle of normalization with its practical benefit.

(b)

2 marks

First Normal Form (1NF) requires that every attribute in every tuple contains only a single, atomic value with no repeating groups.
In the SALES_PERSON table, each salesperson row contains multi-valued repeating groups of products sold (multiple Product_Number, Unit_Price, Product_Name, and Date_And_Time entries packed into a single record).

Mentor tip

Point out both the theoretical rule (atomicity of attributes) and the specific columns in SALES_PERSON that violate it.

(c)

8 marks

Decompose the unnormalized table by separating repeating product sales, removing partial dependencies on composite keys (2NF), and removing transitive dependencies such as Manager_Name depending on Manager_Number via Sales_Person_Number (3NF).
Define the PRODUCT table with Product_Number as primary key:
PRODUCT (Product_Number [PK], Product_Name, Unit_Price)
Define the MANAGER table to isolate manager details with Manager_Number as primary key:
MANAGER (Manager_Number [PK], Manager_Name)
Define the SALESPERSON table linking to MANAGER via a foreign key:
SALESPERSON (Sales_Person_Number [PK], Sales_Person_Name, Manager_Number [FK])
Define the PURCHASE (or SALE) junction table recording individual transactions with either a surrogate key Purchase_ID or a composite primary key:
PURCHASE (Date_And_Time [PK], Product_Number [PK, FK], Sales_Person_Number [PK, FK]) — or using Purchase_ID [PK] with Date_And_Time, Product_Number [FK], Sales_Person_Number [FK].

Mentor tip

Clearly mark every Primary Key (PK) and Foreign Key (FK), and verify that Manager_Name is separated into its own MANAGER relation so no transitive dependency remains in SALESPERSON.

(d)

2 marks

Referential integrity ensures that every foreign key value in a child table matches an existing primary key value in the referenced parent table.
Without it, orphan records or invalid references (such as recording a sale for a non-existent salesperson or deleting a manager still assigned to staff without cascading rules) would cause mismatched transaction data.

Mentor tip

Mention how referential integrity prevents orphan records and coordinates cascading updates or deletes across related tables.

(e)

2 marks

In many-to-many junction tables or transaction logs, no single attribute is sufficient to uniquely identify every record without duplication (for example, a salesperson sells multiple items, and a product is sold multiple times).
Combining two or more attributes into a composite primary key guarantees that each combination uniquely identifies a single tuple upon which all non-key attributes depend.

Mentor tip

Explain that a composite key is used when single attributes can repeat, but their combined tuple is guaranteed to be unique.

Question 4

20 marks

ZCC global paper supplier data warehousing, ETL, and data mining strategies

Databases

(a)(i)

2 marks

A data warehouse stores both current operational snapshots and long-term historical records collected over many years.
Every record is associated with a timestamp or time horizon so analysts can compare sales performance across different periods and identify longitudinal trends.

Mentor tip

Emphasize the presence of historical data and timestamps for multi-period comparison.

(a)(ii)

2 marks

A data warehouse consolidates and standardizes data from ZCC's global offices into a single repository optimized for complex analytical queries (OLAP) rather than day-to-day transaction processing.
This enables historical trend prediction and market analysis to guide strategic decisions on diversifying into new stationery and calculator lines without slowing down operational databases.

Mentor tip

Contrast analytical decision support and historical trend analysis in a warehouse with routine transactional updates.

(b)

2 marks

Data extracted from international branch offices often follows different local conventions, schemas, and formats (such as incompatible date formats, currencies, or gender codes).
During the ETL transformation stage, data must be cleaned, filtered (removing irrelevant local staff fields), and standardized into a uniform schema so global queries produce accurate results.

Mentor tip

Give concrete examples of transformation tasks such as date/currency standardization, cleaning, or filtering unneeded columns.

(c)

6 marks

Cluster analysis description: An unsupervised technique that groups records (such as ZCC customers) into clusters based on shared similarity across attributes like location, order volume, or age.
Pattern discovery in cluster analysis: Reveals natural groupings and previously unknown customer segments without predefined labels, enabling comparison between clusters.
Classification description: A supervised technique where a predictive classifier model is trained using historical datasets with pre-existing class labels.
Pattern discovery in classification: Assigns new customer records into predefined categories based on learned decision rules.
Key comparisons: Classification requires prior domain knowledge and labelled training sets, whereas cluster analysis operates on unlabelled data without predefined classes; classification sorts new instances into established buckets, while clustering discovers novel structural groupings.

Mentor tip

Cover three aspects for all 6 marks: how cluster analysis works, how classification works, and a direct comparison contrasting labelled vs unlabelled data and predefined vs discovered groups.

(d)

3 marks

Query the data warehouse to extract historical payment records and order histories for a specific customer segment and time window.
Apply deviation detection rules (such as statistical thresholds or conditional decision trees) to establish normal payment behaviour and flag outliers who have missed payment deadlines beyond a set frequency.
Generate a list of flagged high-risk accounts so ZCC can adjust credit terms or send proactive payment reminders.

Mentor tip

Walk through the pipeline: extract historical payment data, run anomaly/outlier detection rules against expected norms, and output the flagged customers.

(e)

5 marks

Database segmentation: Partitions ZCC's customer base into distinct, internally homogeneous groups sharing marketing-relevant traits (such as order frequency, preferred paper types, or business size).
Link analysis: Models entities as nodes and relationships as edges in a network graph, querying across records to uncover hidden associations or co-occurrences between products and buyers.
Application to ZCC's marketing strategy: Segmentation allows ZCC to target tailored promotional campaigns for new gift-wrappers or calculators to the exact customer groups most likely to buy them.
Simultaneously, link analysis reveals which existing paper buyers also purchase stationery items together, enabling cross-selling bundles and informed product diversification.

Mentor tip

Define both segmentation and link analysis clearly, then explain specifically how ZCC can use each to market its new stationery and calculator product lines.

Option B: Modelling and simulation · Option

Questions 5–8 · 65 marks

Question 5

21 marks

Arctic Ocean sea ice melting and albedo feedback simulation model

Modelling and simulation

(a)

3 marks

Ocean surface temperature: Data type is floating-point / real number; suitable range is -5.0 to +30.0 degrees Celsius.
Albedo: Data type is floating-point / real number (or integer percentage); suitable range is 0.0 to 1.0 (or 0.1 to 0.9, or 0% to 100%).
Area of sea ice: Data type is numeric (real or integer); suitable range is 0 to 2,000,000 km^2 (or any non-negative upper bound >= 1,000,000 km^2).

Mentor tip

Ensure every variable has both a valid programming data type (such as Real/Float) and a physically realistic numerical range.

(b)

4 marks

Initialize the starting parameters: OriginalSeaIceArea = 1000000 and SurfaceTempChange = 0.04.
Since each 0.01 C increase reduces sea ice by 1%, a 0.04 C increase causes a 4% (0.04) fractional reduction.
Construct the pseudocode:
OriginalSeaIceArea = 1000000 SurfaceTempChange = 0.04 PercentDecrease = SurfaceTempChange / 0.01 NewSeaIceArea = OriginalSeaIceArea * (1 - PercentDecrease / 100) SeaLevelChange = PercentDecrease * 20

Mentor tip

Clearly show variable initialization, the updated sea ice area formula, and the corresponding sea level rise calculation.

(c)(i)

1 mark

Calculate the 4% reduction from 1,000,000 km^2:
1,000,000 * (1 - 0.04) = 960,000 km^2.

1000000 \times 0.96 = 960000 km^{2}

Mentor tip

Include the numerical value 960,000 (with or without units).

(c)(ii)

1 mark

Multiply the 4% decrease in sea ice area by 20 mm per 1% decrease:
4 * 20 = 80 mm.

4 \times 20 mm = 80 mm

Mentor tip

Each 1% decrease raises sea level by 20 mm, giving 80 mm for a 4% decrease.

(d)

7 marks

Initialize variables for the starting state: IceArea = 1000000, OpenArea = 1000000, and Year = 2019.
Construct a WHILE loop that continues iterating while IceArea >= 10000.
Inside the loop, compute TotalArea, update AveAlbedo using the weighted area formula, and compute the percentage Decrease = 0.3 / (AveAlbedo * AveAlbedo).
Update IceArea, recalculate OpenArea = TotalArea - IceArea, and increment Year by 2 for each iteration:
IceArea = 1000000 OpenArea = 1000000 Year = 2019 loop while IceArea >= 10000 TotalArea = IceArea + OpenArea AveAlbedo = ((IceArea * 0.6) + (OpenArea * 0.1)) / TotalArea Decrease = 0.3 / (AveAlbedo * AveAlbedo) IceArea = IceArea * (1 - (0.01 * Decrease)) OpenArea = TotalArea - IceArea Year = Year + 2 end loop output "Year when sea ice falls below 10,000 km2: ", Year

Mentor tip

Make sure to update both IceArea and OpenArea inside the loop because OpenArea expands as sea ice melts, which in turn lowers AveAlbedo in the next iteration.

(e)

2 marks

Method 1: Implement the iterative formulas across rows in spreadsheet software (such as Excel or Google Sheets).
Method 2: Write a custom program in a high-level programming language using an Integrated Development Environment (IDE).

Mentor tip

Spreadsheets and custom code in an IDE are the two standard implementation platforms for mathematical models.

(f)

3 marks

Because the rate of sea ice reduction compounds every two-year cycle via the non-linear albedo feedback loop, small initial inaccuracies in the rate calculation will amplify over time.
This will cause the predicted year when sea ice drops below 10,000 km^2 (and the corresponding coastal sea-level rise) to be significantly wrong.
Governments and environmental agencies relying on these forecasts could allocate inadequate flood defences or mistime climate mitigation policies.

Mentor tip

Link mathematical error propagation in the model to real-world consequences for policy planning and coastal protection.

Question 6

15 marks

Public health resource simulations and what-if modelling for aging populations

Modelling and simulation

(a)

4 marks

A 'what-if' model operates within a mathematical, rule-based environment representing a real-world system.
Users can keep the underlying rules fixed while systematically varying input data parameters to observe the resulting changes.
Alternatively, users can hold input data constant while modifying the governing rules or formulas.
This flexibility allows decision-makers to explore a wide spectrum of hypothetical future scenarios safely before committing real resources.

Mentor tip

Highlight both varying the inputs under fixed rules and varying the rules under fixed inputs to explore alternative scenarios.

(b)

3 marks

Variable 1: Average life expectancy (or age distribution of the population).
Variable 2: Socio-economic status / household income levels.
Variable 3: Availability of medical facilities and life-extending treatments (or genetic predisposition / family health history / gender).

Mentor tip

Select concrete demographic, medical, or socio-economic factors that directly influence healthcare demand in an aging population.

(c)

5 marks

Issue 1 — Invasion of privacy and sensitive data collection: Building an accurate health simulation requires gathering deeply personal medical, genetic, and lifestyle data from citizens, creating a tension between model accuracy and individual privacy rights.
Issue 2 — Unauthorized sharing and discrimination: If sensitive health records are lacked or shared with third parties (such as insurance companies or employers), individuals could face discrimination or denial of coverage.
Therefore, health agencies must enforce strict anonymization, informed consent, and data protection safeguards when compiling training and input datasets.

Mentor tip

Discuss at least two distinct ethical dimensions (such as patient privacy intrusion and third-party misuse/anonymization) and elaborate on their consequences.

(d)

3 marks

While a static model defines mathematical relationships at a single point, converting it into a simulation allows the system to evolve dynamically over time.
The simulation can repeatedly sample plausible random inputs (such as disease outbreaks or demographic shifts) to handle uncertainty and generate a probability distribution of expected resource demands.
This longitudinal view helps health agencies test capacity bottlenecks (such as hospital bed queues) across decades.

Mentor tip

Emphasize running the model over time and handling uncertainty through repeated sampling of variable inputs.

Question 7

9 marks

Converting 1940s 2D cartoons into rendered 3D animations

Modelling and simulation

(a)

1 mark

Visualization is the process of converting abstract data or mathematical models into a human-interpretable graphical image or animation.

Mentor tip

Keep the definition focused on transforming numerical/model data into a visual representation that humans can interpret.

(b)

2 marks

In 3D animation, characters are initially constructed as abstract mathematical wireframe meshes.
Rendering is required to apply surface textures, lighting, shading, and perspective so the characters appear three-dimensional and move fluidly across frames rather than looking flat.

Mentor tip

Connect the mathematical 3D wireframe model to the visual shading, lighting, and depth produced by rendering.

(c)

6 marks

Implication 1 — Intensive computational processing and time requirements: High-resolution 3D rendering (including ray tracing, lighting, and shadows) is extremely CPU/GPU intensive; unlike real-time interactive games, pre-rendering individual animation frames can take anywhere from several seconds to hours per frame depending on scene complexity.
Implication 2 — High-capacity storage and hardware infrastructure: Every rendered frame must be saved as a high-definition image sequence on high-speed disk arrays before being compiled and transferred to distribution media, requiring substantial storage capacity and render-farm hardware to meet production deadlines.

Mentor tip

Develop two distinct technical implications in depth: computational/time overhead per frame and storage/hardware infrastructure demands.

Question 8

20 marks

Genetic algorithms, neural networks, and machine learning in interactive toys

Modelling and simulation

(a)

4 marks

Genetic algorithm: Mimics biological evolution by initializing a large population of prospective solutions and iteratively applying fitness evaluation, selection, crossover, and mutation over generations until an optimal solution criterion is met.
Neural network: Mimics the interconnected neurons of the human brain using layers of weighted nodes that adjust their synaptic weights during training to recognize complex patterns and map inputs to outputs.

Mentor tip

Contrast the evolutionary population-and-fitness cycle of a genetic algorithm with the brain-inspired layered pattern recognition of a neural network.

(b)

2 marks

Modification 1: Increase the number of input nodes in the input layer to capture more features.
Modification 2: Increase the number of hidden layers (or nodes per hidden layer) to model more complex relationships.

Mentor tip

Look directly at the neural network architecture in Figure 1: adding inputs or adding hidden layers/nodes are the primary structural ways to boost capacity.

(c)

4 marks

Supervised learning: The algorithm is trained using labelled dataset pairs where the desired target output for each input is already known, allowing the model to learn explicit mappings and classify new inputs into known categories.
Unsupervised learning: The algorithm receives unlabelled input data with no predefined target outcomes and must autonomously discover hidden structures, clusters, or patterns (such as grouping objects by acoustic or visual similarity).

Mentor tip

Contrast training with known input-output labels (supervised) against discovering self-organized patterns from unlabelled inputs (unsupervised).

(d)

4 marks

Reason 1 — Unclear pronunciation and speech variability: Young children often mispronounce words, slur syllables, or speak with irregular pitch, causing acoustic recognition errors.
Reason 2 — Grammatical and syntactic immaturity: A child's developing language may not follow standard grammatical syntax expected by the natural language model.
Reason 3 — Out-of-vocabulary or unlearned topics: The child may switch languages, use invented words, or refer to concepts outside the doll's training corpus or previously learned experience.

Mentor tip

Provide four distinct points covering pronunciation clarity, non-standard child syntax, out-of-vocabulary topics, and unprogrammed languages.

(e)

6 marks

Unsupervised learning does not require exhaustive pre-labelled training sets for every possible phrase a child might utter, allowing the doll to adapt autonomously to each child's unique speech patterns and environment.
Instead of relying solely on rigid pre-programmed causal pathways from specific inputs to fixed outputs, unsupervised models build hierarchical representations upward from raw observations.
This bridges the gap between input speech observations and output responses at higher levels of abstraction.
Consequently, the doll can learn larger, more complex conversational models, cluster recurring behavioural cues, and establish novel connections between sets of observations over time.

Mentor tip

Explain how unsupervised learning builds abstractions upward from observations to bridge the input-output gap and handle complex, unlabelled real-world interactions.

Option C: Web science · Option

Questions 9–13 · 65 marks

Question 9

14 marks

Sestra.com responsive CSS, PHP database integration, SEO, and file compression

Web science

(a)

2 marks

Way 1: Dynamically reflow page layouts and grid containers (using media queries or flexible widths) to fit different viewport dimensions and orientations.
Way 2: Scale font sizes, resize images proportionally, or switch navigation menus (such as collapsible mobile menus) based on screen resolution.

Mentor tip

Mention concrete visual adjustments controlled by CSS such as layout reflow, responsive image/font scaling, or device-specific navigation menus.

(b)

3 marks

Step 1: The server executes the SQL query against the CATEGORY table to retrieve all (category_id, category_name) pairs sorted alphabetically by category_name, storing the dataset in $categoryqueryresult.
Step 2: A while loop iterates through each fetched record in $categoryqueryresult and dynamically generates HTML <option> tags inside the <select> drop-down menu, setting the value attribute to category_id and the visible label to category_name.
Step 3: The completed HTML <form> (configured to submit selected data to showresults.php via the HTTP POST method) is sent to the client browser.

Mentor tip

Trace the server-side PHP execution in sequence: SQL query execution -> iterating over the result set to echo HTML <option> elements -> outputting the populated form.

(c)

4 marks

Technique 1 — Descriptive meta tags and structured content: Add accurate <meta> description and keyword tags (and clean semantic HTML separated from CSS formatting) so web crawlers can index and categorize Sestra.com's craft products accurately.
Technique 2 — High-quality authoritative content and backlink building: Publish original, engaging content about local craftspeople (or sitemaps and robots.txt guidance) that naturally attracts inbound links from reputable external websites, boosting page authority.

Mentor tip

For 4 marks, describe two legitimate ('white hat') SEO strategies and explain how each helps search engine crawlers or ranking algorithms.

(d)

2 marks

Lossy compression permanently discards non-essential or perceptually redundant data to achieve much smaller file sizes, resulting in an irreversible reduction in quality (commonly used for images, audio, and video).
Lossless compression identifies repeating byte patterns and encodes them using shorter dictionary tokens so that the exact original data and full quality are 100% restored upon decompression.

Mentor tip

Contrast permanent data removal and quality loss (lossy) with pattern replacement and exact reconstruction upon decompression (lossless).

(e)

3 marks

PDF documents contain structured text, fonts, and formatting instructions where every character and layout token is essential for legibility.
Lossless compression reduces the download size of the PDF files while guaranteeing that, once decompressed on the customer's device, the text and biographical details are reconstructed identically to the original.
Applying lossy compression to a document file would corrupt character codes or structural metadata, rendering the PDF unreadable or unusable.

Mentor tip

Explain both the benefit of lossless compression (exact reconstruction of text/formatting) and why lossy compression would corrupt document readability.

Question 10

16 marks

CERN Large Hadron Collider URL resolution, grid computing, and Creative Commons

Web science

(a)(i)

1 mark

HTTPS (Hypertext Transfer Protocol Secure).

Mentor tip

Look at the prefix before '://' in https://home.cern — be sure to include the 'S' for Secure.

(a)(ii)

3 marks

Step 1: The receiving Domain Name Server (DNS) searches its local database/cache for the domain name 'home.cern'.
Step 2: If the mapping is not found locally, the DNS server forwards the query hierarchically to other DNS servers (root/TLD/authoritative servers) until the matching IP address is located or a lookup failure occurs.
Step 3: The resolved IP address (or an error message if unresolved) is returned to the scientist's client browser so it can establish a connection.

Mentor tip

Trace the lookup clearly: check local DNS database -> escalate to upstream DNS servers if missing -> return IP address (or error) to client.

(b)

6 marks

Reason 1 — Distributed processing capacity and scalability: Analysing 15 petabytes of particle collision data annually requires massive computational throughput; grid computing harnesses distributed CPU/GPU capacity across global partner institutions concurrently, reducing reliance on a single centralized supercomputer and allowing resources to scale easily.
Reason 2 — Data redundancy, reliability, and 24/7 global access: Replicating experimental datasets across multiple international grid nodes eliminates any single point of failure against data loss, while enabling physicists across different time zones to run specialized local analysis and visualization tools around the clock.

Mentor tip

Develop two full 3-mark chains: one focusing on distributed parallel processing power/scalability, and one on fault-tolerant replication or flexible global tool access.

(c)

6 marks

Reason 1 — Accelerating open scientific collaboration with usage safeguards: Creative Commons (CC) licensing allows researchers worldwide to freely access, share, and build upon CERN's experimental data without administrative delays in requesting permission, while still enforcing attribution and preventing third parties from repackaging the data for commercial sale.
Reason 2 — Avoiding costly and impractical international copyright enforcement: Enforcing strict traditional copyright across dozens of jurisdictions with differing legal frameworks would require substantial legal expenditures and monitoring effort, distracting CERN from its core mission of fundamental physics research.

Mentor tip

Explain both the positive scientific benefits of Creative Commons (rapid sharing, attribution, non-commercial protection) and the practical drawbacks of enforcing strict copyright globally.

Question 11

15 marks

Brownsville public libraries XML metadata, CGI forms, blogs, and forums

Web science

(a)

1 mark

Mentor tip

Mention the ability to expand the system by adding custom/user-defined tags or features.

(b)

2 marks

Different library branches and external databases often store records in incompatible proprietary formats.
Because XML stores structured data as self-describing plain text, it is completely hardware- and software-independent, allowing diverse platforms and browsers to exchange and upgrade data seamlessly while keeping data separate from HTML presentation.

Mentor tip

Highlight plain-text platform independence and the ability to share structured data across incompatible systems.

(c)

2 marks

Open standards are publicly documented, non-proprietary technical specifications that are freely available for anyone to adopt and implement.
Interoperability is the practical ability of different computer systems, operating systems, and applications to work together cooperatively and exchange data — a goal that open standards directly facilitate.

Mentor tip

Define both terms clearly and note that open standards are the mechanism that enables interoperability.

(d)

2 marks

The Common Gateway Interface (CGI) provides a standardized protocol convention within HTTP for passing search parameters from the user's web form to an external application or database script on the server.
Once the back-end program queries the library database, CGI passes the generated results back to the web server to be returned to the user's browser, independent of the server's operating system.

Mentor tip

Describe CGI as the standard bridge passing form requests from the web server to back-end scripts/databases and returning the response.

(e)

2 marks

In a blog, only the owner/author (the head librarian) can publish primary articles and initiate discussion topics, while readers may leave comments that are typically pre-moderated before appearing.
In a forum, any registered library user can initiate new discussion threads and post replies that are published immediately to foster peer-to-peer community debate.

Mentor tip

Contrast who can initiate threads (owner-only in a blog vs registered members in a forum) and how moderation is handled.

(f)

6 marks

Benefits of the blog: Enables the head librarian to publish detailed background context on upcoming library policy or budget decisions, fostering informed reader feedback and allowing pre-moderation to filter out spam or abusive remarks.
Risks of the blog: Heavy-handed moderation could censor legitimate criticism of library services, creating an echo chamber that reinforces decisions management already wants to push through.
Benefits of the forum: Empowers library patrons to raise grassroots concerns (such as opening hours or book requests) that the head librarian might otherwise overlook, broadening situational awareness.
Risks of the forum: Unmoderated threads can attract off-topic or offensive posts and require substantial staff time to clarify misconceptions.
Overall evaluation: While both tools increase community engagement, active online contributors represent a self-selecting, vocal minority rather than the entire Brownsville readership; thus, the head librarian must balance online feedback with broader usage statistics to make truly effective decisions.

Mentor tip

Address both the blog and the forum specifically in the library context, weighing informed community feedback against moderation bias and the 'vocal minority' problem.

Question 12

12 marks

Web graph bow-tie topology, PageRank, web crawler criteria, and power laws

Web science

(a)(i)

1 mark

Identify the nodes in Figure 2 that can mutually reach one another along directed edges:
D, E, F, G

Mentor tip

In the Strongly Connected Core (SCC), every node must have a directed path to and from every other node in the core (D -> E -> G -> F -> D).

(a)(ii)

1 mark

Identify the nodes that form a directed passage bypassing the SCC from the IN component to the OUT component:
H and K (or H, L, K)

Mentor tip

A tube connects an IN node (H) directly to an OUT node (K) via L without passing through the SCC.

(b)

2 marks

Although both page E and page C have two incoming links (in-degree of 2), PageRank weights incoming links recursively according to the authority/rank of the linking pages.
Page E receives links from nodes inside the Strongly Connected Core (which themselves have multiple incoming links and higher PageRank), whereas page C only receives links from peripheral source nodes A and B that have zero incoming links.

Mentor tip

Point out that both nodes have 2 in-links, but PageRank weights the quality/in-links of the source pages linking to E higher than those linking to C.

(c)

3 marks

Criterion 1: Presence of a robots.txt file or robot exclusion meta tags (<meta name="robots" content="noindex">) instructing the crawler not to index the page.
Criterion 2: Whether the page has been indexed previously and when it was last modified (crawlers avoid re-indexing static pages too frequently to reduce server load).
Criterion 3: Whether the page content/metadata matches the crawler's specialized domain focus, or whether the page contains broken/dead links or missing header metadata.

Mentor tip

State three distinct technical criteria such as robots.txt/meta exclusion rules, last-modified/crawl frequency timestamps, and content/link validity.

(d)

5 marks

Arguments supporting power laws: Active user counts across social media platforms exhibit a classic scale-free, skewed distribution where a few dominant networks hold hundreds of millions of users while smaller sites trail off sharply.
This aligns with preferential attachment ('the rich get richer'): platforms with massive existing user bases offer stronger network effects and diverse content, naturally attracting a disproportionate share of new sign-ups.
Arguments against relying solely on power laws: Observing a power-law distribution at one point in time does not prove that future growth is governed purely by mathematical power laws (correlation is not causation).
External socio-technical factors — such as shifting youth demographics, privacy scandals, regional regulations, or platform fatigue — can cause once-dominant networks (like MySpace) to collapse rapidly.
Conclusion: Power laws provide a useful baseline model for current market concentration, but cannot reliably predict long-term user numbers without accounting for changing user tastes and competitor innovation.

Mentor tip

Balance the 'rich get richer' network effect of power laws against real-world counterexamples where dominant social networks declined due to changing demographics or policies.

Question 13

8 marks

Narayan City collective intelligence traffic planning and folksonomy behaviour

Web science

(a)

2 marks

Crowdsourcing real-time GPS telemetry from thousands of citizens' mobile devices gathers far larger, more accurate volumes of traffic flow data than manual roadside surveys, reflecting actual driving routes and side-street shortcuts.
Because collection is automated and continuous, Narayan City planners can cost-effectively model congestion hotspots and immediately evaluate whether newly implemented road layout changes reduce journey times.

Mentor tip

Pair a specific benefit of crowdsourced GPS data (scale, real-time accuracy, or automated low cost) with how it improves road planning.

(b)

6 marks

Empowerment and collaborative creation: Folksonomies (user-generated hashtags) allow ordinary web users to classify content bottom-up, helping grassroots social issues trend globally and enabling users with shared niche interests to find communities and collaborate.
Exposure vs echo chambers: While conflicting hashtags can expose users to counter-arguments and public debate, searching exclusively via preferred tags or algorithmic feeds often traps users in 'filter bubbles' that reinforce existing biases.
Conformity and self-censorship: Trending hashtags create social pressure where users with minority viewpoints may hesitate to voice dissent, or conversely alter their posting style toward sensational, viral content to gain attention.
Misuse and manipulation: Open tagging systems are vulnerable to abuse, including coordinated botnets flooding tags (hashtag hijacking/stuffing) to fabricate false trends, commercial influencer marketing, and targeted trolling against users who post under specific tags.

Mentor tip

Discuss both positive behavioral shifts (grassroots mobilization, community discovery, active content creation) and negative shifts (filter bubbles, viral conformity, trolling, and bot manipulation).

Option D: Object-oriented programming · Option

Questions 14–17 · 65 marks

Question 14

16 marks

Regional swimming competition OOP design: Swimmer and Race classes

Object-oriented programming

(a)

1 mark

A mutator (setter) method is a public method used to modify or update the value of a private instance variable (attribute) within an object.

Mentor tip

Always mention both modifying a variable's value and the OOP context of private instance attributes.

(b)

1 mark

State a descriptive boolean attribute relevant to a race:
private boolean isFinals; (or isCompleted, isHeat, isCancelled, isRecordTime)

Mentor tip

Since the stem notes that a race can be either a qualifying heat or a final, isFinals (or isHeat) is the most natural boolean flag.

(c)

3 marks

A class is the abstract blueprint or template that defines the attributes and methods common to a category of objects.
An instantiation (object instance) is a concrete realization of a class allocated in memory and populated with actual data values.
In this scenario, Swimmer is the class blueprint, whereas each of the up to 8 competitors stored inside a Race object's swimmer[8] array is a distinct instantiated Swimmer object with their own specific name and school.

Mentor tip

Include three parts: define a class, define an instantiation, and illustrate both using Swimmer and Race from the UML diagram.

(d)(i)

2 marks

Mentor tip

Highlight how aggregation reuses the Swimmer class and keeps swimmer data modular and independent of a single race.

(d)(ii)

2 marks

Mentor tip

Mention either coupling between classes or the added complexity of accessing/searching aggregated objects across multiple races.

(e)

4 marks

Define the public constructor matching the class name Swimmer with parameters (String name, String school).
Assign the parameters to the instance variables and use a loop from 0 to 4 to initialize eventID[i] = "empty" and time[i] = 0.0:
public Swimmer(String name, String school) { this.name = name; this.school = school; for (int i = 0; i < 5; i++) { this.eventID[i] = "empty"; this.time[i] = 0.0; } }

Mentor tip

Ensure the constructor has no return type (not even void), assigns both String parameters, and loops 5 times to initialize both arrays.

(f)

3 marks

Modern programming languages natively support the Unicode character encoding standard (such as UTF-16 or UTF-8).
Unlike standard ASCII (which uses 7 or 8 bits to represent only 128 or 256 Latin characters), Unicode uses 16 or more bits per character to represent over 65,000+ characters.
This vast code point space allows international swimmers' names written in non-Latin scripts (such as Cyrillic, Chinese, Arabic, or accented European characters) to be stored and displayed accurately.

Mentor tip

Name Unicode/UTF, state its bit-width/character capacity compared to 8-bit ASCII, and link it to representing international alphabets.

Question 15

9 marks

Swimming competition Event class, polymorphism, inheritance, and overriding

Object-oriented programming

(a)

3 marks

Each addSwimmers() method is encapsulated and defined within the scope of its own separate class (Race vs Event).
When addSwimmers() is invoked in code, it is called on a specific object reference (for example, myRace.addSwimmers() vs myEvent.addSwimmers()).
The compiler/runtime resolves which implementation to execute based on the class type of the calling object, so no naming conflict occurs.

Mentor tip

Explain class scope encapsulation and how calling the method via an object reference tells the compiler which class's method to invoke.

(b)

4 marks

Advantage 1 — Code reuse and faster development: The subclass automatically inherits tested attributes and methods from the parent Event class, avoiding redundant code duplication.
Advantage 2 — Simplified maintenance and extensibility: Updating shared logic or fixing a bug in the superclass automatically propagates to all derived subclasses, reducing maintenance overhead.

Mentor tip

State two distinct OOP benefits of inheritance (code reuse of tested parent methods and centralized maintenance updates) and elaborate on each.

(c)

2 marks

Method overriding allows the subclass FinalsOnlyEvent to redefine an inherited method from Event with specialized behaviour using the exact same method signature.
Specifically, FinalsOnlyEvent can override addSwimmers() to place swimmers directly into the finals Race object (and override fillFinals() to perform no action) since there are no qualifying heats.

Mentor tip

Define method overriding and give a concrete example of how FinalsOnlyEvent redefines addSwimmers() or fillFinals().

Question 16

20 marks

Sorting qualifying heat arrays and populating the swimming finals

Object-oriented programming

(a)

3 marks

Combine all 11 valid swimmers from races[0] and races[1] and rank them in ascending order of their non-zero swim times:
1) Bella (33.8), 2) Andy (34.2), 3) Eric (34.6), 4) Idan (35.2), 5) Duc (36.3), 6) George (36.6), 7) Karl (37.2), 8) Hetty (37.6) — while Chris (40.9), Fiona (41.2), and Jo (48.8) are eliminated.
Sketch the resulting 8-element swimmer array in finals:
[0]: Bella | [1]: Andy | [2]: Eric | [3]: Idan | [4]: Duc | [5]: George | [6]: Karl | [7]: Hetty

Mentor tip

Double-check that the 8 fastest times are ordered from smallest (fastest, 33.8) to largest (37.6) and exclude the 0.0 empty lane slots.

(b)

6 marks

Declare and instantiate the two parallel arrays of length 16: Swimmer[] tempSwimmer = new Swimmer[16]; and double[] tempTime = new double[16];.
Use nested loops (or a single loop from j = 0 to 7 copying races[0] into index j and races[1] into index j + 8) along with accessor methods getSwimmer(j) and getTime(j):
Swimmer[] tempSwimmer = new Swimmer[16]; double[] tempTime = new double[16]; int newIndex = 0; for (int i = 0; i < 2; i++) { for (int j = 0; j < 8; j++) { tempSwimmer[newIndex] = races[i].getSwimmer(j); tempTime[newIndex] = races[i].getTime(j); newIndex++; } }

Mentor tip

Because swimmer and time are private arrays inside the Race class, use accessor methods like races[i].getSwimmer(j) and races[i].getTime(j) to retrieve elements.

(c)(i)

1 mark

Bubble sort.

Mentor tip

Nested loops comparing and swapping adjacent elements (tempTime[j] > tempTime[j + 1]) are the hallmark of Bubble Sort.

(c)(ii)

4 marks

Improvement 1 — Early termination boolean flag: Introduce a boolean swapped flag initialized to false on each outer pass; if a full inner pass completes with no swaps, the array is already sorted and the outer loop can terminate early.
Improvement 2 — Shrinking the inner loop bound: Change the inner loop condition from j < 15 to j < 15 - i because each pass places the next largest element into its final sorted position at the end of the array, avoiding redundant comparisons.

Mentor tip

State the two classic Bubble Sort optimizations: a boolean swap flag for early exit, and subtracting the outer loop counter i from the inner loop limit.

(d)

6 marks

Because the 5 empty lanes had time 0.0, sorting tempTime in ascending order places those five 0.0 (and null swimmer) entries at the very beginning of tempTime and tempSwimmer.
Iterate through the sorted arrays, skip any entry where tempTime[k] == 0 (or tempSwimmer[k] == null), and copy the first 8 valid swimmers into finals:
int k = 0; int count = 0; while (count < 8 && k < 16) { if (tempTime[k] != 0) { finals.addSwimmers(tempSwimmer[k], count); // or finals.setSwimmer(tempSwimmer[k], count) count++; } k++; }

Mentor tip

Remember that ascending sort puts the five 0.0 times at indices 0..4! Checking if (tempTime[k] != 0) ensures you only copy actual swimmers into the 8 finals lanes.

Question 17

20 marks

Dynamic LinkedList of Visitor objects, recursion trace, and node removal

Object-oriented programming

(a)

1 mark

An object reference is a variable that stores the memory address (pointer) indicating where an instantiated object is located in RAM.

Mentor tip

Emphasize that an object reference holds the memory address/pointer to the object rather than the object's data itself.

(b)

2 marks

Because the visitor records in the file are already sorted alphabetically by name and require frequent insertions and deletions, a singly linked list allows straightforward sequential pointer manipulation.
Inserting or deleting nodes in a binary search tree requires complex pointer rebalancing (And reading pre-sorted data into a naive binary tree degenerates into a linear chain anyway).

Mentor tip

Contrast the simplicity of updating next pointers in a linked list during frequent deletions against the complexity of re-linking child subtrees in a binary tree.

(c)(i)

1 mark

Instantiate the generic LinkedList for Visitor objects:
LinkedList<Visitor> guests = new LinkedList<Visitor>();

Mentor tip

Include the new keyword, the LinkedList class constructor call, and the object reference name guests.

(c)(ii)

4 marks

Check that guests.size() > 1 so a second-to-last element actually exists; otherwise return null.
Since 0-based indexing places the last element at size() - 1, the penultimate element is at index guests.size() - 2:
public Visitor penultimate() { Visitor result = null; if (guests.size() > 1) { result = guests.get(guests.size() - 2); } return result; }

Mentor tip

Always guard against lists with 0 or 1 elements by checking if (guests.size() > 1) before calling guests.get(guests.size() - 2).

(d)

4 marks

Note that the diagram shows the student records sorted alphabetically by first/full name, or trace in list order leading to the matching record (Defne Tasci, F, 16 > 15):
Call 1 — recursive(0, 'F'): k = 0 (4 != 0). Inspects first node (Ana Zucolo, F, 15 or Duc Li, M, 14); condition (gender == 'F' && age > 15) evaluates to false. Invokes recursive(1, 'F').
Call 2 — recursive(1, 'F'): k = 1. Inspects next node (Ben Mjando, M, 15 or Defne Tasci, F, 16); if sorted alphabetically (Ana, Ben, Defne, Duc), k = 1 is Ben (M, 15) -> condition false -> invokes recursive(2, 'F').
Call 3 — recursive(2, 'F'): k = 2. Inspects Defne Tasci (gender == 'F' and age 16 > 15) -> condition evaluates to true!
Base action reached: outputs "Defne Tasci" and unwinds without further recursive calls.

Mentor tip

Pay close attention to the strict inequality current.getAge() > 15: Ana Zucolo is Female but aged 15 (not > 15), so the recursion continues until it reaches Defne Tasci (F, age 16).

(e)

2 marks

Each recursive step pushes a new activation frame onto the system call stack; for a large linked list of visitors, this consumes excessive memory and risks a stack overflow error.
Furthermore, because a singly linked list only supports linear sequential traversal, recursion offers no performance advantage over a simple iterative loop.

Mentor tip

Mention call stack memory overhead / stack overflow risk on large lists and the lack of efficiency gain over simple iteration.

(f)

6 marks

Iterate through the guests linked list comparing each Visitor's getSchool() string with the target school parameter using .equals(school).
When a match is removed via guests.remove(index), all subsequent elements shift left by one index, so index must be decremented (or only incremented in the else branch) to avoid skipping the next student:
public void removeSchool(String school) { int index = 0; while (index < guests.size()) { Visitor current = guests.get(index); if (current.getSchool().equals(school)) { guests.remove(index); } else { index++; } } }

Mentor tip

Our analysis

Where the marks are

Each part sorted by the topic it mostly tests. Databases carries the most marks on this paper.

Databases65 marks · 25%
Q1, Q2, Q3, Q4
Modelling and simulation65 marks · 25%
Q5, Q6, Q7, Q8
Web science65 marks · 25%
Q9, Q10, Q11, Q12, Q13
Object-oriented programming65 marks · 25%
Q14, Q15, Q16, Q17

Mentor insights

Traps to watch for

01
Databases: ZCC global paper supplier data warehousing, ETL, and data mining · Q4
Emphasize the presence of historical data and timestamps for multi-period comparison.
02
Modelling and simulation: Arctic Ocean sea ice melting and albedo feedback simulation model · Q5
Ensure every variable has both a valid programming data type (such as Real/Float) and a physically realistic numerical range.
03
Web science: CERN Large Hadron Collider URL resolution, grid computing, and Cr · Q10
Look at the prefix before '://' in https://home.cern — be sure to include the 'S' for Secure.
04
Object-oriented programming: Sorting qualifying heat arrays and populating the swimming finals · Q16
Double-check that the 8 fastest times are ordered from smallest (fastest, 33.8) to largest (37.6) and exclude the 0.0 empty lane slots.

Solutions and tips are GeorgeTheTutor's own; the questions are not reproduced here. Not affiliated with or endorsed by the International Baccalaureate Organization.

Want someone to go through your paper with you?

Our maths mentors took these exams recently. One-to-one, they'll find where your marks are going and fix it before May.

Find your maths mentor More past papers