An interactive study companion for Martin Kleppmann's classic on reliable, scalable, and maintainable systems.
Data systems have societal impact — predictive analytics, privacy, and the ethics of data collection.
Building data systems is not value-neutral. Recommendation engines shape behavior. Predictive policing encodes historical bias. Tracking infrastructure enables surveillance. Engineers must consider who benefits, who is harmed, and what data is truly necessary.
Collect minimum data, encrypt at rest and in transit, provide deletion mechanisms, and make data practices transparent to users.
GDPR requires deletion APIs and data processing agreements — engineering teams implement right-to-erasure in PostgreSQL with soft-delete + purge jobs. AWS offers region-locked storage (data residency) for regulated industries. Snowflake column masking and row access policies enforce authorization at the warehouse layer. Recommendation systems at Netflix and TikTok raise ongoing questions about filter bubbles and engagement optimization vs user wellbeing.
Recommendation models trained on viewing history must honor GDPR deletion requests — purge user rows from warehouses and retrain without retained PII. OAuth scopes limit which services access profile data.
// Netflix / Auth0 — OAuth 2.0 authorization code flow for user identity
const authUrl = oauth.authorizeUrl({
clientId: process.env.OAUTH_CLIENT_ID!,
redirectUri: "https://app.example/callback",
scope: "openid profile email",
state: crypto.randomUUID(),
});
// Token exchange happens server-side; access token scoped to minimum permissions