CDC | The Event Loop & Matia

18:00 - Mingling 🎉, pizza 🍕, and beers 🍻

18:30 - lectures

Write Once, Publish Twice (Maybe)by Yedidya Rashi

Your service saves an order, then crashes before telling anyone. Or it publishes twice. Or someone renames a column and breaks three consumers. Change Data Capture (CDC) and the Transactional Outbox both promise to keep your database and your events in sync, but they fail in very different ways.

This talk starts from the basics, so no prior knowledge is needed.

We'll walk through how each pattern works, then look at where each one hurts: duplicate and out-of-order events, schema changes leaking into business contracts, replication slots that fall behind, and outbox tables that grow without bound. You'll leave with a practical way to choose between them, or to combine them.

Capturing a database change is easy. Turning it into a reliable business event is the real challenge.---Zero Latency, Zero Privacy: The CDC Trap Nobody Warned You Aboutby Oren Efraim

Replacing traditional batch polling with Change Data Capture (CDC) solved our latency bottlenecks, but it silently opened a backdoor to one of the subtlest privacy threats: social linkage attacks. Drawing lessons from the infamous Netflix Prize de-anonymization disaster, this talk exposes how continuous, row-level streaming turns innocent millisecond timestamps and real-time DELETE events into high-entropy digital fingerprints that unmask "anonymized" users across staging lakes and public feeds. We will explore practical, in-flight architectural defenses, from connector-level transformations and timestamp coarsening to crypto-shredding, ensuring that regardless of the architecture we build, protecting user privacy remains our core engineering responsibility.