When young engineers and data scientists learn about databases they are always presented with examples that use non-termporal models and update data in-place. Modelling time is considered hard and is often avoided in database courses. Many of us carry this bias for decades into our careers. When faced with challenges of audit and data analytics that require understanding of how data evolved over time we often rush to counter the loss of temporality using techniques like periodic snapshots, change data capture, and historizations, without fully realizing their flaws.

The mismatch between non-temporal services (OLTP) and temporal analytics systems (OLAP) is the main point of contention between software and data engineers in many companies. It leads to multiple reworks and expensive system redesigns, results in high complexity, and prevents us from using data efficiently.

Like with any other information, it's much easier to preserve temporality than to reconstruct it later. In this talk we will look at the evolution of temporal modelling techniques that happened in parallel in software, data analytics, version control, and blockchain domains. We will see how they are all about to converge on a key set of principles that just might be the one "true form" of data we all seek.

We will also look at the tools currently available for working with temporal data to highlight major flaws in batch processing and peek at the next generation of tools that will take us towards real-time and more reliable data.