The theme of @CMUDB's ongoing Future Data Systems seminar series is "what should a database system look like in the future to handle more of our past so we never forget" - clearly I'm biased, but the answer has to be @xtdb_com ! Join us live on 24th November to learn more...

I'm giving an online tech talk with @FINOSFoundation this Wednesday (4pm GMT) on 'as-of reporting' in financial services - but really this applies to all organisations who care about the auditability of systems. XTDB helps you to quickly figure out how data has changed and why /1

Model training requires accurate history, right?

Ultimately though I suspect the 'best' protocol will be very data & situation dependent. It feels like all the same set of trade-offs you face when thinking about optimal LSM-trees and compression more generally

@rwmarti @lukaseder

> recalculate the indexes on each insert In principle this shouldn't be a blocker. Simpler is better if there's any hope for a database to accelerate the workload. Like it might be okay *if* you could tune an LSM index to match the workload. > no free lunch aka RUM Conjecture

Just in case anyone was in doubt about the stability argument, Q1 (1974) still works great: dl.acm.org/doi/pdf/10.1145/8…

SQL is turning 50 years old later this week 🎉 In your opinion, which are the best bits? Which are the worst?

Related - I've only just realised that @MapsAsCode now supports Iterations 😍 docs.onlinewardleymaps.com/2…