July 15, 2026

Git, but make it database drama

Show HN: SirixDB 1.0 Beta – Git-Like Versioning, Diffs, Time-Travel Queries

A database that remembers everything has coders arguing if it’s genius or just more homework

TLDR: SirixDB says it can store every version of your data and let you check the past as easily as the present, which could make audits and history tracking much simpler. The comments quickly turned into a fight between impressed builders and skeptics asking the big question: is this a breakthrough, or just another clever tool with tradeoffs?

A new project called SirixDB just rolled into Show HN promising something every stressed-out developer dreams about: a database that never forgets. Change a piece of data today, and you can still ask what it looked like last week, last month, or at 3 a.m. on some cursed Tuesday. The pitch is basically “what if your database worked a bit like Git,” the file-history tool programmers adore, except you can actually ask it questions like a normal app. And yes, the creator came in swinging with big claims about fast old-version lookups, tiny storage growth, and a single neat package you can run without turning your setup into a science project.

But the real entertainment is in the comment section, where the vibe instantly split into “this is wild” and “okay, but what’s the catch?” One side leaned hard into the project’s nerdy flexes, with the creator casually mentioning there’s basically no practical ceiling “besides running out” of giant ID numbers. That kind of swagger absolutely reads like catnip for database fans. The other side showed up with the classic Hacker News energy: polite, surgical skepticism. One commenter, building a rival-ish audit-log tool, basically asked why this matters if it skips familiar SQL standards and doesn’t visibly include tamper detection. Translation for civilians: cool demo, but can I trust it and does it play nice with the rest of my stack?

So the mood is deliciously split: part hype, part interrogation, part “please explain this to me like I’m five.” No giant meme outbreak yet, but the biggest joke is the usual one in database land: every “simple history feature” somehow becomes an existential debate about standards, trust, and whether developers are about to adopt one more very clever thing that also ruins their weekend

Key Points

  • SirixDB 1.0.0-beta is presented as a usable, actively developed bitemporal database with stabilizing on-disk format and public APIs.
  • The system claims to make every revision directly queryable by revision ID or timestamp without replaying logs or events.
  • Its storage model is described as structural sharing with sub-page versioning, copy-on-write, and a sliding-snapshot algorithm that limits page reconstruction work.
  • SirixDB tracks both transaction time and valid time so users can distinguish between when data was recorded and when it was true in the real world.
  • The article reports benchmark improvements for concurrent reads under a committing writer, including 11,198 reads/s on a 12,800-revision database after performance fixes.

Hottest takes

"no upper limit besides running out of 48bit nodeKeys" — lichtenberger
"what is the advantage of this project?" — aforwardslash
"It doesnt seem sql-compliant" — aforwardslash
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.