Journal
Notes from real projects - AWS, DynamoDB, Rust, home servers, and whatever open source tool I'm currently building.
Parity Suite 3.0.0: measuring the wrong thing
The board scored emulators on a pass rate over what they attempted, so a score could jump 8 points without the engine changing a line. 3.0.0 replaces that with two figures, letter grades, and 56 new tests.
Read article →
When a DynamoDB vector index is actually ready
AWS's docs contradict each other about what searching a backfilling vector index does, describe a state the index never enters, and name a readiness signal that arrives too early. Here's what I measured instead.
Read article →
What a write to an indexed DynamoDB table costs
Adding a GSI to a DynamoDB table doubles the cost of some writes and changes nothing about others, and which is which isn't well documented. So I measured it against real DynamoDB and pinned every number.
Read article →
Dynoxide 0.12.0: PartiQL RETURNING, wasm conformance
RETURNING finally does something, the wasm build lands on the conformance board at 100% over three quarters of the suite, and PartiQL UPDATE stops creating rows it shouldn't.
Read article →
You can't pin DynamoDB to one region
My DynamoDB conformance suite took one AWS region as ground truth. Then that region started accepting a value the rest of AWS rejected, and weeks later reverted it. AWS ships changes region by region, so one region was never a safe baseline.
Read article →
Dynoxide 0.11.3: closing a Windows port hijack
Issue #23 asked for a Windows socket option investigation. It found the opposite problem: dynoxide's ports could be hijacked by a same-user process. 0.11.3 closes that, and the test suite now runs on Windows for the first time.
Read article →
Dynoxide 0.11.2: a stack of conformance fixes
Last week I pinned what a conditional DynamoDB transaction costs against real AWS. Dynoxide 0.11.2 makes the engine report the same numbers, fixes an UpdateTable bug that dropped your table keys, and closes a stack of projection and validation gaps.
Read article →
What a conditional DynamoDB transaction actually costs
Does a conditional write inside a DynamoDB transaction cost read capacity on top of the write, or is it write all the way down? The docs aren't clear, so I ran it against real DynamoDB and pinned what it costs, re-checked on every run.
Read article →
Dynoxide 0.11.1: four conformance fixes
A small dynoxide patch. Four fixes where the engine evaluated an expression or validated a number a bit differently from real DynamoDB, most of them checked back against real AWS.
Read article →
Learn DynamoDB by running it - accesspatterns.dev
DynamoDB is a brilliant database and a hard one to learn. So I built accesspatterns.dev - a real engine in a browser tab where you learn it by running it, from your first GetItem to single-table design. No AWS account, nothing to install.
Read article →