In writing this up - https://simonwillison.net/2024/Jul/30/aws-codecommit-quietly... - I found out about a much more significant deprecation: Amazon QLDB (Quantum Ledger Database - a blockchain-ish thing they launched in GA in 2019) is being hard-deprecated too! They're shutting it down completely on 31st July 2025, having announced the shutdown a few weeks ago (on July 18th).
It wasn't a blockchain per se. It basically was Amazon's Version of Rich Hickey's Datomic and them pointing people to Aurora for Migration is a bit of a joke. It's a completely different product.
QLDB is an immutable database with allowing you to move back and forward in history and do live streaming of events etc with cryptographic verification built in.
You _can_ set up Postgres like this but it's a lot of hoops and not very user friendly (some monstrosity involving DMS, Kinesis Firehose and what not).
> QLDB is an immutable database with allowing you to move back and forward in history and do live streaming of events etc with cryptographic verification built in.
I used it for a subset of the features. Financial auditing and risk rating of destination banks for a particular network.
Having an immutable database that was able to track fraud ratings, along with when exactly it occurred with the data available at the time was pretty cool. You could probably do it with other tools, but having it all in one product was nice.
The cryptographic verification was helpful for legal defence. I can't think of any other uses so we may have been their only customers.
QLDB is the 'blockchain' businesses need. If you think of blockchain design choices as a set of dip switches, QLDB is the only one with all the correct settings for distributed business transaction value at scale.
Eh, most people would like their apps to provide them with a browsable history of their data, and why the changes happened.
It's just increases scope x100 to get anything done, so few do it... but if the tooling was available, almost all stateful applications would benefit from it.
Most people? There is a Django-history module, and yet it’s not nearly as popular as Django. I’ve used this on two out of 100 projects. No one ever actually wanted the history, it’s there for “just in case of an audit” and compliance, which never comes.
lol no. And the DB isn’t going to tell you who what why without you writing code to support that audit in which case… that special database has solved exactly 0 problems that a traditional RDBMS already solves.
I was addressing the point "nobody wants that". And I disagreed, because: if it existed, almost everyone would want it.
The tooling just doesn't exist. Doing it with half-assed tooling increases the scope of literally every development you'd be doing, so nobody bothers with it. Ive never worked with this particular database, so no idea how it would work in practice here.
It's basically a schema-less database which you could query with SQL-ish syntax (PartiQL) and which provided additional views to let you query the history of changes. Git is a good analogy. I've never used the signing for anything but the ability to query the history makes it trivial to implement things like point in time recovery (essentially equivalent to a git checkout) or audit logs (essentially git blame or git log -p), which would otherwise require a much more complex schema in something like DynamoDB: https://docs.aws.amazon.com/amazondynamodb/latest/developerg...
Straight callback to all the "you don't need a blockchain, you just need a database".