How is this different to aptly, especially if one already has object storage?

One of the tools we tried was Aptly! A couple issues we had with Aptly:

1. Aptly needs to rebuild the entire repository before it can do any changes. One of our customers builds packages for every Linux OS, and every architecture, and each release contains 10-something separately packaged tools, and they do regular releases. This means their repository of packages is _huge_. When they try to do small maintenance actions like removing a package that has a bug, they need to rebuild their entire repository to do this, which takes upwards of 30 minutes. On Attune, since we have a centralized control plane, it takes about 10 seconds.

2. Aptly doesn't have concurrency controls. This makes it annoying to run in CI. Imagine you're setting up a CI pipeline that runs on every commit and publishes a package. You need to make sure you don't have two different instances of Aptly running at the same time, because their updates will clobber each other if they publish at the same time. Can you ensure that your CI only ever runs one publish at a time? This is easy in some setups, but surprisingly hard in GitHub Actions - concurrency keys don't quite work the way you'd expect them to.

3. Aptly uses its storage as its source of truth for the current state of packages. This is a great design choice for simplicity! However, if your CI job dies halfway through during an Aptly publish, this can leave your repository in a corrupted (in this case, half-updated) state. To recover from this, you need to rebuild the repository state... which you can't, because the ground truth is contained in the storage which was just corrupted! We mitigate this issue by tracking metadata in a separate transactional database so CI crashes don't cause corruption.

These are some examples of little cuts that build up over time with all of the tools we've found so far. Lots of the open source tooling is great at low scale, but the cuts start to build up once you are pushing a meaningful amount of packages to production.