Here’s my one prediction for 2021 and beyond (and will take more than one year to be obvious):

AWS’s “selling managed OSS” strategy will go from “they’re ruining/screwing people” to “not very effective strategy.”

We already have enough evidence to see this.

1/

The high-level thing to understand is that OSS projects are not built to be run as managed services. So there is a permanent gap that makes OSS less appropriate and targeted than directly-built solutions, and that wedge is more and more obvious.

2/

One characteristic of OSS projects is that they’re not built to be serverless. So there tends to be a requirement that AWS only make them available like marketplace software: you have to launch them on VMs and deal with VM failure.

3/

Another is that OSS projects love to make you configure an endless number of things with obscure naming and weird parameters (and understanding this is often essential to getting it to work at all). AWS can pick decent defaults, but that only takes you so far.

4/

Another characteristic of many OSS projects is that they expect to run custom code a lot of the time / a lot of solutions for them require some writing and running of custom code around them.

5/

All of this means that OSS projects are great if you are ok spending a time learning their idiosyncrasies, operating them, and writing and running custom code for them. None of these make a great managed service today, though.

6/

We can compare Algolia to Elasticsearch, or BigQuery to Airflow/Presto, or even Serverless Aurora to MySQL or Postgres to see how custom-built managed services, especially built for the serverless mindset/JAMStack/“API Economy” are substantially better.

7/

In my experience, it takes 5-10 years for pundits and leading-edge enterprise to get these things, and I think we are probably quite near the beginning of that timeline (maybe 2 years in at most).

end/