Over on Twi… what-shall-not-be-named, Kelsey Hightower argued that companies want on-premise back and that this is happening by on-premise product vendors copying cloud APIs in their products:
It might just turn out that the cloud was the best way to research and design better ways of managing our systems, and thanks to the open source community standardizing the APIs on top, we might finally have the blueprints we need to close the gap on-prem.https://twitter.com/kelseyhightower/status/1714685230991446212
His first example then is MinIO, a commercial provider of an S3 API compatible storage solution, with an open source GTM strategy.
Let’s think it through.
- Are APIs standardized now? I’m not sure. Why would MinIO remain compatible with S3? Even if we tried to move defacto standards (S3) into a true standard managed by a standard’s body, I think neither AWS nor MinIO will comply with it. Standards for APIs are only ever truthfully adhered to if they remain a pure cost factor and do not serve product differentiation. But AWS’ S3 and MinIO’s offering are hot products so they’ll most certainly vary and eventually diverge in their APIs to corner their market and lock users in.
- No bother, the open source community will do it for you (establish an API as a defacto standard). No such luck in the example case, because MinIO is a traditional proprietary vendor who owns its software. Like many, they provide an open-source software version of their product, but there is no true open source community nor widely shared copyright. Unless vendors come up with a better solution, a future relicensing to a source-available non-compete license may be a foregone conclusion.
- There will be an open source community version. You might also think that once MinIO relicenses, the rest of the industry will step up and fork the last good open source version and keep developing it as a proper open source community. In the example case, MinIO’s license will make it hard though (pure AGPL) to motivate industry. Certainly one important player, AWS, who in other cases has an incentive to lead an open source effort (e.g. OpenSearch), does not have a motivation in this case (they still have S3).
Kelsey then has other examples, but it is a mixed bag of community and commercial open source projects. I therefore think that only part of his argument holds: Yes, we can learn from the cloud to have better on-premise software for in-house clouds than past more traditional on-premise vendors could deliver. But I don’t think there will always be a useful community open source version of such software.
The real issue, thrown in almost as a side remark, and which I think is the main reason why public clouds will keep winning, is not artifacts, but people. It is simply impossible for a random company to reach the same levels of operational excellence that public cloud vendors and their outsize profit margins can achieve. While we can open source educational materials and operational blueprints, we can’t improve intelligence and capabilities of people through open source.