At the expense of losing what little reputation I have on HN I will say this:
As many others on here seem to be correctly saying, i think this article amounts to fear mongering of vendor lock in. The modern public cloud is very different from the Oracle/IBM mainframes of yester-year.
The whole point of the public cloud is to leverage managed services to their fullest extent so you can move incredibly fast. As a startup, you’ll run laps around your competitors doing all of this from scratch simply to preserve their non vendor lock in.
The notion that removing that glue code that glues your code to AWS or Azure managed services amounts to vendor lock in, that is no more true than any other code running on any VM that talks to those same managed services. Except the main difference here is that your not wasting time writing the glue code.
Additionally Azure Functions or AWS Lambda, or even Functions on Kubernetes, which are meant to be the smallest unit of work when used correctly (similar to a MicroService) and should contain only your application logic are “vendor lock-in” is absolutely rediculous to me. If anything when you do decide to move vendors this will be the easiest code in the world to migrate, inputs and outputs.
I will concede that it is hard to see this the way I’m describing if you haven’t actually worked on the modern public cloud and are not actively taking advantage of managed services on there for speed of delivery.
A little self promotion: as an example of what’s actually possible with these Serverless frameworks I recently built a cross platform app, as a side project in just a few months nights and weekends with the entire backend as Serverless Functions, the app can read any article to you, using some open source ML models for text to speech, and can be found https://articulu.com if you want to check it out.
As many others on here seem to be correctly saying, i think this article amounts to fear mongering of vendor lock in. The modern public cloud is very different from the Oracle/IBM mainframes of yester-year.
The whole point of the public cloud is to leverage managed services to their fullest extent so you can move incredibly fast. As a startup, you’ll run laps around your competitors doing all of this from scratch simply to preserve their non vendor lock in.
The notion that removing that glue code that glues your code to AWS or Azure managed services amounts to vendor lock in, that is no more true than any other code running on any VM that talks to those same managed services. Except the main difference here is that your not wasting time writing the glue code.
Additionally Azure Functions or AWS Lambda, or even Functions on Kubernetes, which are meant to be the smallest unit of work when used correctly (similar to a MicroService) and should contain only your application logic are “vendor lock-in” is absolutely rediculous to me. If anything when you do decide to move vendors this will be the easiest code in the world to migrate, inputs and outputs.
I will concede that it is hard to see this the way I’m describing if you haven’t actually worked on the modern public cloud and are not actively taking advantage of managed services on there for speed of delivery.
A little self promotion: as an example of what’s actually possible with these Serverless frameworks I recently built a cross platform app, as a side project in just a few months nights and weekends with the entire backend as Serverless Functions, the app can read any article to you, using some open source ML models for text to speech, and can be found https://articulu.com if you want to check it out.