I’m not buying that. Lambda is merely an execution environment. In most Lambda functions I write, the Lambda-specific bit is tiny, and could be easily replaced without affecting business logic.
On the other hand, most Lambdas I write interact with other AWS APIs, which is where the real lock-in is. The effort to eg. move the data off Dynamo is substantially higher than what’s required to switch that bit of code to run on k8s and consume a Kafka topic.
Cool. What part aren't you buying, the title of this post, or the actual five paragraphs in the article that actually give you the context of that title?
On the other hand, most Lambdas I write interact with other AWS APIs, which is where the real lock-in is. The effort to eg. move the data off Dynamo is substantially higher than what’s required to switch that bit of code to run on k8s and consume a Kafka topic.