Thread by @TejasKumar_

@mlfrg @zeithq λs all the way my brother

@mlfrg @zeithq The problem that I have with running a container in the cloud is that it's harder to manage: your container dies (maybe insufficient resources?), your whole system dies. Of course, you could have microservice containers, but they're still harder to scale than λ IMO.

@mlfrg @zeithq Since λ are just functions, they:

- boot up
- run to completion
- shut down

I consider this _really_ predictable with very low screwup potential in contrast to:

- deploy
- start
- ...anything can happen

The above holds state (over time) that could be corrupted, etc.

@mlfrg @zeithq IMO that's the greatest appeal of λ: you are not in charge of supervising – they wake up, do their jobs, go to sleep. It makes me so happy.

I think it's nearly impossible for me to think so far in advance and create something that runs perfectly stable as a long running process.

@mlfrg @zeithq Philosophically, I've often found greater joy when I've STOPPED, CLEANED UP, and RESTARTED various things: a gym routine, reorganized my room, etc.

λs do their job, and RESET, serving people anew, clean, each time. :justright:

@mlfrg @zeithq Finally, the single responsibility principle is a WONDERFUL principle as we engineer software.

I feel like λ really captures the spirit of this principle: I can/should not be responsible over so much ceremony with my containers and my clusters.

@mlfrg @zeithq Instead, I write a λ to do a _single responsibility_ and do it well, test it, `now` it, and it continues to serve me faithfully for _years_. I don't have to concern myself with process management, self healing, scaling, etc. It quite literally JustWorks™.

@mlfrg @zeithq Thank you for your tweet, @mlfrg. You've really encouraged me to appreciate λ more than I already do. It's such a wonderful concept.