When I did a cost calculation 7 years ago, even the AWS spot market was 3x as expensive as renting CPU time in a datacenter.

So, size-competence matrix aside, companies for whom the main cost is computing time would be ill-advised to use a cloud provider (except maybe as fallback to absorb rare load spikes that correspond to income spikes).

And Ruby is still among the 30x slower languages in the benchmarksgame¹, so I expect that the break even is hit earlier with Ruby. And if you have a significant amount of essential logic in Ruby, moving to a datacenter might be the cheaper option compared to rewriting in Java or Rust.

¹: https://benchmarksgame-team.pages.debian.net/benchmarksgame/box-plot-summary-charts.html

Expand full comment

While I think Rails was a truly great creation and something that has shaped a lot of frameworks and influenced a lot of other tech, and as someone who was a big Rails fan and used it for arguably my most successful company, I couldn't agree more with your statement about DHH "insisting that his personal experiences are universal principles." Anytime someone writes these kinds of things where they make blanket statements that they think apply to every other company, it instantly significantly reduces the value, weight, and trust I'd give that article. I think the cult of DHH is long gone, but his writing obviously still gets attention. Thank you for writing this.

Expand full comment

Isn't he the same guy that wrote the screed hoping DEI would just go away so he can go back to hiring people without caring who they are again? He clearly has a thing for the good old days that weren't all that good. Also, I recently went from working in an entirely cloud-based consulting arm of a tech company to a mid-sized business that is hoping I can help them move more to the cloud, and the amount of time/energy/resources spent just patching, talking about patching, planning patching, etc. is mind-boggling to me.

Expand full comment