Yeah that’s fair if you need that. I don’t, which is why I’m saying all this.
Yep I’m sure. Each paid workers plan has 10 millions requests included per month and 30 million cpu milliseconds.
Even the free plan has 100k requests per day.
Details here. Workers & Pages Pricing
I think there’s a lot to be said for being as independent and not reliant on others as you can - and that circles back nicely to your original question “How does Elixir keep going?” - I’d say for pretty much the same reasons it got off the ground to begin with; it’s incredibly empowering for bootstrappers and small teams.
Look at what Erlang did for WhatsApp - Elixir made that power accessible to the rest of us… and that was over ten years ago. Now, it’s beyond even that, with a whole ecosystem behind it 
And do I really need to mention community - the Elixir community is still one of the best around : -)
This is true, but WhatsApp grew up in a completely different world than we live in now. Even AWS was barely a thing in 2009. It’s essentially a legacy codebase now.
That’s not enough for me. Although I don’t have that much real traffic there are bots and bots traffic is significant.
Cloudflare has the problem of starting all kinds of projects and never finishing them, leaving you to clean up the warts. Just the insanity of discrepancies between workers and pages and the not updated documentation drove me crazy when I used it last.
I’m not anti-Cloudflare, we do use it, and I use it personally, but it’s not as magic as this thread makes it sound.
I don’t think it’s that different. I doubt you could run WhatsApp on AWS in some other language without having deep pockets.
Elixir and Erlang lets you take on gigantic tech - it was true ten years ago and perhaps even truer now, given the cost of dedicated servers has sky-rocketed…
What do I run is beside the point. The fact is that my app runs in low end VPS and will cost more to run in CF. Consider the situation of some nasty bots come and overwhelmed my server. With a low end VPS, I see my server is struggling, grep the log, see what is happening, and deploy a hotfix to mitigate. With CF, I don’t even see that coming (because everything chugging along), and got hit by a massive bill at the end of the month. Do you see why I prefer the first situation much more?
I beg to differ for the most part. The app was started in an era where it was normal for people to run their own hardware on-prem, without globally distributed apps or customer bases. Before containers really became a thing.
The economics of running software has changed massively. Large capital investment (at least in computation) isn’t really a thing anymore for small businesses in most industries.
Now it’s almost exclusively manageable fixed costs on rented hardware, or exclusively variable costs with serverless.
Obviously if you’re running a large company though, 100% variable cost is a shitty idea because of poor operational leverage.
But for startups with 0-a few customers, it’s great. Also depends on the kind of app for sure. There are some where I’d definitely consider Elixir still.
For the land of SaaS MVPs with few customers, imo serverless (finally) wins.
Fair. Just curious what kind of hobby app you’re running that’d exceed 10 million requests per month, or even 100k per day tbh.
It’d almost need to have some sort of event or data ingest round the clock to reach those numbers. But that’d be one of the apps I probably wouldn’t use CF for.
You have absolutely no leverage if you build your application on serverless or any proprietary technology. This has been proven again and again, if the company doesn’t die due to poor decision you might be lucky enough to force your customers to eat the bill, but why would you make that sacrifice, its even less attractive today then it ever was as you can just get help from AI.