How does Elixir keep going?

Not everyone wants to deploy a monstrosity of a javascript app and couple it to Cloudflare.

Not everything is a website on the public internet.

Avoiding vendor lock-in is a good one. It would be interesting if an elixir-based system can be shown to be provably less expensive to run at scale too. Not sure if that’s true or not with how things like Cloudflare price.

On a related note, it’d be nice to see an elixir-tailored host that makes clustering dead simple and helps leverage the BEAM. Maybe that’s potions.io? I assume fly.io has abandoned the cause.

If you’re brand new in the field the yes infra and deployment is a gotcha, but that is not language specific, personally I think most of the BEAM and Elixir suffer from not showcasing the real potential of the language.

Some people would call it insanity

@budgie, are you forgetting half the internet went down last year because they were tied to CloudFlare?

Elixir is a firm favourite with those who want autonomy and reliability

They don’t have to be, it was just a figure of speech due to the AI slop involved.

It just makes the most sense these days. I get vendor lock-in is a problem at scale, but the truth is paying $100/mo with no users is a lot more impactful than paying $100k/mo at scale.

For most early-stage apps, these days I wouldn’t use anything else.

I didn’t even mention their durable workflow engine or queues. They have literally everything you need in one place.

They don’t have to be public.

Vendor lock-in is the only good argument for caution imo. But it’s also typically something that comes up later. For 99% of your apps, the problem isn’t will I be able to afford $1k/mo with 100 paying users. It’s can I justify spending $100/mo with zero paying users.

Potions looks cool, but even so, on 1 vps you’re not fault-tolerant. Not to mention you still likely need a HA postgres setup, and that’s not cheap.

On CF, availability isn’t something you actually need to think about in most cases.

Autonomy yes I get it, but reliability is relative. The internet is unpredictable, and everyone everywhere has downtime.

This is the wrong perspective imo. You don’t have to be a noob to benefit from simplicity. It’s not that it’s a gotcha, it’s just drastically simpler to go with other options, especially when the marketability of what you’re selling is unclear.

If you’re fine paying for a personal project regardless of whether it makes money, that’s fine. But there’s no contesting the value of what you get for $5 from CloudFlare.

I used to be hard in the other direction, but after a while it becomes hard to ignore the benefits.

I payed $5 per month for a vps to which I deployed several personal projects, db included, and I can switch among several vendors painlessly. What do you get for $5 from Cloudflare for your personal projects?

A shitload, tbh. Lot of stuff I probably don’t even know about.

Global edge deployments. DNS utilities. Transactional email sending, but also for receiving and processing emails. Secrets management. Managed HA ACID-compliant database. KV store. Queues and durable multistep workflow execution. Media storage and delivery. Basically free traffic.

I also have zero administrative responsibility for the infra. I could add 15 more apps, serverless endpoints, etc and pay the same thing.

For some apps you could even get away with being on the free tier tbh.

There are many people running commercial apps on the $5/mo plan.

My point is you cannot host on-device or on-prem applications on Cloudflare.

Cool, I don’t need any of those for my personal projects though. My understanding of serverless is they are billed by usage; are you sure the $5 per month come with reasonable usage included? if, so, how much usage are we talking about?