How do I start reinventing the wheel by writing a simple web server, but not too low-level and still stay safe?

Welcome. I like to re-inventing wheels too; however, I prefer to understand how the existing wheel works first. For example, I still use Phoenix, but I have replaced many functionalities of Phoenix with my code to scratch my itches.

There’s a great video series by Andrea Leopardi if you want to get into lower level networking (tcp/udp):

He also has a book on the topic.

I find it less enjoyable to figure out how to do something in a prebuilt system than to figure out things from scratch all by myself. And this decreased enjoyment makes me quit or not persist. On the contrary, if there is time constraint and it is too challenging, it becomes off putting to do everything from scratch. I am trying to find that sweet spot for me.

I will consider again though. Thanks for sharing your thoughts!

Thanks, I have skimmed through one of his videos. They might be useful, will check them out again in a more spacious time.

Merhaba, hoş gel!

Personally despite many years working in web space, I have never kindled much of an interest in networking deeper than that of a user. However, at framework level, Phoenix is already built on Cowboy/Bandit so you could start a project with one of those (probably Bandit) and design your own request parsing/routing/response handlers. Not something I’ve done myself full disclosure but that’s where I would start.

To be honest, I am not particularly interested in writing a framework from scratch. The whole thing is… I want to understand what is happening under the hood. Not just on the surface though. I appreciate understanding why a wheel has no corners, and why it has evolved into its current form by trying to push a car with rectangular tires

By recreating its initial form, I get to see how things emerged. And I like it. I do it to some extend, just enough to satisfy my desire to internalize.

Haven’t heard of Bandit before, will check it out! I guess I will start with either Cowboy or Bandit as you suggested, and see if it is the sweet spot. If I find it too challenging to my taste, will consider Phoenix with many rewrites.

Thanks for sharing.

P.S: Hoş bulduk. ^^ Appreciate the warm welcome.

Slightly different but if you don’t really care about the “end result” but just want to experiment with some building blocks:

It might be fun to just build a simple TCP server and play around with that and “actors” and (binary) pattern matching. That’s basically a web server, isn’t it? Can go as simple or as hard with your network protocol as you like.

Should give you a sense of how programming and building something like that works out in Elixir.

Something similar is actually one of the introduction chapters of Elixir and specifically Mix / OTP: Task and gen_tcp — Elixir v1.20.2

That’s something no one else can help. Since everyone has different backgrounds and interests, I’d suggest you to come up with 2~3 ideas, and people on this forum can tell you how much time/effort each would require, so you judge for yourself?

That would be interesting and suit my case probably the best, though my only concern is how safe it is. Even though I don’t care about selling the software I make with it, it should be safe (for me) to deploy to the internet. But that would also be the charm to it, right? (Making it safe.) Thanks for the link, I will check it out before all.

Yes, the reason I wanted to open this thread was to get as many ideas/input as possible, and see who has done what, or what kind of paths may I take.

Instead of people judging a case for me, I am more interested in what people has got to share. For example, thanks to you, I know I could start with Phoenix and just replace anything I would like to tinker with.

Should you told me “Just use Phoenix”, that would not make much of a difference to me, but you showed a different lense, and it widened my perspective on the topic. That is exactly what I was hoping to find, and you gave it to me, thanks!