Ask HN: Hunting for a Framework

create-t3-app is a great starting point for folks going this path. Has optional checkboxes for db/ORM/migrations (prisma.js), auth/sessions (next-auth.js), tailwind CSS, typed client/server communication (trpc).

https://create.t3.gg/

https://github.com/t3-oss/create-t3-app

I hadn't touched much webdev for a few years, the ecosystem / build tooling has gotten a LOT better. It's back to being fun and painless to spin up a new webapp.

I recently created ~25 full-stack example projects using the most recent technologies and frameworks. I believe the following stack is the most productive, yet very scalable (doesn't use "leaky" abstractions).

Frontend:

  - Good, old, Single Page App. I believe that all these SSR/isomorphic rendering frameworks are a lot of times just unnecessary complexity.
  - Vite.js + your preferred framework (Vite supports React/Vue/Svelte, etc).

Backend:

  - Node.js + Typescript.
  - Trpc.io - while it's not "automatic generation of REST endpoints", it's very close. It's very easy to write your endpoints, and the added flexiblity/control is definitely worth the time spent writing them.
  - Trpc also has a built-in support for subscriptions: https://trpc.io/docs/subscriptions
  - Prisma + Postgres. Good, old, proven, flexible SQL database is always a safe bet. And Prisma makes interacting with it an order of magnitude easier.
  - The whole "network layer" is completely and very efficiently handled by Trpc. It also uses React-query under the hood: https://github.com/TanStack/query.

Infrastructure + deployments:

  - I would recommend going for AWS.
  - Frontend (statically built) can be hosted in S3 bucket behind a CDN. You should also properly configure caching headers returned by Cloudfront to make it work with SPA.
  - You can run your TRPC API in AWS Fargate container (behind ALB, or behind HTTP API GW). Alternatively, you can also run it in a lambda function.
  - To make the configuration of your Infrastructure/Packaging/Deployments 97% easier, have a look at

https://stacktape.com

(disclaimer: I'm a founder)

The Supabase CLI does have migrations (merge-diff approach that is wonderful) and codegen for DB types for the front-end to use. We've even dispensed with Prisma recently as the ORM and just use the Supabase Client for data access on the front-end. Whole API layer disappears.

Add in a SvelteKit and deploy to a Vercel and you have a minimal yet powerful stack with everything you need for 90% of business CRUD app demands.

I’ve been using razor pages. For me it takes the abstractions I enjoyed with aspnet MVC, with less of the boilerplate- controllers, etc

I love reading the comments. There are no simple answers.

I was and still am in a similar situation. Since performance, ease of use and resource consumption are important my backend language of choice is Go.

So I wrote a simple backend generator with RESTful CRUD endpoints. The workflow is, you define the data models in your models package. The generator then analyzes those models, creates all the routes and repositories and handlers, then creates Angular model definitions (interfaces) and finally vuex-orm model definitions.

It's very simple and primitive. If you want to change anything you have to regenerate. So this gives you a skeleton backend app. No auth. But that's the general idea. It uses mongodb as a database target. And most importantly it supports data updates via websocket. I initially worked with Angular on the frontend but have been using Vue for the last 2 years. Angular sadly only has ngrx-data which is incomplete. Relationships are not supported and have to be done manually. So I found vuex-orm, which is far from perfect, but works well and is productive, supports relationships. However vuex has been deprecated in favor of Pinia. Someone did a Pinia version of vuex-orm-next, but I haven't tried it yet.

So that's all I have.

I've been thinking to rewrite it. I wanted a simple tool, that doesn't get in my way. I've been thinking to have a clumsy model definition way like entgo.io . And the more I worked with it, the more I see that their way is a better way. Have elaborate model definitions and relationship definitions. I was thinking about that for a while and ent made it. But ent is far from perfect too. It doesn't generate client models, it's complicated, the documentation is lacking. The developers assume you're smart. I'm not smart when I have to crawl through someone else's code. They're going a certain route, and both their graphql and grpc implementations are lacking. I tried to write an extension but the learning curve was too steep. And now they have migrations with some external library, am I'm just sitting there thinking, why do you assume your libraries are a part of my world. The whole thing is a mix of marketing and hype development. They have their own company now and of course they want to make money. I prefer something less hype driven. Something where when you criticize you don't self censor in order to not hurt their reputation. And I hate that.

So I'm thinking make such library myself but how can I monetize? Who would pay me to create a useful tool to solve this 20+ year old problem? This a serious question.

I could open source my tool so far. But I didn't because it could be used by competitors to provide what I provide. Currently my income is so low that I'm actually seeking either employment or freelance work, which I haven't done in at least 8 years.

An alternative would be Spring Boot 3. Learn Java. But could I handle the Java bs on a regular basis? Sometimes when I hate Go, I experiment doing that same backend with Spring Boot. I do the JPA definitions, do the security setup and then I just say fuck it because deep diving requires more Java knowledge and the community is very unfriendly and walled-off.

I'll continue reading the other comments now.