Solana Fellowship Week 2: Creating mini dApps

This week's goal was to start creating mini dApps on Solana. There are 2 parts to this - interacting with Solana using @solana/web3.js@solana/web3.js library and creating on-chain programs.

Thoughts on web2 vs web3 development

On Monday, Preethi Kasireddy gave a talk about how web3 development is different from the classical client-server programming model. On one hand, web3 makes it extremely simple to have a global computer with persistent storage, but on the other hand, doing iterative development is significantly more difficult. Another major difference is the security model and how much auditing is needed before deployment. Because iterative upgrades are difficult, most of the auditing and security reviews need to done prior to deployment.

My experience so far with Solana and Ethereum aligns with that: a lot of the basic concepts are the same - you are still writing rust/solidity/js, storing/retrieving data, using a database as a source of truth - but things change a lot when you start building, deploying, combining these components. Be sure to have an open mind if you want to go deep.

Solana web3.js library

This week we did 2 quests - creating an airdrop program and creating a roulette game. Both of them were about how to use @solana/web3.js@solana/web3.js to interact with Solana and on-chain programs. The quests are relatively easy if you know javascript.

My quick review of the library - very easy to use, powerful (you can move real money in a couple of lines of code), and even better with typescript with all the autocompletion goodness.

On chain programs

We had to build a voting program with support for delegation (same thing implemented in Solidity). I wanted to do this exercise without using the Anchor framework. Even though it makes program development much simpler and secure, it abstracts away quite a few things. I wanted to learn how those things handled by the solana program sdk, without the Anchor abstractions.

It turned out to be more challenging than I expected. I couldn't find a good starting point, so I went into a rabbit hole and explored the solana program library. I found the name-servicename-service program most helpful. It's features were relatively straightforward and the implementation was simple to understand. I created a template for myself at templates/program.

I completed the program after a lot of browsing, going through open source Solana and Rust programs. I had no way to test it through a UI, so I ended up learning how to write integration tests for Solana. It was satisfying to see the green check marks after some challenging work.

That being said, I'll be using the Anchor framework from now. If anything, not using it showed me how useful it is.

The fellowship seems to be moving into top gear, with each week being more challenging than the last one. I enjoyed this week's exercise a lot. The only thing lacking was a UI to make it an end-to-end voting system. I'll continue writing more Solana programs next week, building on this week's work.