Zane ∞

This is one of my favorite spaces that I've participated in. We had multiple people in the round-table, discussing ICP, ZK, product-market-fit, especially focused on some of the more pervasive challenges facing ICP, and future outlook. Very informative space, recommend a listen

What you're saying sounds lofty, and I was drawn into ICP because of this vision, but we have some real practical issues that need to be resolved: 1. Instruction limits need to be entirely removed, not just lifted. There's a fundamental boundary on the epoch boundary. Not all software can be rewritten to get around these limits, and requiring libraries to be rewritten is a major source of complications, frictions, time, and effort 2. We need to be able to bring already-existing software to ICP. Right now most software has to be written from scratch for ICP. Anything having to do with HTTP has to be rewritten, and not all of it can be rewritten because of ICP/replicated compute. No TCP is currently possible, anything below HTTP is basically out of the question. 3. Databases are still necessary, you often say they aren't necessary but this is not the case. People really want to use the databases that already exist, and storing everything in global variables is not a solution. I've been working on this problem on ICP since 2021, and we've made progress. But it isn't helpful when you claim that databases aren't necessary. It's more like they're not possible right now (they are, just very difficult) "The ICP environments enables the ultimate simplification of code" My experience shows me that this is not the case generally on ICP. There are so many limitations compared with traditional cloud platforms like AWS, Azure, etc. ICP can't even compete with these right now because they aren't even equivalent in features. I don't see why DFINITY doesn't pour resources into resolving these fundamental issues, summed up like this: you should be able to pip install, npm install, cargo install anything that you would run on AWS. If that's not the case, ICP will continue to rule out vast swaths of use cases. I've been trying to make this vision a reality, and Azle's experimental mode shows a path forward: TypeScript or JavaScript, many more npm packages available, you can even build Express servers and use SQLite right from within your canister, with the limited feature set it provides it feels just like writing a Node app to deploy to a traditional cloud. ICP should fully embrace Wasi and implement HTTP and other networking protocols at the OS level, remove instruction and memory limits (memory coming along well), and become a truly capable cloud platform. If it doesn't do that, then where will it stand? It can't compete with cloud because of these limitations, and it struggles to be as robustly decentralized as the leading blockchains, with small replication. So it occupies a small middle space right now, which is seeing very little adoption. Something needs to change, from my point of view.

Clarifying my thought process on ICP vs $ICP and the Internet Computer: If anything I'm more of an ICP (protocol) maximalist. I would love to see ICP become a truly open source set of protocols. Being an $ICP or Internet Computer maximalist is different. ICP is not fully liberally licensed at this point. It would be amazing to see full Apache or MIT licensing. Currently some code is licensed under the Internet Computer Community Source License, which essentially prevents permissionless forking. I want the power, and I want everyone else to have the power, to fork ICP at will without permission from @dfinity. I want to see many different instances of ICP deployed in the wild, instead of just the Internet Computer. The Internet Computer and $ICP are one configuration of the ICP protocols, with a very opinionated set of design decisions. Unfortunately it's very difficult to get changes through to this one instance of ICP. This might be limiting experimentation and progress. Imagine a thriving ecosystem like Bitcoin and Ethereum with many forks and competitors. Has Bitcoin or Ethereum's lack of patents and full open source nature harmed or blessed them more? Let's make ICP full open source! @UtopiaCloudLabs (or its basic idea UTOPIA) would then just be yet another instance of ICP, or just a way to deploy ICP into a private configuration. Anyone, not just DFINITY, could push this narrative and solution forward. We could have many of these, and many other public networks, anything! Without this ICP feels somewhat brittle and centralized and risky, as the one instance must succeed and skills and code developed for it might not be able to be taken elsewhere. Embrace open source! Reject patents! Am I alone?