There is a principle in program design: defer decisions to the last possible moment, because this is when you will have the most information about how to make the right choice.
If I were to design a new programming language today, I would apply this to a programming language environment. It would look something like this:
- A runtime that is small – optional, ideally – but can grow and be configured, incorporating ideas from different communities.
- A language that is equally small, but growable. This means Lisp, but I include it from the standpoint of being a reasonable place to start. Include facilities to extend the actual textual syntax of the language, not just manipulating sexps.
One of the more frustrating aspects of programming languages is that they make tradeoffs up front. The design decisions are made at the earliest possible moment. These decisions often are as permanent as the languages themselves are, which is a shame. Programming is a constantly growing, dynamic field. We will never know less about it than we know now.