That line is blurring to the point where it barely exists any more. Compiled languages are becoming increasingly dynamic (e.g. JIT compilation, code generation at runtime) and interpreted languages are getting compiled. JavaScript is a great example: V8 uses LLVM (a traditional compiler) to optimize and compile hot functions into machine code.
IMO the only definition of “real” programming language that makes any sense is a (Turing complete) language you can realistically build production systems with. Anything else is pointlessly pedantic or gatekeeping.
So the only ways that the distribution mechanism matter are really a difference between How does the distribution mechanism matter beyond that? And even those points are
I’m assuming you are not saying “real” languages should be hard to write…
Objective-C and Go run slower than C and they’re all compiled languages. Sure, an interpreter will be slower than a compiled language but modern languages aren’t simply interpreted (i.e. JIT, etc).
There are dynamically typed compiled languages, and high level compiled languages.
Calling one class of languages “real” and another class something else is inherently demeaning. I wouldn’t have cared enough to type this if you used “compiled vs scripting” instead of “real vs scripting”. Though I disagree with using “scripting” at all to describe a language since that’s an assertion of how you use the language, not of the language itself. “Interpreted” on the other hand is a descriptor of the language itself.
I personally can’t stand Java because the language designers decided to remove ‘dangerous’ features like pointers and unsigned integers because apparently programmers are children who are incapable of handling the risk. On the other hand I love Go. It’s high level enough to be enjoyable and easy to write, but if you want to get into the weeds you can.