damien

It does close the gap.

Before:

  cc -shared -fPIC -Os -s -o empty.so empty.c
  cc -std=c99 -Wall -Wextra -O3 -g3 -fpie -pie -o benchmark benchmark.c ./empty.so -ldl

  jit: 1.541608 ns/call
  plt: 2.309939 ns/call
  ind: 1.540583 ns/call
  

After:

  cc -shared -fPIC -Os -s -o empty.so empty.c
  cc -std=c99 -Wall -Wextra -O3 -g3 -fpie -fno-plt -pie -o benchmark benchmark.c ./empty.so -ldl

  jit: 1.541836 ns/call
  plt: 1.539239 ns/call
  ind: 1.542874 ns/call

It's likely talking about all future dividends (discounted for time value). Startups will naturally reinvest their earnings to grow, but at some point, pursuing growth will be too costly and the best use of their earnings will be to return it to the owners as dividends.

If a company were guaranteed to never return anything to owners, then it would be more like speculating in the future value of a baseball card than an actual investment.

It may not be a practical way to calculate value, but it's not crazy either.

Probably the largest application written in Rust is the Rust compiler itself.

Yes, I'd say a language runtime definately falls in that critical 3% Knuth referred to in the original quote.