I’m curious about what you think would be a meaningful definition of a “perfect encoding”.
Part of the point of the excerpt you quoted from Aaronson is that in any notion of an encoding, some of the computational work is being done by the decoding procedure, whatever that is. So e.g. you can specify a programming language, and build a compiler that will compile and execute programs in that programming language, and then talk about a program perfectly encoding something if it outputs that thing when run. Some of the computational work is being done by the program but some of it’s being done by the compiler.
Part of the point of the excerpt you quoted from Aaronson is that in any notion of an encoding, some of the computational work is being done by the decoding procedure, whatever that is. So e.g. you can specify a programming language, and build a compiler that will compile and execute programs in that programming language, and then talk about a program perfectly encoding something if it outputs that thing when run. Some of the computational work is being done by the program but some of it’s being done by the compiler.