Does source code have to be stored in files?

Is it specified in any C++ standard? Would compiling OCR-ed code written on a sheet of paper be legal?

Who cares about legality? Valid code is the one the compiler accepts. If you find one that has OCR built in and eats pdf go ahead.

>>1
Thats a weird question. I assume you're trying to archive
code on paper, then at some later date OCR it and feed it to the compiler of the day. There are a few flaws in this design:
1.C++ compilers of the future will likely not compile it.
2.OCR will introduce subtle bugs(is it ; or :) where C++ will show
a 12page template instantiation error completely unrelated to the mistake.
3.Paper doesn't scale;Microfiche/Ultrafiche microfilm will be more compact and capable of storing a modern C++ codebase for centuries:
Microfilm and Microfiche Storage When properly stored, microfilm and microfiche can accurately store data with a shelf life of about 500 years
https://en.wikipedia.org/wiki/Microform

one and zeroes on stone tablets is the best longterm storage

>>4
binary to text implies a knowledge of specific encoding which will
likely not survive the time periods,
future readers of stone tablets will have to try all possible
encoding to decipher ancient sepples codebases.
The amount of single-byte ascii encodings its 255!,
whereas unicode will create additional requirement of testing
all variable encodings upto (256^4)!(4byte utf-8).
With periods more than few thousands of years, the hardware/software
stack may also change radically due new paradigms, making
such Sepples code unusable and undecipherable due its complexity tied
to C++ compiler/stdlib. C code on the other hand will be much
more intuitive to read(except for preprocessor of course) and
could be converted to the language of the day by mechanical translation: writing a C compiler is much easier than C++ compilers.
(of course something like Forth will be easier still)

>>5
You are lack of punctuation makes me not care about whatever you where trying to say.