I would have loved to have been exposed to some CS in school. For some reason, my otherwise rather privileged schooling didn't feature any computing, and nobody in my direct environment did any. Only after starting a Physics undergrad did I meet some proper hackers, and years onwards I still feel like I'm years behind on them.
I'm pretty comfortable with that conclusion.
Anti-teaching, the teaching of things that are actively misleading about the nature of the subject or even actively wrong, is a bad idea. It doesn't matter how much "prior art" it has. It doesn't matter how much we have to just fill time while the students are sitting there. It doesn't matter whether there isn't a "better choice", because nothing really is a better choice than anti-teaching. Anti-teaching is a bad idea.
Besides, ask around. Ask your non-geek friends simple, basic physics questions. If you like, even ask if they took physics in high school. If the goal of physics is for students to pass a test in school, we're OK. If the goal of physics is to give the general public an understanding of simple, basic Newtonian physics... we already aren't teaching physics, we're just going through the motions. The "threat" you raise is already the situation we are in.
There were attendance problems in Oakland due mainly to external factors, like parents not wanting their kids walking home after dark. Only 7 of the original 13 Oakland students finished the class, compared to 12 out of 14 in Daly City, but everyone who finished the class in both Oakland and Daly City had a working program by the end of the 10-week course. I think it's fair to say that all of the students who finished have a reasonably good understanding of some basic computer science concepts: functions, variables, applying operators to values, evaluating expressions, and a simple notion of types (strings, numbers, booleans and "images"). The obvious next steps would be iteration/recursion and some simple compound data structures, but we didn't have time for that in just 10 lessons, which is all that this particular after-school program permitted.
My point is this: middle school students can handle a fairly rigorous approach to introductory computer science. Saying we shouldn't teach something because it will be taught poorly is no excuse not to teach it the right way -- we just need people qualified to do that. That's the real problem, but most of you who're reading this could help do something about that by volunteering to teach a curriculum like Bootstrap.
It was way too formal to boot, too! If I want to create a specification, I write them in code, either as unit test or as a prototype.
What would such a "computer science" look like? I don't know, but the odds of it being actively anti-educational are far too high for my liking. Can you imagine what a computer science optimized for testability would look like? I can see it now:
"Fill in the blank with the correct variable name."
"Organize these lines of source to correctly sort the list."
"Find the thirty-three keywords of $CRAPPY_LANGUAGE_WE_STANDARDIZED_ON_TWENTY_YEARS_AGO in this word scramble."
"Which of the following is the definition of 'strong typing'? A: wrong answer B: wrong answer but defendable C: essentially wrong answer from 30 years ago the curriculum officially calls correct D: a more right answer than C but still wrong"