I agree. What would prevent something like 20q.net? Compile a database of make/model/year, symptoms, and eventual repairs... People enter their car type then answer questions like "Did the car start? Yes/No/Sometimes?" "Does it make a grinding noise when you turn the key? Yes/No"... then using past results it tells you it's 85% likely to be the alternator, etc...
I used to work for a car manufacturer where we made such a diagnostic database for mechanics. At the end of all the decision trees was "hmm. call factory." Instead of following the suggestions, the mechanics routinely skipped to the bottom. Several dozen of my field trips to dealerships were resolved with a $2 part that they skipped over because reading was hard.
