I will bet a pint of Guiness if we state a requirement for SGML/DocBook, less than 5% of new contributors will complain over the first year, and in two years, no one will think twice about it.
This is serious. A pint of Guiness makes this more than a theoretical discussion. The problem with your proposed bet is that the people who get turned off won't complain, they'll just go away. You won't find out anything unless you have two avenues for contributors and don't bias either of them by making one or the other purposely harder or less attractive. Then you'll be able to tell by the way people choose. They'll vote with their feet.