I know it sounds like a stupid joke, bear with me for a second…
UPDATE: the link of the tutorial have changed. Thanks Roger for the info!
A few weeks back Tim Elliot got in touch for using rubyinstaller sandbox project (which uses MinGW) to build Shoes the awesome GUI library created by _why
Now, current master branch of RubyInstaller targets Ruby version 1.8.6.
Tim took the time, in his fork to workout the details and the dependencies to build Ruby 1.9.1!
Pointed here by _why to awesome Tim’s work:
Again, that’s what I call community!
This is one of the reasons I didn’t keep the sandbox environment for myself and seems that GitHub experiment has proven to be a success for RubyInstaller!
Of course, Ruby on Windows couldn’t make Tim’s life a joy, here is a post of something I believe is candidate for filling a Ruby bug report, don’t you agree?
Keep the good work Tim, hope next set of recipes for RubyInstaller will make your life more easy 
Really cool, but the UAC problem is occurring under Vista. When Tim’s rake tasks tries to run ‘patch’, UAC is triggered. If let patch run, then another cmd window opens, but nothing happens. I have to close the window, at which point the rake task fails.
It runs great if I first run the console elevated, though. Perhaps this needs to be pointed out in Tim’s instructions?
Yes, patch, install, installer and setup are reserved words that trigger UAC
That’s plain silly.
Any candidate for a Pure-Ruby Unified Patching solution?
It sounds almost too obvious, but could patch not be temporarily renamed (or permanently symlinked or something) so that it doesn’t trigger UAC?
Patch is part of MSYS.
If we play with it and break it we need to alter the recipes to take care of that.
AFAIK the patching is only required while building Ruby 1.8.6…
I’ll see what can be made to workaround this, but getting the new recipe format working has highest priority.
wow that is dang cool