In various research projects the advantages of having a mirror API to separate reflection from a language implementation have been discussed, and "industry grade" implementations exist for Java and C#. This project aims at providing a number of specs and classes that document a mirror API for Ruby. The mirror implementation that is part of this project will use only those language facilities that are available across Ruby implementations. The specs, however, will also test behavior that cannot be provided in such a manner. The idea here is that in time, all implementations provide their own implementation of the mirror API, and all implementations collaborate on this one spec.
Why do this, you ask? Because Ruby needs tools, and those tools need to be written in Ruby. If they are not, then people will be excluded from tinkering with their tools, thus impeding innovation. You only have to look at Emacs or Smalltalk to see what's possible when programmers can extend their tools, all tools, in a language they feel comfortable in. If we have a standard mirror API, all tools that are written for Ruby, in Ruby, can be shared across implementations, while at the same time allowing language implementers to use the facilities of their platform to provide optimal reflective capabilities without tying them to internals.