On Jan 27, 2012, at 01:31 , Gunther Diemant wrote:

>> You're right, that's much better -- apart from the fact I have yet to
>> come up with a single good reason for dynamic class names.
> 
> Test cases in a dsl test language like MiniTest/Spec or RSpec
> 
> describe "Foo" do
> end
> #=> Class FooTest < MiniTest::TestCase

Not (currently) _strictly_ true. describe does make a new class and gives it a "name" but it doesn't actually bother to set_const on anything. It used to, way back in the day, but I gave it up because I couldn't decide what to do given A::C vs B::C and I didn't want anything breaking because of it. 

That's not to say that dynamically creating named classes is a bad thing, just that it didn't fit in this case given the inherent complexities of tests.