What makes 0.2.x special

The 0.2.x branch is maintained separately. Because 0.2.x continue using java regex grammar to define your lexer rules.
Any version newer than 0.2.x(that is, at least 0.3.0) would use a simpler and more restricted lexical rules' syntax to describe your DSL's token schemes.
The implementation behind 0.2.x's generated lexer is also the java.util.regex functionality stuff.
This doesn't mean that the 0.2.x way is totally worse. It depends, I think. When jdk1.7 being popular, 0.2.x could gain a performance boost in its newly 'capturing groups' way. So, 0.2.x would continue evolving and growing in the future unless some day it is proven being out performed by newer versions in both convenience and speed.