You can clone with
In Java land, it makes sense to move to a maven-based build. This would make the build simpler for pycascading and also make it easier for external contributors to build the project. At the moment, the build consists of a lot of hard-coded paths in Ant files which isn't good. Furthermore, moving to Maven would allow us to contribute the pycascading artifacts to Maven central which would make things even easier for potential consumers.
I agree, I expect to see a few of the supporting libraries at certain places within the three main libraries that I use. If the dependencies.properties file is pointing to the right places where Jython/Hadoop/Cascading is downloaded, this should be OK, but in general a Maven build would be awesome and a todo.
Between maven central and conjars.org, I'm pretty sure the necessary libraries are available (even cascading 2.0 stream).
This is more of a nice to have at the moment until you decide to tag and release pycascading somewhere.