From Kyle Grucci from Oracle:

I've read through the doc a few times. I think the info from this email
(and more) needs to be explained in the TCK user's guide. Optimally, the
TCK bundle should run out of the box (without the need to download/clone
anything else aside from an implementation to run against). For
example, the CTS has everything needed to run against a Java EE impl
aside from an impl to run against (i.e. the RI). In our user's guide,
we then explain what porting interfaces and ant targets must be
implemented to run the CTS against another impl. There is no assumed
knowledge of Maven dependencies, external git workspaces, etc. The
steps to get up and running must be explained in detail without the
assumption of maven. If you want people to use maven to execute the
TCK, then I would think all steps to get up and running must be itemized....

-install maven
-create a runner dir for your impl
-<describe how to create a runner>
-describe how to have the tck execute in your container of choice
-describe where your arquillian.xml file must live, what's in it, etc.
-describe how to create a LoadableExtension and how the Java
ServiceLoader mechanism will use it

Just telling licensees to clone a ws, cd to a dir, and run mvn might get
the tests to run against the ri, but leaves out alot of info on how to
get the tests running against another impl. In our case, how would you
describe (in detail) the steps to setup the TCK and run it in the
containers of Glassfish?

Thanks,

Kyle