For runtime depends, I completely agree that we should discuss before
the code goes in.
For test and build time depends (like maven plugins and test harnesses),
I don't see why we need prior review, as long as the artifact is
available in the public maven repos and uses an ASF friendly license
(where something that used LGPL or GPL would warrant discussions) and is
not required by the runtime.
Guess the grey area, is optional runtime depends, like the OSGi Core jar
in the recent openjpa-example changes, where it was my bad that the
checked in code required OSGi at runtime instead of optionally
supporting it by adding the optional BundleActivator support....
-Donald
Pinaki Poddar wrote:
Hi David,
This discussion also brings up a more generic point.
As a group we should adhere to the discipline of bringing *any*
dependency upon build/test/runtime *after* discussing it in the appropriate
forum. The process of A stumbling upon B's commit to notice such 'dependency
injection' is inherently non-sustainable.
Do other members agree that, in future, before we need to any dependency,
we validate/discuss it in the group?
Regards --
Pinaki Poddar wrote:
Hi,
Noticed that some recent changes made OpenJPA HelloWorld to depend upon
OSGi.
Impressive but questionable change.
And the example JavaDoc says:
/**
* A very simple, stand-alone program that stores a new entity in the
* database and then performs a query to retrieve it.
*/
The classic "Hello World" was built on KISS -- and that principle still
remains valid especially for getting started examples. A separate example
should be constructed to demonstrate OSGi integration capability.
-----
Pinaki Poddar http://ppoddar.blogspot.com/
http://www.linkedin.com/in/pinakipoddar
OpenJPA PMC Member/Committer
JPA Expert Group Member
