On 6/13/2013 10:02 AM, Alexandre Rafalovitch wrote:
P.s. As an aside, being relatively new, I do wonder what kind of event/discussion will trigger version 5 branch-off. I guess it would actually be more of a Lucene decision these days.
I foresee two likely reasons for a branch-off. Based on what I've noticed over the last year or so, the latter reason is probably more likely. I don't think that the decision to branch will depend much on whether the improvements are in Solr or Lucene, mostly because they tend to improve together, not separately.
1) Trunk gets really awesome features X, Y, and Z that aren't included in 4x, and it becomes stable enough that the committers believe the entire codebase is almost ready for ALPHA. This reason is more likely if those features are major performance enhancements.
2) Backporting fixes from trunk to branch_4x becomes too difficult due to code divergence. Right now, almost all changes backport with no difficulty, and even when there are conflicts, they tend to be easy to resolve.
Over time, various factors will cause trunk code to become very different from 4x, and backporting will become a nightmare. Examples:
A) Pieces of code that get deprecated in 4x are removed completely from trunk. B) Trunk requires Java 7 and 4x must work with Java 6. C) There are differences that appear because some changes are too major and not tested enough to include in the stable branch.
Thanks, Shawn
