Understanding Lucene up and down migration

[Stuart Halloway <st...@gmail.com>]

I have several related questions about migration, at the file format level.

1. Having gotten my 3.x codebase working with 4.x APIs, is it OK just to
open existing files created by 3.x and keep on indexing?   The answer
appears to be yes, with some performance hit, per
http://lucene.apache.org/core/4_0_0/changes/Changes.html#4.0.0-alpha.changes_in_backwards_compatibility_policy
.

2. Having written some indexes with 4.x, is it OK to open and use them with
3.x?  The answer appears to be no, by testing and inference from e.g.
http://lucene.apache.org/core/4_0_0/core/org/apache/lucene/codecs/lucene40/package-summary.html#package_description
.

3. My codebase is a library, so I cannot stop people from using e.g. maven
exclusions to pick their own favorite version of Lucene (and they often do,
since lots of different libraries use Lucene).  This, combined with #2
above, creates a problem.  Is there a recommended way to *detect* that an
index was created with an incompatibly-newer version of Lucene?

4. It seems like the answers to these questions would be important to ...
everybody upgrading Lucene, ever.  I expected to find the answers at
http://lucene.apache.org/core/4_0_0/MIGRATE.html.  The fact that others
don't seem to have these issues makes me think I am missing some key fact
or piece of documentation.  If so, please edify me.

Thanks!
Stu