> In an earlier mail I gave the following example: If we had a separate > analyzer module, and we add an analyzer in a new language to it, it would be > cool to release it soon, without having to wait until Lucene/Solr are ready > for a release. The pace here can be faster, because I imagine in such an > Analyzer module it's much less common that a patch "touches everything". > What do people think about this? Maybe it's just a nice wish, but not > realizable, because there'd be a lot of version management overhead. But > maybe not? >
Hi Michael, this is a great question. In my opinion, it would be nice for any analyzer to have some time to sit around in trunk before being released. In general we have been trying to treat contrib/analyzers with some special care and also maintain more backwards compatibility, etc. It helps if we can somehow shake out any problems before releasing them. But there is a a slight problem: I feel that new analyzers/improvements to existing ones in lucene trunk are really not being tested by many people, because Solr is not using lucene trunk. I think if they were available to the many Solr trunk users, then this would give us more feedback and better quality releases, just from an analysis-only perspective. -- Robert Muir [email protected]
