On 4/26/11 3:46 PM, Mark Miller wrote:
It really seems to come down to robert/simon and yonik being really opposed here. This bubbles up every few weeks. I just don't know how to fix that - does anybody? I want to fix things as much as anyone - but what now? We have hashed everything over and over.
I totally agree with Robert and Simon that it is currently very frustrating that moving code to Lucene is being veto'ed on. They're investing time in realizing the modularization goals we had for the Lucene/Solr merge, but there are always roadblocks. We should really thank them for this effort, because it's very honorable work: making an existing feature available for more users.
I think, Yonik, that you ignored the concerns of other committers in SOLR-2272. It shouldn't be a surprise that this leads to frustration. If it's a goal of the Lucene/Solr project to have shareable modules for common code then in order to get a new feature committed discussions about where the code should live have to happen very early in the phase of a new feature. And then you need to scope the necessary development time accordingly.
It doesn't necessarily even have to take longer to develop a new feature as a shared module vs. one that only lives in Solr. What usually takes much longer is to develop it first in Solr, then after the release was made refactor it and move it to a module, with the burden of having to maintain backwards-compatibility with the original implementation.
It's a responsibility of every committer to eg. make sure that Solr tests pass when a new Lucene feature is developed. It should also be the responsibility to figure out early on where code should live. I can not emphasize this enough.
