On Thu, Sep 17, 2009 at 2:56 PM, Chris Hostetter <hossman_luc...@fucit.org> wrote: > : If it's an example with all of it's jars, it seems like it's still a > : copy of all those jars, right? > > correct -- either way *we* have to make a copy of those jars somewhere > in the release in order to demo Solr Cell.
We could do it like how I changed the clustering component to work... it no longer copies jars (but we do copy config around then): $ cd example $ java -Dsolr.solr.home=../contrib/clustering/example -jar start.jar That would shave off 30MB from the download. > My point is that right now anyone who downloads a release, and makes a > copy of "example/solr" to use as a set of starting configs (which is > what we frequently advocate) winds up with a really *huge* solr home dir > right off the bat because of all the extraction libs -- and extraction > isn't exactly a "core" feature that everyon is going to want, so it > doesn't really seem appropriate to provide it in the main example. > > If the main example was a kitchen sink example showing off *everything* > and there was another "skeleton" exampel that would be fine ... but we've > already gutted a bunch of cool stuff from example/solr in the interest of > makining it "svelt" so this seems contradictory. My main concerns with the explosion of the example directory was more one of complexity, not size. A person looking through that directory may have a difficult time telling what is what. It started off as one logical server, but morphed into multiple. A single server that had more capabilities out of the box (but didn't use resources if you didn't use those features) seems desirable to me. If it weren't for the extra 30MB download size, I'd probably argue for keeping SolrCell in the example server... but I think I'm leaning toward trying to configure it like the clustering example. -Yonik http://www.lucidimagination.com
