I'm ok with the ivy thing, I just wanted to raise the non-disconnected build issue, since moving to ivy is not 'no cost' as proponents would like to paper over. I just wanted to make sure everyone realizes this will hurt productivity of some people.
On Thu, Jan 7, 2010 at 10:52 PM, stack <st...@duboce.net> wrote: > On Thu, Jan 7, 2010 at 9:42 PM, Ryan Rawson <ryano...@gmail.com> wrote: > >> So given all that, I'd rather do the whole hog, rather than the half >> and half thing. Ie: maven over ivy. >> > > > I'm not opposed to a move to Maven. I've not really given it much thought, > but I think that a separate discussion and undertaking. It would take some > work doing it properly. We would need to take on the maven layout at a > minimum -- move all to src/main/java, etc. -- and we'd need to purge any > deviation from the maven way because the alternative is hours burnt > wandering in the weeds of poorly documented plugin xml configs., or worse, > hours writing custom maven plugins to pull Maven in alternate directions. > For one, our notion of contrib (src/contrib/*), IIRC, does not map to > maven's notion of subprojects. Its been a while but with Maven subprojects > notion, you could not without backflips have the parent project build its > jar and then have subprojects depend on parent. > > If someone wants to take on the Maven work, well and good but for me the Ivy > work is done. Lets commit it. The way it does its dependencies is > Maven-like (you list them in pom for maven, in properties for ivy; both pull > to local caches, etc.). Committing Ivy gets us working with external > repositories, pulling and publishing. So, the ivy commit takes us some of > the ways toward a mavenized hbase while meantime, making hbase build like > its hosting project and its siblings. > > St.Ack >