Separator committer lists are generally frowned on by Apache (ref Lucene/Solr and also Mahout).
They also are essentially unnecessary. It isn't that big a deal to have a single committer list and depend on reversion to enforce community standards. Core ZK is review-then-commit and should continue to be. Add-on modules that are relatively independent of the core can quite reasonably be commit-then-review or whatever level of care is implied by module's content. The review can include a determination of who should actually do the commit. Moreover, projects can easily have more than one artifact. Mahout has three independent artifacts (mahout-math, mahout-collections and the core stuff with examples and integration code). That works well. There are different sets of committers who are typically the ones who work on different components. The overall level of diligence required in Mahout is much lower than for Zookeeper as you might expect, but the principle that multiple artifacts are fine is still the same. Likewise, the idea of committers with specialized expertise also applies here. There is a tradition in Zookeeper of having a very high bar for committership, but experience in some other projects indicates that this doesn't actually help quality or security all that much and it can be argued to decrease community involvement a bit. On Wed, Nov 30, 2011 at 5:11 PM, Patrick Hunt <ph...@apache.org> wrote: > Curator has been getting a lot of interest and was just officially > announced by Netflix: > > http://techblog.netflix.com/2011/11/introducing-curator-netflix-zookeeper.html > > I chatted briefly with @adrianco @chad_walters @randgalt (Jordan > Zimmerman) on twitter and the idea of moving Curator into > Apache/ZooKeeper came up: > https://twitter.com/#!/search/phunt%20curator > > I wanted to restart this thread as a way of discussing this idea in > more depth, and to gauge the interest of both communities. I > personally think this would be a great thing: both for users and for > the development community itself. Thoughts? > > Re the mechanics. See my original post below. Seems to me we (ZK PMC) > could sponsor Curator as an incubator project with the intent to bring > it to ZK as either a subproject (it's own committer list) or as a > separate release artifact of the ZK TLP itself. Or just shortcut the > incubator process altogether. My preference would be to go incubator > first. At some point we would deprecate the existing ZK recipes (ala > Curator had suitable replacements). > > Patrick > > On Fri, Jul 22, 2011 at 10:59 AM, Benjamin Reed <br...@apache.org> wrote: > > while i agree with the sentiment of not fragmenting the zookeeper > > community and recipe committers also moving into core development, i > > also think it would be good if a strong community of developers grew > > up around zookeeper recipes. to do that they need a sense of identity, > > and i'm not sure if they can get that with this proposal. > > > > on the other hand the proposal does address all of the other issues i > > have with recipe maintenance. the key is to grow a committer base. > > > > ben > > > > On Wed, Jul 13, 2011 at 11:04 AM, Patrick Hunt <ph...@apache.org> wrote: > >> During the recent developer meetup we discussed how to more > >> effectively grow the recipes. > >> > https://cwiki.apache.org/confluence/display/ZOOKEEPER/June2011DeveloperMeeting > >> > >> The concerns seemed to be around managing changes, community growth, > >> and releases. We discussed a number of approaches: > >> * move recipes into apache-extras > >> * or github > >> * or incubator > >> * or sub-project > >> > >> each of these has it's good and bad points. One alternative that I put > >> forth was to create a separate release artifact for recipes as part of > >> the ZooKeeper project itself. What would this look like? > >> > >> * zookeeper/trunk/src/recipes would move to > >> zookeeper/recipes/{trunk,branches,tags,...} > >> * there would be distinct/separate releases of zookeeper and > >> zookeeper-recipes artifacts > >> ** these releases would not necessarily be synchronized, although we'd > >> want to ensure that particular versions of each release work together > >> ** these releases would go through our standard release process - ie > >> creation of release candidate by a committer and approval by the PMC > >> ** recipes would no longer be included in "zookeeper" (core) releases > >> * continue to use the ZOOKEEPER jira, with "recipes" component > >> * space in the web and wiki sites for recipes specific pages > >> * recipes continue to use the std zookeeper mailing lists > >> * commits to recipes would continue to use our regular commit process > >> (jira, rtc, etc...) > >> * we would accept new "ZooKeeper committers" with the intent that they > >> focus on their area of expertise, whether this be core (zookeeper > >> trunk) or recipes. > >> > >> This last item is the more interesting I think, the other issues are > >> all pretty straightforward. Although moving to sub (or out entirely) > >> would allow us to have a separate commit list it would fragment the > >> ZooKeeper community. I also believe that over time committers to > >> recipes would more likely get experience with core and start > >> contributing there as well. The only issue really, if you want to call > >> it that, is that we'd have committers with "guardrails" implied. ie > >> someone with more expertise in recipes than core. However there are > >> many other Apache projects that take this approach, and do it > >> successfully. Regardless RTC allows for review both before and after a > >> change is committed. > >> > >> What do you think? > >> > >> Patrick > >> >
