lower the barriers to contribution, branching
I'd like to promote discussion on this topic again with a little haiku:
lives on, makes a difference
share a branch, git-style
practical beats purity*
just gate-keep for quality
coding wise DRY KISS!
community over code
*derivative attribution to line 9 of Zen of Python
--- to quote myself, as poets may do:
> What we do not want, and must try to avoid, are hard forks by the users
> (entities that take the code and deploy in the real world), where they
> long standing unmerged changes, and worst that these changes are
> incompatible with the upstream changes that are on the main fineract dev
> branch. This then leads to harder to maintain code at the users and more
> costly duplicative development for all. This is the opposite of the
> virtuous cycle.
On Fri, Aug 24, 2018 at 2:03 PM Myrle Krantz <myrle@xxxxxxxxxx> wrote:
> Hey all,
> Just FYI, I forwarded James Dailey's mail to Ross, and here was his
> response. (There's a little plug in there for ApacheCon as well.
> Anyone who hasn't registered yet: there's still time. : o)
> "Thanks Myrle,
> You are correct I do not actively follow the Fineract dev list
> anymore. Feel free to share this back to the broader community list if
> it helps.
> The below email from James is a good summary of our conversation at
> OSCON. This conversation was a general one about “The Apache Way” and
> how Apache projects deal with growing pains as their code matures. I
> cannot comment on how well James has applied those general lessons to
> Fineract today, but I certainly see plenty of good content within this
> If I were to summarize that general guidance in a few sentences it
> would be to remove as much friction from the contribution process
> (code reviews, merit recognition, community alignment) as possible.
> Generally speaking the lower the barriers to contribution the faster
> the community will grow. This does depend on people actively
> monitoring the project, but monitoring is less work than gate-keeping.
> Adopting things like Lazy Consensus can be key
> In the days of SVN we were forced by the tooling to operate with a
> branching model. This works really well. All changes are visible in
> one place. Work in progress can be easily discovered and reviewed. It
> means those monitoring the project have the opportunity to review work
> as it happens, thus enabling them to raise concerns about a design
> decision or implementation weakness early in the process. This in turn
> meant that when it was time to consider a merge most of the rough
> edges had ben talked about *before* they had become deeply imbedded in
> finished code. It was easier to fix and people would work together to
> design a fix that worked for everyone.
> As James indicates Git does not force this way of working. It has
> excellent support for the SVN concept of branching, unfortunately
> GitHub has driven most people to click “fork” (mostly invisible to the
> community) rather than branch. GitHub, therefore, has encouraged us to
> work privately then issue a pull request for review when the work is
> “finished”. This often means people are not keen to redo their
> implementation to satisfy the broader needs of the community. They
> will blame the community for “blocking” their improvements.
> I am a strong believer in doing as much as possible in the open at all
> times and reducing barriers to collaboration.
> Interestingly, I have a talk on this topic at ApacheCon this year, if
> only I’d written it already
> On Wed, Jul 25, 2018 at 7:00 PM James Dailey <jamespdailey@xxxxxxxxx>
> > Hi All -
> > For the good of this project, I'd like to share some ideas gathered and
> > shared in a side meeting at OSCON18 with Apache President Ross Gardler
> > was one of the champions of this project. You can read the official PMC
> > reports that go to the Apache Board here -->
> > https://cwiki.apache.org/confluence/display/FINERACT/Board+Reports
> > I am not a member of the PMC, nor a committer, but I have been involved
> > from the beginnings of this in 2002. So, I am hoping to share both the
> > short term and the long term view. As most of you know, the Mifos
> > Initiative contributed the code to Apache and remains - as an external
> > entity - highly interested in ensuring the continuation and growth of the
> > project. In the Apache worldview, Mifos offers a kind of
> > of the *project*, and the hope is that many more such entities - for
> > companies in particular - will productize, and contribute back via the
> > community of developers, requirements and ideas. In other projects we
> > within Apache, contributors can be paid by companies to make sure that
> > their priorities get attention. Those companies and entities provide a
> > of "wrapper" around the project and can provide things like dashboards,
> > add-ons, and deployment scripts. Thus a virtuous cycle is born and
> > supported.
> > What we do not want, and must try to avoid, are hard forks by the users
> > (entities that take the code and deploy in the real world), where they
> > long standing unmerged changes, and worst that these changes are
> > incompatible with the upstream changes that are on the main fineract dev
> > branch. This then leads to harder to maintain code at the users and more
> > costly duplicative development for all. This is the opposite of the
> > virtuous cycle.
> > If there are large unmerged changes that can be proposed for either
> > Fineract1.x or for Fineract-CN, I believe a key way forward would be to
> > make those branches visible. Fortunately, and tongue firmly in cheek,
> > is a mechanism available in git conveniently called a "branch". I think
> > the PMC should consider this approach to bring into the fold those
> > entities that are on forks (via the individual contributors) and then to
> > have a clear process by which a serious attempt to evaluate and accept
> > changes into the main branch are undertaken. It is probably naive of me
> > think that the point of forking is that clear based on a defined release,
> > but one can hope. In any case, the project would be much assisted if code
> > that is written for real world situations is made visible for merit-based
> > evaluation and inclusion. That can, and probably should, exclude
> > productizations (plug-ins, deployment scripts, UIs, report
> > that give companies a differentiation in market. However, underlying code
> > changes that make those things work better need to be contributed back so
> > that the “wrappers” can be a kind of patch that is easily maintained on
> > of the fineract release. If you are part of one of those companies,
> > now do comment on what is holding you back, and make an attempt to move
> > of your infrastructure to the latest stable fineract release (identifying
> > issues as they arise).
> > The other thing that we should strive to avoid are PRs that sit around
> > remain un-merged. (as noted by PMC) This is an obvious problem made
> > I believe, by having some number of contributions that may not have
> > anything to do with the needs of the broadest set of actual users. If
> > community is out of sync with the users, which is possible in a project
> > that is NOT involved in a direct "scratching the itch" kind of thing.
> > (referring to the axiom that most opensource projects are developers
> > scratching the itch for software that works for their needs). To solve
> > problem, the Apache board recently heard about a cool innovation that
> > obvious in retrospect: allow non-committers to review and comment on
> > proposed Pull Requests, thereby determining their priority and earning
> > non-committer points and merit towards committership. Also, we should
> > a cultural project norm here where a few things can change around PRs:
> > 1.
> > Committers should be free to merge if no objection is heard (a time
> > frame of 72 hrs is probably ok, to be set as “community norm”)
> > 2.
> > Merge and review - rather than review and merge should be adopted by
> > PMC
> > 3.
> > Releases must be scrutinized but the “tip” or “head” of dev can be
> > merges that may be subject to review and revert-backs
> > 4.
> > If you break it, you unmerge it (tests coverage is your friend!!)
> > For more information on project maturity at Apache, please read
> > By the way, and now speaking with my non-profit Mifos hat on, a key
> > of moving the code over from Mifos to Apache was to broaden the community
> > and broaden the appeal. When I say "Mifos contributed" I also mean to
> > all of the contributors to the Mifos project, who worked on it as an open
> > source project from 2005 to present, are part of that. It is accurate to
> > say that the Mifos community was already an active one and a key
> > accomplishment over the past two years is bringing over the code and the
> > community to Apache. But more needs to be done to clarify.
> > Mifos community code (also released under apache 2.0) is now a wrapper on
> > top of fineract. Fineract can include binary releases for convenience but
> > the code is the thing, not the productization. Mifos is also continuing
> > play an important role in organizing the community of financial inclusion
> > around fineract - I submit that that is not inconsistent with the PMC
> > trying to market fineract to both the financial inclusion community and
> > private sector interested in payments, banking, etc. As a non-profit,
> > Mifos has much of the same operating imperatives as Apache, but with a
> > narrower focus on financial inclusion. We probably need advice from our
> > apache friends how to address this dual role.
> > Finally, I am inspired by what so many have accomplished on the mifosX →
> > now fineract 1.x codebase and what is promised by the fineract-CN code.
> > continue to envision fineract in the broadest and more inspiring terms.
> > Opensource will eat the world and the financial world is only beginning
> > be heard from us.
> > Please do comment on this post and suggest ways to operationalize or
> > and say why. THANK YOU!!
> > James Dailey
> > Fineract’r
> > Board Chair, Mifos