Open Stack

On 2014-02-12 13:48, Adrian Otto wrote:
> On Feb 12, 2014, at 10:13 AM, Ben Nemec <openstack at nemebean.com>
>  wrote:
> 
>> On 2014-02-12 09:51, Jesse Noller wrote:
>>> On Feb 12, 2014, at 8:30 AM, Julie Pichon <jpichon at redhat.com> wrote:
>>>> Hi folks,
>>>> Stefano's post on how to make contributions to OpenStack easier [1]
>>>> finally stirred me into writing about something that vkmc and myself
>>>> have been doing on the side for a few months to help new 
>>>> contributors
>>>> to get involved.
>>>> Some of you may be aware of OpenHatch [2], a non-profit dedicated to
>>>> helping newcomers get started in open-source. About 6 months ago we
>>>> created a project page for Horizon [3], filled in a few high level
>>>> details, set ourselves up as mentors. Since then people have been
>>>> expressing interest in the project and a number of them got a patch
>>>> submitted and approved, a couple are sticking around (often helping 
>>>> out
>>>> with bug triaging, as confirming new bugs is one of the few tasks 
>>>> one
>>>> can help out with when only having limited time).
>>>> I can definitely sympathise with the comment in Stefano's article 
>>>> that
>>>> there are not enough easy tasks / simple issues for newcomers. 
>>>> There's
>>>> a lot to learn already when you're starting out (git, gerrit, 
>>>> python,
>>>> devstack, ...) and simple bugs are so hard to find - something that
>>>> will take a few minutes to an existing contributor will take much
>>>> longer for someone who's still figuring out where to get the code
>>>> from. Unfortunately it's not uncommon for existing contributors to 
>>>> take
>>>> on tasks marked as "low-hanging-fruit" because it's only 5 minutes 
>>>> (I
>>>> can understand this coming up to an RC but otherwise 
>>>> low-hanging-fruits
>>>> are often low priority nits that could wait a little bit longer). In
>>>> Horizon the low-hanging-fruits definitely get snatched up quickly 
>>>> and I
>>>> try to keep a list of typos or other low impact, trivial bugs that
>>>> would make good first tasks for people reaching out via OpenHatch.
>>>> OpenHatch doesn't spam, you get one email a week if one or more 
>>>> people
>>>> indicated they want to help. The initial effort is not 
>>>> time-consuming,
>>>> following OpenHatch's advice [4] you can refine a nice "initial
>>>> contact" email that helps you get people started and understand what
>>>> they are interested in quickly. I don't find the time commitment to 
>>>> be
>>>> too much so far, and it's incredibly gratifying to see someone
>>>> submitting their first patch after you answered a couple of 
>>>> questions
>>>> or helped resolve a hairy git issue. I'm happy to chat about it 
>>>> more,
>>>> if you're curious or have any questions.
>>>> In any case if you'd like to attract more contributors to your 
>>>> project,
>>>> and/or help newcomers get started in open-source, consider adding 
>>>> your
>>>> project to OpenHatch too!
>>>> Cheers,
>>>> Julie
>>> +10
>>> There’s been quite a bit of talk about this - but not necessarily on
>>> the dev list. I think openhatch is great - mentorship programs in
>>> general go a *long* way to help raise up and gain new people. Core
>>> Python has had this issue for awhile, and many other large OSS
>>> projects continue to suffer from it (“barrier to entry too high”).
>>> Some random thoughts:
>>> I’d like to see something like Solum’s Contributing page:
>>> https://wiki.openstack.org/wiki/Solum/Contributing
>>> Expanded a little and potentially be the recommended “intro to
>>> contribution” guide -
>>> https://wiki.openstack.org/wiki/How_To_Contribute is good, but a more
>>> accessible version goes a long way. You want to show them how easy /
>>> fast it is, not all of the options at once.
>> 
>> So, glancing over the Solum page, I don't see anything specific to 
>> Solum in there besides a few paths in examples.  It's basically a 
>> condensed version of https://wiki.openstack.org/wiki/GerritWorkflow 
>> sans a lot of the detail.  This might be a good thing to add as a 
>> QuickStart section on that wiki page (which is linked from the how to 
>> contribute page, although maybe not as prominently as it should be).  
>> But, a lot of that detail is needed before a change is going to be 
>> accepted anyway.  I'm not sure giving a new contributor just the bare 
>> minimum is actually doing them any favors.  Without letting them know 
>> things like how to format a commit message and configure their ssh 
>> keys on Gerrit, they aren't going to be able to get a change accepted 
>> anyway and IMHO they're likely to just give up anyway (and possibly 
>> waste some reviewer time in the process).
> 
> The key point I'd like to emphasize is that we should not optimize for
> the ease and comfort of the incumbent OpenStack developers and
> reviewers. Instead, we should focus effort on welcoming new
> contribution. I like to think about this through a long term lens. I
> believe that long lived organizations thrive based size and diversity
> of their membership. I'm not saying we disregard quality and
> efficiency of our conduct, but we should place a higher value on
> making OpenStack a community that people are delighted to join.

I'm not disagreeing, but there has to be a balance.  tldr: there are a 
lot of pitfalls in the submission process, and if they aren't 
well-documented they're going to frustrate new contributors to no end.

For example, let's say someone reads through just the Solum page (if 
they follow the link to GerritWorkflow then they're back to the big, 
scary documentation we have now).  They try to submit their change.  It 
fails because they have no ssh keys on Gerrit so they can't connect.  
Casual contributors probably give up here because they followed the 
instructions and they didn't work.

But let's assume they visit review.openstack.org and get an ssh key 
registered on Gerrit.  They try again.  This time they fail due to an 
unsigned CLA.  Okay, sign the CLA (assuming their employer doesn't 
restrict them from doing that, but that's a whole other discussion).  
Hooray, Gerrit will accept their submissions now!

But wait, now they put the entire commit message in one block of text 
because they don't know any better.  Let's say worst-case-scenario that 
they're submitting to a project that's gated on Tempest so it takes an 
hour or more for Jenkins to vote, and an hour later it fails due to the 
pep8 violation in the commit message (note that there doesn't seem to be 
a mention of running tox before submitting in either set of 
documentation).  Fair enough, they fix their commit message to have a 
first line <72 characters as pep8 tells them to.

They submit again.  Their submission sits for a week because the project 
they're working on has a long review queue.  Maybe a reviewer comes 
along and notices their commit message has an improper bug reference and 
-1's.  At this point I'd be throwing things around the room. :-)

Granted, this is a very pessimistic scenario, but I'd bet most new 
contributors hit some subset of these issues on their first submission 
because our existing documentation is either too concise and doesn't 
cover these steps, or is too long and rambling and therefore they miss 
important details.  As I said below, as simple as possible, but no 
simpler.  Where that line is probably varies by person, but I think we 
can all agree we're missing the mark right now.

> 
> Focused efforts like the one outlined by Julie Pichon are exactly what
> OpenStack needs to make on-boarding smoother, and more pleasant. Thank
> you Julie for sharing what you are learning, and helping us improve.

Absolutely +1!

> 
>> That said, the GerritWorkflow page definitely needs some updates and 
>> maybe a little condensing of its own.  For example, do we really need 
>> distro-specific instructions for installing git-review?  How about 
>> just "Install pip through your distro's package management tool of 
>> choice and then run pip install git-review"?  I would hope anyone 
>> planning to contribute to OpenStack is capable of following that.
> 
> Optimize public facing documentation for throughput of attracting new
> members, not maximizing efficiency of the existing ones. We can decide
> how to deal with internal inefficiency as it presents as a problem.
> Our default mode should be welcoming.

No argument here, but I still think we should be able to assume a 
certain level of competency in our prospective contributors.  In a 
process as complex as the one we're dealing with, a digression to 
discuss every possible way a package might be installed is just noise 
IMHO.  Right now the length of GerritWorkflow is not welcoming.

> 
>> I guess my main point is that our contributing docs could use work, 
>> but there's also that "as simple as possible, but no simpler" thing to 
>> consider.  And I certainly don't like that we seem to be duplicating 
>> effort on this front.
> 
> All efforts that result in more open source, and more community
> participation should be encouraged. Don't worry as much about
> duplication of recruiting efforts. Trying a few different approaches,
> and seeing what works best is a great approach to something like this.
> Attracting a community is not a science. It's an art form. Making art
> is about trying a bunch of things, and seeing what works best.

Sure, but it's a wiki.  Rather than creating a completely separate page 
that almost entirely duplicates an existing one, why not fix up what we 
already have so everyone benefits?  Right now the Solum page is only 
useful if you're looking for Solum information, and the instructions 
there apply to every single OpenStack project.  And if someone disagrees 
with your changes they're welcome to make their own.  It's the wiki way. 
:-)

I guess I see the Solum page as a sort of cheat-sheet for people who 
have already been through the longer form Gerrit workflow page, and I do 
think that would be useful.  I just think it would be useful to 
everyone, not only Solum developers.

> 
> Thanks,
> 
> Adrian
> 
>> /2 cents
>> 
>> -Ben