[Qgis-developer] Project quality discussion

[Alex M]

On 11/06/2015 05:21 AM, Hugo Mercier wrote:
> Hi all,
> 
> We had a discussion here in Las Palmas about the overall quality of the
> project.
> 
> The main concerns / questions (I had) were:
> 
> - big organizations are starting to fund QGIS. This is great but it is
> still a bit hard for a company to sell the development of a new feature,
> because it is hard to guarantee it will be integrated.
> 
> - we still rely too much on volunteer work. And the situation becomes
> complicated when a paid development depends on volunteer work.
> 
> These are my main conclusions.
> For people who attended, don't hesitate to complete if necessary. For
> the others, you are welcome to react of course.
> 
> - every new feature introduced by a core developer should be sent as a
> Pull Request first. With a given "quarantine" delay after which the PR
> will be merged, even if no reaction. It will allow to share information
> and react in case of problems (and encourage people for good work as well :)
> -> should I create a QEP for that ?
> 
> - if a company with no core developer wants to ensure a new feature is
> accepted, it should pay another core developer for the reviewing part.
> Ideally the money should go to the project and the project would decide
> what core developer(s) to pay.
> The details of this process are not very clear. It still has to be
> discussed. But the goal is to make clear for everyone that if you want
> guarantee: you have to pay for it and there is a clear process to handle
> that.
> 
> - writing a QEP before adding a new feature is a good way to increase
> its acceptance. But some people have to review it. We may come to the
> same process to pay for QEP reviews.
> 
> - at which point we rely on volunteer work is not yet clear. But the
> current guess is: still too much. Having a better idea of the ratio
> between free work and paid work would be profitable for the project: it
> would allow to make clear what the reality of an open source project
> like QGIS is and that too much free work is not sustainable. Paolo's
> mail is about that. The goal is to (begin to) separate clearly what is
> the part of free work and the part of paid work in the project.
> 
> - see on the PSC side if it is possible to pay some people to handle
> global maintenance : PR triage, reviews, small bug fixes and so on. It
> does not have to be only one developer.
> 
> Thanks for participating in this discussion.
> 

If a company wants something that makes it into Core, it should be
pre-approved via a QEP. The fallback should be a plugin.

I caution the use of $ to directly influence the QEP review and Pull
Request reviews. Some sort of bounty pool, allowing core devs to get a
stipend for clearing big reviews might be an option. So companies can
contribute to QGIS and that money can go to expedite all QEP and Code
reviews, but not a specific one. I'm mostly concerned with paid funding
driving the direction of the project into strange niches and increasing
the core code maintenance tasks. Plugins seem a much easier way to pay
for exactly what you need.

Free work vs volunteer work is tricky to draw a line between in a
project like this. I think you'd have to survey the devs about what %
they think comes from each in order to account for time spent at
salaried job helping QGIS, time spent improving QGIS to help you sell
services (consultants), time spent being directly paid to do X with
QGIS. There are tons of ways that time is not exactly volunteer.

Thanks for bringing up good things for discussion.
-Alex