[QGIS-Developer] A plea: more volunteers needed for reviewing backports

[Martin Dobias]

Hi all

On Tue, Mar 23, 2021 at 12:07 AM Nyall Dawson <nyall.dawson at gmail.com>
wrote:

>
> This is a public plea for more developers who are very familiar with
> different parts of the QGIS codebase to become actively involved in
> backport PR management.
>

(Nyall later clarified this is not only about backport PRs, but all reviews
in general)

Thanks for starting this thread - it is a discussion we definitely need to
have. (And apologies for getting back to this soooo late!)

Pull request reviews are absolutely vital part of the QGIS development, a
chance to get bugs fixed before they even get into QGIS code. Quality
reviews also need a good amount of expertise of the QGIS code - often the
hardest part of a review is not the code included is the pull request, but
figuring out what is missing...

Speaking of myself, I used to review pull requests regularly... But after
several years I have to admit I mostly gave up doing that unless someone
asks me to do a review. The pace of QGIS development is not getting any
slower (which is great!), so there is a constant flow of new pull requests
and doing code reviews regularly is not something I want to do in my free
time... I am happy to do some QGIS work in my free time, but only doing
what I want to do :-)

For a company, strictly business speaking, sparing 15 minutes a day of a
senior developer is roughly equivalent to lost profit of few thousands of
EUR (assuming ~50 hours / year). And many reviews need much more time than
15 minutes... Moreover companies doing QGIS dev are often already donating
to QGIS as sustaining members...

In a mail in the thread it was suggested that companies doing QGIS
development should add extra cost to quotes to accommodate the time for
reviews (of unrelated pull requests). Not sure I agree with that - if a
company had constant income from QGIS dev, that's doable, but if we are
talking about occasional QGIS dev work, that is hard to plan.

>From all of that above, my thinking is that in order to make things
sustainable, regular pull request reviews should be ideally funded by
QGIS.org similarly to how paid bug-fixing sprints work. It is the kind of
project maintainance work that needs to be done, it is not always super fun
and it requires input of someone from a small group of people that are
already donating lots of their free time.

My proposal would be therefore along these lines:
- PSC allocates annual budget to reviews
- core devs interested in participating would indicate their availability
(eligibility may be the same as with paid bug fixing)
- PSC tells devs how much paid time they can spend on reviews
- paid devs should do reviews regularly, e.g. at least twice a week,
ideally every day - not just once a month or so
- paid devs would self-assign themselves to PRs and do reviews
- if a PR is not picked up by anyone e.g. within 3 days, PR queue manager
would assign it to one of the paid devs
- paid devs keep track of their time in a spreadsheet and invoice
(quarterly?) up to the amount they were allocated

I believe this approach should solve our problems:
- remove stress from growing PR queue and reviewer burnout
- get more core devs (who otherwise may not be available) to do reviews
- reduce frustration from devs submitting PRs when their PRs are not
getting attention

In my humble opinion, good quality reviews are even more important than the
regular paid bug fixing or grants. A review that is rushed due to lack of
time may omit important code details, or focus only on code style...

We could start with a relatively small budget and compensate the extremely
valuable work that reviewers (Nyall and others) are already doing.

Regards
Martin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/qgis-developer/attachments/20210501/63a2defc/attachment.html>