[Qgis-developer] Release schedule discussion - again

[Nyall Dawson]

On 12 October 2015 at 22:47, Andreas Neumann <a.neumann at carto.net> wrote:
> Hi all,

>
> So should we really keep this fast release cycle with releases every 4
> months with the risk that these in between releases are barely usable for
> professional users or should we rather have one release every 6 to 12 months
> and invest more (financial) resources into these fewer releases? I would
> personally vote for fewer releases and increased QA resources. Maybe change
> to a release every 6 months with four months of development and 2 months of
> bug fixing/testing - and see how this works?

My thoughts (note that I'm not in a position where I have to maintain
an enterprise-wide rollout, so my viewpoint may be skewed):

- If you're using QGIS in a professional setting, you SHOULD be using
2.8 ONLY. I wouldn't recommend non-LTR releases as suitable in any way
for professional use. Treat the non LTR releases as betas or for hobby
use only.
- I also think LTR releases should be every 2 years rather than every
year. I also think that having a new release tagged as the new LTR
release shouldn't be immediate or automatic. Eg, let's say 2.22 (or
whatever) is the next scheduled LTR release. I don't think the LTR
should switch immediately when 2.22.0 is released. Instead we should
let the release settle and manually decide when it's suitable to move
the LTR from 2.8.x to 2.22.x. Maybe this will be a few weeks after
2.22.0, or maybe when 2.22.3 is released. In the meantime it's just a
LTR candidate, and not the recommended release for corporate use.

That said, I think the much more important issue which needs to be
settled ASAP is timing of QGIS 3.0 and the move to Qt5. Honestly, we
are extremely lucky that QGIS isn't totally broken on the latest OSX
release. Qt4 can't build on it, but the homebrew crew have managed to
patch out the broken sections and at least get Qt4 compiling on 10.11.
Who knows what kinds of broken behaviour and regressions are waiting
to be uncovered though... I'm VERY concerned that a future OSX release
(or windows update, or distro decision,...) will totally break Qt4 on
one of our platforms and we'll be forced to rush release a Qt5
version. This would be a huge shame - for QGIS 3.0 to be done properly
and be a nice stable base for the next couple of years we need a
double-cycle to clean up all the deprecated code and API hacks which
have built up over the 2.0 cycle.

We need to be in control of this decision and not have it forced upon
us, and it needs to be made soon...

Nyall