[Qgis-psc] PSC on Monday

[Richard Duivenvoorde]

On 04-02-18 00:04, Nyall Dawson wrote:
> On 3 February 2018 at 03:52, Paolo Cavallini <cavallini at faunalia.it> wrote:
>> Hi all,
>> on Monday I'd like to talk also about the proposal of removing grass and
>> saga providers from Processing. Is that OK for all, or better keep on
>> discussing on qgis-dev?
> 
> 
> Hi Paolo,
> 
> I fully support discussing this option and making an official decision
> - BUT - will Victor or Alex be invited to discuss? It seems odd to me
> to discuss the future of an area of code of which they are the
> maintainers without their direct input.
> 
> I'm very concerned that the outcome of this discussion will be "keep
> the providers in core, QGIS can maintain". Which effectively is just
> saying "Victor, Alex, here's a bunch of work we've decided you HAVE to
> do". (Slightly exaggerated in order to convey what I'm trying to say
> better).
> 
> If the discussion is just "on a theoretical level, do we want to ship
> out-of-the-box providers which rely on 3rd party applications which
> are not always available (or usable) on QGIS installs, or should these
> ALWAYS be published through the plugin repo", then ignore my concerns!
> The issue of manpower and if anyone wants to maintain them can be
> decided later.

Yeah, I'm with Nyall the burden is now always on the same shoulders...
But taking it OUT of QGIS would make a default QGIS install really much
less valuable isn't it?

I like the idea of having some plugin like solution.

Isn't the core problem:
- either QGIS or the 3rd party app changes api
- processing uses 'providers' as glue, so they should go with the changes
- different platforms have different QGIS / 3rd party combinations
(installed)

Could a ( theoretical :( ) solution be:
- have providers be (some kind) of plugins(sets) of which are only
'valid' for given QGIS - 3rd party app VERSIONS
- QGIS (or better processing) upon first start checks if the provider
plugis is for given QGIS/appl VERSIONS, and if not disables it with a
warning?

Maybe if we then provide good documentation for the providers more
people can pick up the burden of keeping api changes running?

The keeping stuff uptodat/running is always a rather 'dull' part of
software developement, which we can maybe encourage more by providing
bounties of targetted funding?

Just my 2ct,

Regards,

Richard Duivenvoorde