Even Rouault

In the GEOS context, a custom memory allocator would require all references to C++ containers to be changed from ``std::vector`` to ``std::vector`` (cf https://stackoverflow.com/a/826635). For example the ``std::vector vect`` member...

> I think that GDAL will align if GeoZarr emerges as a standard. :) It depends on what emerges :-) but yes the current status of the Geo part in...

fix in https://github.com/qgis/QGIS/pull/53566 . Opening time goes down to ~ 10 s on my machine

The issue with coerceToType() is that it is sometimes real coercitiong, like transforming a polygon to a line, or the reverse, which is probably you don't want to do behind...

When I look at https://docs.ogc.org/is/18-000/18-000.html "Figure 2. Related Tables UML Diagram", there's no "references" relationship from gpkg_contents to a user mapping table, which would suggest they shoudn't be referenced in...

I was trying to think about the drawbacks of referencing a mapping table in gpkg_contents. One that I see is that a user could potentially drop such table, if not...

Actually another possible issue with adding the mapping table to gpkg_contents is that: - if a hypothetical software would manage the Related Table Extension according to its letter - and...

> So that means that both existing GDAL releases and ESRI software can't access the mapping tables **at all** if we follow the specification exactly if you need a workaround...

> Should we then force remove the table from gpkg_contents? 🤷‍♂️ Probably ? I dunno... Let's just put it in gpkg_contents: there's no explicit requirement preventing it from being added...

@nyalldawson I guess the reading part will be for another pull request? (not totally sure if we'd want to backport it)