Robbi Bishop-Taylor

That's a good idea! Plus it would avoid needing to come up with/use some kind of consistent naming convention for those fields

@ateucher This would be fantastic! Python's `pystac_client` handles pagination automatically, and it would be fantastic to have it in `rstac` too

Update: turns out that a missing nodata attribute on floating point rasters also impacts `gdal` too - e.g. overview generation is messed up if a floating point raster contains nodata...

> NB. In 1.9, these functions should be reimplemented to use `odc.geo.xr.write_cog` and `odc.geo.xr.to_cog` respectively. > > (Do these odc-geo functions already have this functionality?) Good call - it would...

Hey @Kirill888, the original issue we encountered is that floating point COGs without a NODATA value flag are incompatible with ESRI software (ArcMap, ArcGIS Pro), which is unable to recognise...

Here's a clearer example of the GDAL incompatibility: Example `float32` array with mostly NODATA on the left and values of 1 on the right, except with a narrow strip of...

> > The question is then should we set nodata=nan by default when creating float32 arrays with nan based masking also? > > I think not. This is a workaround...

@SpacemanPaul @Kirill888 This seems like a bug - writing out datacube data should preserve those important datacube attributes. However, it also makes me think we need to think about relocating...

Copying chat from Slack here @SpacemanPaul @Kirill888: 

Superceded by #1152