You may use the eodatasets3 library to index data hosted remotely, I've done it by indexing official Copernicus Global Land Cover data you can find in this S3 bucket (data access is documented here).
By using the
DatasetPrepare object, specifying a remote dataset location, you can call the
note_measurement method by passing a remote file (i.e. a GeoTIFF tile), the library takes care of downloading the file, reading the metadata, and generating an EO3 document you can index in ODC, without the need of keeping the file locally afterward.
DATASET_LOCATION = 'https://s3-eu-west-1.amazonaws.com/vito.landcover.global/v3.0.1/2015/E080N60/'
TILE_URI = 'E080N60_PROBAV_LC100_global_v3.0.1_2015-base_Discrete-Classification-proba_EPSG-4326.tif'
p = DatasetPrepare(
p.note_measurement(measurement_title, DATASET_LOCATION + TILE_URI)
dataset_doc = p.to_dataset_doc()
Notice the usage of
allow_absolute_paths as specified in the docs.
Unfortunately, the library doesn't work for multi-band files (see issue), in that case, you have to implement a preparation script to create the EO3 documents, but I'm sure this doesn't prevent you to reference remote datasets in the EO3. Hope this helps.