Improve Federation aspects

[m-mohr]

see https://docs.openeo.cloud/federation/#eodc

[sophieherrmann]

Fixed with 44465ce

[m-mohr]

We should also include:
Are there any unsupported features that VITO supports that you should mention in the federation aspects? Especially thinking about some important processes, file formats (e.g. COG support) or other API functionality that is crucial to common use cases?
I.e. this page should document the differences between back-ends the user could stumble across, this may also need some additions on VITOs side. Right now the whole page is pretty high-level.

[soxofaan]

As noted in #12: I don't have a good view on the practical differences between the back-ends that are relevant to end users, as I mostly play with the VITO back-end.

Maybe people that have more experience with using multiple back-ends like @clausmichele have some quick notes to share?

[clausmichele]

The main difference I have experienced are:

1. Different processes available: VITO has the highest number of processes, whereas EODC now has some particular ones dedicated to UC6 not available at VITO.
2. Different extent in the data coverage
3. Different band naming for the same satellite, for example Sentinel-2:
   https://openeo.vito.be/openeo/1.0/collections/TERRASCOPE_S2_TOC_V2
   https://openeo.eodc.eu/v1.0/collections/boa_sentinel_2
   at EODC there are some missing bands and the name of the presents now is not correct
   

[m-mohr]

Thanks @clausmichele - quick comments on your points:

1. This should be neglected in the aggregator for now due to the fact that the intersection of the processes is exposed. Although that makes it confusing if processes used for a use case are missing there.
2. & 3 I don't see as big issues right now as users would specifically select a collection and then check the corresponding metadata. So this is a bit more clear to users as e.g. process differences. But long-term this is something to improve for sure.

[clausmichele]

Well, in my opinion it would be better to avoid confusion with different band names, we can't assume that the user will look at the metadata if he sees that the band names provided are the ones he knows already.

[m-mohr]

Sure, I agree with that. I just meant to say that at least it's documented whereas other aspects/differences are not even documented right now.

[soxofaan]

Some quick stats about the differences in processes between EODC and VITO/Terrascope at the moment (just based on process id, not parameters):

• EODC lists 102 processes
• VITO lists 113 processes
• 90 processes are common : ['absolute', 'add', 'add_dimension', 'all', 'and', 'any', 'apply', 'apply_kernel', 'arccos', 'arcosh', 'arcsin', 'arctan', 'arctan2', 'ard_normalized_radar_backscatter', 'array_apply', 'array_contains', 'array_element', 'array_filter', 'array_find', 'array_labels', 'arsinh', 'artanh', 'atmospheric_correction', 'between', 'ceil', 'clip', 'cos', 'cosh', 'count', 'divide', 'e', 'eq', 'exp', 'extrema', 'filter_bands', 'filter_bbox', 'first', 'floor', 'gt', 'gte', 'if', 'int', 'is_nan', 'is_nodata', 'is_valid', 'last', 'linear_scale_range', 'ln', 'load_collection', 'log', 'lt', 'lte', 'mask', 'max', 'mean', 'median', 'merge_cubes', 'min', 'mod', 'multiply', 'ndvi', 'neq', 'normalized_difference', 'not', 'or', 'order', 'pi', 'power', 'product', 'quantiles', 'rearrange', 'reduce_dimension', 'rename_labels', 'resample_cube_spatial', 'round', 'run_udf', 'sar_backscatter', 'save_result', 'sd', 'sgn', 'sin', 'sinh', 'sort', 'sqrt', 'subtract', 'sum', 'tan', 'tanh', 'variance', 'xor']
• 23 only at VITO: ['aggregate_spatial', 'aggregate_temporal', 'aggregate_temporal_period', 'apply_dimension', 'apply_neighborhood', 'array_concat', 'array_create', 'constant', 'date_shift', 'discard_result', 'filter_spatial', 'filter_temporal', 'get_geometries', 'histogram', 'load_disk_data', 'mask_polygon', 'mask_scl_dilation', 'raster_to_vector', 'read_vector', 'resample_spatial', 'resolution_merge', 'sleep', 'vector_buffer']
• 12 only at EODC: ['ard_surface_reflectance', 'create_raster_cube', 'cummax', 'cummin', 'cumproduct', 'cumsum', 'dimension_labels', 'drop_dimension', 'fit_curve', 'load_result', 'predict_curve', 'resample_cube_temporal']

[sophieherrmann]

As mentioned in https://github.com/openEOPlatform/architecture-docs/issues/32 the Level-2 Sentinel-2 / Landsat-8 band names are now aligned between VITO and EODC.