FillValue

[castelao]

During the meeting on Jun 14 it was raised the question if FillValue should be locked, i.e. fixed value, or allow the user to define. There is an advantage to leaving it free, since it is clearly defined in the file and any software can parse it, as long as it is the same data type. Seems like we are inclined toward a free value.

Emma raised an important point that if it is a free value it would be a problem when aggregating multiple deployments. For instance, ERDDAP would take the first or the last attribute.

Since we haven't clear a path to go, we decided to move the discussion here to give us some time to think about it.

[vturpin]

During our last meeting, we agree that aggregating data set together should not be our driver to define the format. I do not have strong opinion on this.
This is a benefic / risk balance I guess.

[castelao]

Well, as long as it is clear that this is a free value, clearly defined on _FillValue, we can transfer the responsibility to the aggregator to inspect and conform the aggregating elements.

My suggestion is to let it be a free field, suggesting the use of NaN for float and double.

[vturpin]

I am fine with that !

[castelao]

If no one step in, I will open a new PR on the first chance.

[callumrollo]

I believe this was resolved in #114 @castelao can you confirm? If so, suggest we close this Issue

[justinbuck]

Discussion 24th April 2023 - recommended fill values are agreeable.
Fill values to be defined in OG 1.1