No lowercase on key name in bulkloader

[tucotuco]

Right now the source_id is lowercased before using it to create the key name. Since this is the only record we have of the source_id, it should not be altered.

[eightysteele]

Yeah, so in thinking about this some more, Record should probably have a couple of new properties like source_id_column, source_id_value, and maybe even source_filename.

[tucotuco]

Don't think any of that is necessary. Need a compelling use case. But, definitely need an unaltered source_id somewhere.

[eightysteele]

Don't think any of that is necessary. Need a compelling use case. But, definitely need an unaltered source_id somewhere.

Right, so a publisher downloads record annotations and now needs to link them back to their source data. If they've published records with different source_id columns, then they'll need to know which source_id_column belongs to which record so that they can make the join on that column with the source_id_value.

[tucotuco]

Not a realistic scenario. source_id is the collection-level unique identifier, by definition - the closest thing to stable information in the source data. If they are changing source_ids - all bets are off, and knowing what file and field it came from originally doesn't help to re-integrate annotations with the original data.

[eightysteele]

So @tucotuco and I face-to-faced this issue a bit. The basic use case to support here is publishing CSV files that are generated from different data sources having different source_id columns. This comes into play when publishers want to link these records in VertNet back to their data sources. I think we agree that it's unclear how useful this would be. So, we're tabling the issue for a bit while we think on it and get some feedback from others on the team.

[tucotuco]

+1