We should provide incremental dumps also for the JSON dumps.
This would probably be implemented like this: Have a script that dumps all entity ids that have been changed since the last incremental dump. Then just dump all entities on that list.
The first script is not yet implemented, but that shouldn't be to hard.
Potential shortcomings of this (that may or may not also apply to the other incremental dumps, I have no idea): Deletions and merges (that turn things into redirects) wouldn't show up that way.
I believe I originally asked for this, but current WDQ wouldn't use these anymore, and SPARQL replacements are on the way. In case I would have been the only customer, this task could be closed now.
Report here what I have writed in Wikidata:
The actual JSON dump compressed is more than 6 Gigabyte so, it's possible to create json dumps with only item changed/added from the previous week/dump? This allows for smaller files, and then you need less time to download and decompression. Useful for those who have slow connections
Is useful for bot operator that done periodic task of maintenance