Working copy in multiple GPKG

[pfw]

I've been having some discussions with some folks who would like to use Kart to manage some datasets but they are being tripped up with a couple of things, one of which is how the working copy is laid out relative to datasets when they are dealing with multiple GPKG files which are part of a project. These files have overlapping table names across the multiple files, eg. something roughly similar to having multiple country files from https://gadm.org/download_country.html where each file looks a bit like

ADM_ADM_0                    rtree_ADM_ADM_0_geom_node  
ADM_ADM_1                    rtree_ADM_ADM_0_geom_parent
ADM_ADM_2                    rtree_ADM_ADM_0_geom_rowid 

The files can have up to 5 tables but not all files have all tables.

Where this causes issues is that the Kart import will clash by default and will put all tables into one working copy even if the clashes are dealt with at import time with TABLE:NEW_TABLE. Having the files merge into a single working copy messes with the existing processes they have and since there is a long history of they way they do things they are disinclined to change that.

Is there something fundamental which requires all datasets to live within a single file for the working copy? I been thinking through and the best I could come up with is that having a single working copy file means that all of a 'pull' can be done in one transaction and having multiple files might mean some are updated and others not if the 'pull' fails. That failure case sounds like it is something that can be dealt with through some atomic flag files to at least indicate an out of whack state.