Re:Dolt is Git for Data
by hinkley on Thursday June 23, 2022 @12:08PM
(#31851109)
In a similar vein, I am trying to work on a tool for a volunteer group, and one of the problems I'm trying to figure out how to model is not just tracking when the data changed, but why it changed.
Sometimes when you're first entering data you just get it wrong. You said the wrench was in storeroom A1 or the customer lives on 3rd St or the tree was five feet from the sidewalk. If the state of the asset changes due to human action, that's a new thing. But if you opened the wrong list taking inventory, or you missed a keystroke for 33rd St or you just eyeballed the coordinates, then that row was simply wrong, and the correction should be back-dated for most things.
But if I emptied out A1 because we aren't renting that space anymore, the customer moves or the tree got replanted, then it was here and now it's over there. Which might be important for figuring out things like overstock, taxes or success rates.
Similarly if the migration introduces derived data, then the migrated data is assumed/inferred, whereas if we took that data from a user, that information is confirmed, which might introduce subtle differences in how best to relate to the user. Things a mediocre business could easily ignore but a quality establishment might be ill-pleased with such a request.
Sometimes when you're first entering data you just get it wrong. You said the wrench was in storeroom A1 or the customer lives on 3rd St or the tree was five feet from the sidewalk. If the state of the asset changes due to human action, that's a new thing. But if you opened the wrong list taking inventory, or you missed a keystroke for 33rd St or you just eyeballed the coordinates, then that row was simply wrong, and the correction should be back-dated for most things.
But if I emptied out A1 because we aren't renting that space anymore, the customer moves or the tree got replanted, then it was here and now it's over there. Which might be important for figuring out things like overstock, taxes or success rates.
Similarly if the migration introduces derived data, then the migrated data is assumed/inferred, whereas if we took that data from a user, that information is confirmed, which might introduce subtle differences in how best to relate to the user. Things a mediocre business could easily ignore but a quality establishment might be ill-pleased with such a request.