Obviously being a newbie (to OSM, not mapping itself), I wonder what is the preferred way of adding new (POI) nodes, like shops/restaurants in a strip mall or around an intersection?
Assuming that the strip mall or intersection in question doesn’t show any POI type nodes at all yet, is it better to add each single shop/restaurant/feature of it one by one, or is it preferred to add a complete set of them at once?
I am sure the preferred way is to map all the businesses in every strip mall on all corners of each intersection complete with full addresses, opening hours, etc. I fall short of that ideal almost all the time.
My feeling is that the map is an exercise in successive approximations where as long as my edits make the map better (more complete and/or more accurate) then I’ve helped the cause. So if I notice unmapped businesses at an intersection I will take a photo and then when back home use satellite imagery from Bing or Mapbox to place the building(s) and then add as many of the POIs I can remember or see on my photo(s). Usually that is not all of them. And, unless explicitly mapping the strip mall, I probably don’t have a street address or opening hour information, etc. And I might have missed many, even most, of the businesses in the complex. But the map will have been improved. And later some other mapper, or maybe even me when I have more time, can fill out that area more completely.
In the meantime, mapping and routing software like Osmand will show the POIs I’ve added so people can find food, fuel, etc. and get directions to them which they couldn’t before.
Sure, completeness of the end result is the ultimate goal.
My question was more about “how big to make the individual change set”, which as I understand the process is from one “SAVE” to another “SAVE”.
Being well established in Open Source, I personally appreciate “one feature” commits. But that approach may not apply to OSM.
Saves can be as huge as you want.
Actually big saves makes you work more understandable, as long they are properly commented and centred around some topic, like editing numerous different features around limited area, or maybe some feature around huge area, e.g. “roads, sidewalks, transit, POIs near Dmitryvska street” or “railway electrification in NE USA”.
Having an individual save per each POI, especially without comments makes you profile and edit history unreadable.
I wouldn’t worry too much about this, although I would say it is better not generate very large numbers of change sets. If you do generate very large numbers, I would expect them to have commit comments that describe the scope of each one, not just “adding more POIs”.
What you mustn’t do is generate very large change sets that sneak in things other than ones matching the primary nature of the change. Also, if you do anything that might be controversial, you should keep those items in separate change sets. Basically consider how you can break things down so that reverting the controversial part will not lose the other information you have added.
From my point of view, the main thing people do wrong with change sets is doing a large number of changes without an adequate and detailed comment, and often with no comment at all. If can be very difficult to work out what a large edit was actually trying to do. Looking at your history so far, it seems to me that you do understand how to use change set comments.
Thanks for the quick replies.
This matches the practice of large open source projects, I am working on. Easy enough to follow.