[OSM-talk] Good practice, and should we rely on defaults?

[stevea]

On Apr 10, 2022, at 2:31 AM, Martin Koppenhoefer <dieterdreist at gmail.com> wrote:
>> On 10 Apr 2022, at 04:02, Minh Nguyen <minh at nguyen.cincinnati.oh.us> wrote:
>> 
>>> name-suggestion-index had already problems with some items used by it getting
>>> deleted by Wikidata admins as were no considered as passing some standards there.
>> 
>> In every instance I'm aware of, it was either by mistake and quickly restored, or because the Wikidata item was insufficiently tagged and indistinguishable from spam.

See, in many cases here, what isn't clear is "whether I should use defaults like others do, too." And then it gets into whether you MUST and in what circumstances and special case exceptions and the logic can get nightmarishly complex quite fast.

Some will either invent or latch-onto-and-use "default lists" and others will wonder whether they should use them, or maybe prune them so they work better for THEM (and maybe not any longer for the use-case they were invented for / initially tuned for). And then you collide with how the invention of these particular defaults for this particular use-case did a terrible job of working well for that particular use-case. This is an old problem in data science: predicting how data will be used and thinking you can predict the future (maybe well, maybe not-so-well) by "goosing ahead" a slippery, easy path for those data to fit into your scheme. Then somebody else comes along with something even slightly different and maybe your invention for the scheme was poorly designed. But you don't always have the foresight to design for everything, that's what's hard.

It's not just that default data go stale / obsolete. It's that their scheme (as invented and deployed) might not be the best for everybody. Then, a tweak / tune of the data might be required, or a whole new scheme must be invented. Do we win doing this? Maybe yes, sometimes. Maybe no, sometimes. I think it's impossible to predict such things. However, there is something to be said for good (elegant, deeply future-looking, to the extent it can be, unambiguous...) design. I'm not saying we must not. I am saying it's hard: there will be success, there will be failure.