[OSM-talk] Mechanical Edit?

stevea steveaOSM at softworkers.com
Tue Jul 27 22:31:12 UTC 2021

On Jul 27, 2021, at 3:17 PM, Frederik Ramm <frederik at remote.org> wrote:
> Yes, we have definitely reverted undiscussed map edits where 99% of edits were an improvement and 1% were making things worse, just to prove a point. This is necessary, otherwise people just ignore the rules willy-nilly and say "but you didn't revert it when that guy did it".
> Almost every time someone proposes to do something stupid in OSM, when faced with criticism, the response is "but a similar thing has been done in the past".
> My response is always, the fact that some things are shit in OSM isn't a license to add more shit.

The difficult part of this is being certain that "crap data are crap data" and "let's improve the map."  Perfectly good intention (with the latter) doesn't ALWAYS imply the former (that crap data are entered).  There are exceptions to this.  It is difficult to characterize this, and even as one tries, it looks like either favoritism or randomness.

I know OSM isn't (the fictional book) "Zen and the Art of Motorcycle Maintenance" but there is something elusive and important (at the same time) about "Quality."  We are somewhere around "there" in this discussion and strong language (or DWG membership) doesn't make one person more of an expert than another.  "Standing tall" (as a very experienced editor, longtime contributor, DWG member...) helps, it offers historical perspective, yes:  but it does not guarantee that someone's distinctly clever "gather what appears wrong" (even if mechanically created, as by Validator) and "then, fix these apparent errors" is ALWAYS wrong.  It isn't.

There really are ways to long-term-improve our map data.  Rather than take easy shortcuts for overburdened honestly-inspired do-gooders (as-well-for-the-map-as-is-possible), we must discuss how to deprecate obvious noise.  Granted, given history, given experience, given the present and future, this remains very difficult.  It is the widest spectrum of lifecycle-oriented data gardening possible we might discuss (old cruft finally gets tossed), this makes what is best to do quite difficult.  (There are many perspectives and potential perspectives as input to the equation).

More information about the talk mailing list