[OSM-talk] [Imports] Mechanically Cleaning Up FIXME Tags

[moltonel 3x Combo]

On 27/02/2015, Christoph Hormann <chris_hormann at gmx.de> wrote:
> fixme=stream␣attributes␣missing
> fixme=stream␣attribute␣data␣missing
>
> have not been added by an import but in an attempt to fix a broken
> import.  As far as i can see they indicate the waterway in question
> lacks a tag indicating if it is permanent or intermittent while the
> source data (NHD) generally contains such attributes and in the area in
> question a large fraction of waterways are non-permanent.
>
> Sensible ways to fix this particular problem would be
>
> (1) on the ground survey/check via aerial images (obviously)

Aerial images won't let you decide between stream=intermitent and
stream=ephemeral. At best you can rule out stream=permanent. So on the
ground survey it is.

> (2) deriving the corresponding tag from existing 'nhd:fcode' tag (like
> for example here: http://www.openstreetmap.org/way/45949047)
> (3) newly getting the corresponding tag from the NHD source data (by
> matching with geometry or id)

That looks like a good idea, that kind of data would be interesting to
import. 3) should be more robust than 2).

Note that in this context, having the objects tagged with fixmes is useless :
 * the list of stream ways with no stream type can be easily obtained
using an overpass query
  * if importing the stream types from some existing source, the
import won't care wether stream=fixme is set or not (it should of
course warn about other stream=* values and leave them as-is).

Even though there may not be an existing QA tool that checks the
presence of stream=*, these tags seem to be as useless as
fixme=name_missing. That doesn't mean that I'm eager to see them
mass-deleted. IMHO the harm they do is about on par with the version
churn of a mass delete would do.

As a comparison, getting off-stream-topic for a bit, in Ireland we
imported placenames from GNS a while ago, with a bunch of gns tags
included just in case at the time and now deemed useless. We still
have 7k such objects but we're not planning a mass-cleanup, we're
quite happy to manually cleanup whenever we happen to be updating the
place for some other reason.

> (4) remove the whole data where it has not been been modified by manual
> edits
> Given the  generally sorry state of the NHD/Canvec waterway data in America option
> (4) might be the better thing to do (we are talking about stuff here
> that has not been touched for five years).

Is the data so bad that you would consider removing it regardless of
its fixme tags ? That seems a bit strange to me, but I admit not
having looked at the data much. I've said it before, but the presence
of bad fixme tags *should not influence* the decision to delete the
object or not.