[Talk-GB] Deletions and newbie editors (was: Vandalism in London)

David Woolley forums at david-woolley.me.uk
Sun Oct 5 09:47:08 UTC 2014


On 05/10/14 00:55, Frederik Ramm wrote:
> If you want to delete something for a legitimate reason - meaning:
> because it just isn't there on the ground

Even then it is a dangerous operation for newbies, as objects may have 
tags that need to survive, and which may not be displayed in iD's 
default view.  A classic example is NaPTAN stop data, where the rule for 
one that  has gone away is to invalidate the bus stop tag and add 
physically_present=no, but leave the node present.  I think I've seen 
cases where a stop being moved has triggered an delete/add operation 
that has lost he NaPTAN tagging.

Although I can't remember seeing this, I can also imagine a newbie 
deleting the node when a shop closes, rather than keeping it as a vacant 
shop.

Maybe newbie editors should clear the tags and set a fixme, rather than 
actually deleting tagged features.  They could then hide the feature 
from the newbie, whilst it is still visible in advanced editors.  I 
know that real deletions don't actually delete the object from the 
database, but the object is not loaded into editors, so people cannot 
see that somebody removed something, so it is much more difficult for 
someone familiar with the area, but not with the specific object, to 
spot that, sometime in the past, the map was damaged by a newbie 
deletion, and it is much more difficult for them to find the object and 
the changeset to revert, to avoid a re-add.

I've excluded untagged nodes, as deleting them may be expedient when 
re-modelling a way, although I'll try to avoid deletion, even then.  A 
difficult case is something like a gate, where keeping the node, either 
means detaching it from the way, or leaving a node that affects the path 
of the way.

A weaker approach would be to require newbies to delete all tags and 
relation memberships, one by one, before the editor will accept a 
delete.  Unfortunately newbie cookbook guides tend to come up with the 
easiest solution, not the right one.  In the contexts I most see, that 
is to turn off all the security features, but in this case, it might be 
a mechanical deletion of all tags and memberships, without applying any 
thought.

iD's relation deletion is a sort of mechanically applied example of the 
cookbook approach I fear.  The API won't let it delete the object, so it 
removes the memberships rather than questioning the operation.

Having several decades of experience in the software industry, starting 
with JOSM was very easy for me, but I wonder whether newbie editors 
really benefit newbies.  To me, iD is actually a little more difficult 
than JOSM to use, even for simple things.




More information about the Talk-GB mailing list