[josm-dev] Validator

Dirk Stöcker openstreetmap at dstoecker.de
Fri Jul 13 11:32:11 BST 2012


On Thu, 12 Jul 2012, Paul Hartmann wrote:

> I'd suggest an alternative: Remember the original version of each
> primitive; on upload run 2 passes for the validator: one for the
> (reconstructed) original dataset and one for the modified one.
>
> This solution isn't necessarily easier to implement, but it solves
> other problems as well: Currently, if you add a tag to a large number
> of objects, and later remove that tag, JOSM still keeps the modified
> flag for these objects, so they get uploaded and get a version bump,
> although the properties remain unchanged. In addition it would be
> useful to display the changes made to an object before upload (e.g.
> for a large relation; much like "svn status"). The history GUI could
> be reused for this.
>
> This still requires an extension of the data file format used in JOSM,
> but I think it's inevitable. There are a couple of other things that
> need to be saved to file, e.g. object histories and edit conflicts.
> The plain .osm format can be offered as export option instead.

I agree that this would be an overall improvement - To keep server version 
for each changed object. This would allow several improvements including 
prevention of unnecessary changes.

But I still would not reduce the displayed validator warnings. Maybe we 
could add a "this one is new by you", e.g. by different color, but overall 
I don't want to reduce the impact of the validator.

Again having a look at my father. When he changes objects and there are 
validator warnings, he will fix them (most of the time :-) But he probably 
never will call validator for the data set to start finding issues.

I think the validator helped a lot to improve the current map. I read the 
discussion in talk-us and especially the "crossing ways" is not an 
theoretical issue, but is essential to allow proper routing. We had lots 
of these issues some years ago, we have only very few which today 
improving the overall quality a lot.

For me the most disturbing warning is "missing name" and usually this is 
always right. OSM is missing a lot of names.

In my holidays is used Garmin+OSM for routing:
  * We have very good data (except in south Ireland)
  * Garmin's routing engine and display of routing commands is crap :-)
  * We are missing tons of meta-data like street names

Users changing data need to be notified about this or the situation will 
not change. And a user modifying an object may know the name and can fix 
it. It is not his fault that name is missing, but he may nevertheless be 
in the position to fix the issue.

Todays errors are usually hidden in the data and not clearly visible. And 
not everybody will add an OpenStreetBugs entry when something wrong 
happens on routing. I think the validator is an essential tool and I don't 
want to reduce its impact. And to say it again: We did a lot in the last 
years to make it more userfriendly and I don't like these discussion 
coming up again and again which are based on the assumption that the 
validator is a silly programmers perfection tool. It is not!

Ciao
-- 
http://www.dstoecker.eu/ (PGP key available)




More information about the josm-dev mailing list