[OSM-talk] Coastline Update

[Paul Norman]

> > Toby wrote:
> >
> >> Assuming you used the data I supplied this morning, it is actually
> >> from 12:30 AM CST last night when I started the jxapi query before
> >> going to bed.
> >>
> >> Also, it looks like http://www.wightpaths.co.uk/coast/CT-only.php
> >> has
> >> been updated with your new files since you sent your email.
> >
> > Perhaps I don't understand what I'm looking for, but have checked
> > the 4 points on the Scottish coastlines indicated on the link above.
> > I can't see what is wrong with any of them. Some of the ways were
> > edited on the 24th March 2012 which I guess might have fixed
> > whatever the issue was, but if the data is from (consults timezones)
> > yesterday-ish then I'm not sure why they would still show. The
> > points seem to correspond with Paul's Europe png too.
> 
> 
> I think the answer lies in the original comment from Paul [1]
> 
> "This is somewhat more aggressive than the rebuild will be,"
> 
> As far as I can tell the error points sometimes get displayed where
> there
> are what the JOSM  relicencing plugin calls "possible data loss".
> 
> As such the error points produced by Paul, and the two maps built upon
> that
> data are an indication of possible problems, not a 100% definite
> statement
> of actual problems.
> 
> Furthermore, as I understand it, none of the tools which find "dirty"
> ways
> and nodes are actually using the same logic as the actual rebuild will
> use,
> so its impossible to get a 100% accurate picture of what will happen.
> 
> Regards
> 
> David

Nodes and ways are removed by cleanway. The exact logic used is as follows:
Do nothing to v1 objects created by acceptors
Do nothing to objects reported as clean by WTFE the last run

Then, for ways:
Do nothing to ways with odbl=clean
Drop ways where WTFE reports a severity=normal problem with the first
version
Record the ways that were clean to avoid querying them next time

For nodes:
Drop nodes where WTFE reports a severity=normal problem
Record the nodes that were clean to avoid querying them next time

A dropped end or start node or a dropped way will result in a break and an
error.

If anyone has suggestions for improving the algorithm I would welcome them.
Keep in mind that I only have access to the current version of the object
and the WTFE response.