[Rebuild] We're better off detecting splits and merges

ant antofosm at gmail.com
Sun Feb 5 11:56:10 GMT 2012


Hi,

On 05.02.2012 08:37, Ilya Zverev wrote:
> On Sat, 04 Feb 2012 18:30:25 +0100, ant wrote:
>
>> If splits and merges are detected, the result will differ in the
>> following cases:
>>
>> 1. Dirty way is split by agreeing mapper
>> - Currently: The newly created way is marked as clean.
>> - If detected: The newly created way is marked as dirty.
>
> Let's have an example: a dirty way (v1 by non-agreeing mapper) is split
> in two by agreeing mapper, who put lit=yes on one part and lit=no on the
> other. In this case both parts would be considered dirty and removed?
> So, if we implement this algorithm without reconsidering removing v1
> objects by non-agreers, we'll lose a majority of surveyed data?

Yes... that's why I think the v1-by-decliner thing is useless. I'd 
rather regard the creation of an object as a modification like any 
other, in order to give subsequent modifications a chance of survival.
On the other hand though, we'd be able to preserve a considerable amount 
of clean ways split by a non-agreer - regardless of the v1 thing.

>
> Right now splitting a way does not remove dirty status from the nodes
> (obviously). So, which part of the result of a split is copyrighted? The
> list of nodes has changed and tags were checked by agreeing mapper (who
> would fix them if they were wrong, because he should select way to
> split). I'd say, both of the parts from splitting a way should be
> odbl-clean. Or, maybe, with some tags lost, but not entire ways as objects.

Splitting itself is not protectable, only the tags added/removed - the 
"lit" tags in your example - are accountable to the mapper performing 
the split.
Whether existing tags were checked against better sources can, 
unfortunately, not be detected by algorithmic means, so we must assume 
they are the work of the mapper who added them first.

cheers
ant



More information about the Rebuild mailing list