[Rebuild] About keeping and removing tags

[Dermot McNally]

On 2 March 2012 17:53, Frederik Ramm <frederik at remote.org> wrote:

> The reason for this rather cautious approach is that I thought it is too
> difficult to determine whether the decliner's work was nullified by the
> later change, or built upon.

I understand the reason for your caution...

> Eg:
>
> decliner adds "name=Pepe's"
> agreeer changes to "name=Pepe's Pizza"
>
> certainly we cannot say "the agreer has overwritten the decliner's
> contribution and we can keep the name" in this case.

Agreed (well, sort of), and agreed also that it's hard to tell the
difference between such a case and a clear change in name (which BTW
I've sometimes seen in cases of node recycling [kids, don't do
that!]).

Why "sort of"? Because in the specific example given, the subsequent
agreeing mapper seems to demonstrate primary knowledge of the name of
the restaurant and has judged the shorter version worthy of a change.

But that doesn't get around the fact that some changes really are
nothing more than adding an apostrophe.

> They suggested that we use a more fine-grained approach that would somehow
> compare the strings, and only remove the name if the current version had
> sufficient similarity.
>
> I think this is worth discussing.

Me too. We'd have to get happy with an algorithm that seems to strike
the right balance, but it seems feasible.

> There are other tags where the situation is just as difficult. If someone
> tags "addr:housenumber=45" and someone later changes that to "13" then the
> value 13 can hardly be a derived value but what if the original is 45 and
> the corrected version is 450?

String similarity may work for these cases too.

> There are also tags where the situation is much easier. For example, any
> change from one of the ~ 20 common "highway" values to another one of these
> values could surely be said to remove the original mapper's copyright on the
> highway type because changing that is hardly possible without first-hand
> information. Or is it?

I tend to consider the changing to another highway tag as a "cleaning"
act. You might certainly argue that the non-agreeing mapper has told
us that the way is a highway and not something else. In any area
covered by Bing, though, this is a red herring, since this fact can be
verified by anybody (and presumably has been by the subsequent mapper
who chose to change the highway tag).

Dermot

-- 
--------------------------------------
Igaühel on siin oma laul
ja ma oma ei leiagi üles