[Rebuild] Logging - post-Rebuild OSMI?

[Martijn van Exel]

Hi,

On 4/8/2012 1:40 PM, Frederik Ramm wrote:
> Hi,
>
> there will be some damage after the rebuild process.
>
> There will be many *wrong* things in OSM after the change. Disfigured
> ways that have lost one or several nodes; roundabouts that aren't round
> or buildings that aren't square; forests that self-intersect; ways that
> were closed but aren't now; former T-shaped junctions where both streets
> are now not connected any more.
>
> I'm not as concerned about nodes or ways entirely missing because *not*
> having some information is less of a problem than having erroneous
> information; but still ways that go missing will occasionally break a
> relation or cause a routing engine to go crazy.
>
> Much has been said about remapping prior to the license change but what
> kinds of tools will we want to have to help us *after* the change?

That's a very valid question. I am just assuming (but I never actually 
asked, doing that now) that CLEANMAP / BADMAP will remain available in a 
frozen state after the redaction bot has started doing its work.

That's a start, but after the 'damage is done', there may be a few 
things you would want to know about a feature when editing.

> Will we perhaps want some kind of logging output from the bot that runs
> the redaction?

I would say definitely. I don't know if we can legally add redaction 
information as a tag to redacted nodes (like redaction:nodes_removed=* 
for a redacted way, redaction:tags_removed=*, ...) - but that would make 
it easier to spot how much was lost in the process, and give (re)mappers 
something to focus on.
>
> Some operations - e.g. "removed a start/end node from a way" would
> almost definitely lead to someone having to check on the result because
> such an edit is likely to break an intersection or open a formerly
> closed way.
>
> Some things would not be precisely loggable because the log would then
> have to be CC-BY-SA licensed, but even then one could could record of
> often certain operations happened in a given map square or so.

It would be interesting to know to what extent the redaction information 
(not the deleted data of course, but what and how many redaction 
operations were applied to an object) can be included in the ODbL data. 
But in order to focus remapping efforts after the fact it would be nice 
to have an 'affectedness index' with map tiles (z14 and up or 
something?) as the unit of measure. To keep it relatively simple you 
could use a limited set of metrics, like number of deleted nodes with a 
weighing factor where deleted nodes that are intersections (2+ records 
in way_nodes) or that meet some requirement for a POI are applied a 
greater weight.

Another thing that would be nice to see is ways that have more than 
25/50/75 percent of nodes removed, or as you say the first or last nodes.

I wonder though, would you need redaction logging for this? You can get 
a pretty good estimation of all these things from WTFE + a snapshot of 
the data before the redaction starts, or can't you?

> Bye
> Frederik
>


-- 
Martijn van Exel