[OSM-talk] compatibilit necessary OR reversing the meaning of tags: oneway on motorway_link (was: New dimension of vandalism)

[Michael Kugelmann]

Mikel Maron wrote:
> Your change earlier this month was pretty substantial.
[...]
> Personally, I'm not partial to either side of the argument, but willing to work with whatever the consensus is.
>   
I want to give a hint on a change that was made some times ago and is 
still today considered not OK by myself.
As Mikel I am willing to work with a lot of definitions. But it is not 
OK IMO to reverse definitions that had been existing and used for years 
=> afterwards our data are likeley to be wrong.

What happened?
The standard definition since I joined the project years ago was that an 
attribute has to be added to a object (node/way/area). This means for 
"oneway" that "oneway = no" was the default for any object. Suddenly 
some people decided that a "highway=motorway_link" would automatically 
include a "oneway=yes" even w/o having the attribute attached to the 
object (also valid to other "XXX_link").
This is a real problem as after this change a lot of motorway exits are 
simpley tagged wrong!
My conclusion is that we have to be very (!) careful when we change 
definitions, there has to be a compatibility between the old definition 
and the new one. If this is not considered, our complete planet data has 
to be reworked because we change the data to wrong by a "simple change 
of the definitions". I am not willing to check and correct the whole 
planet...


Best rgerads,
Michael.

PS: my sollution at that time was to explicitley add a "oneway=yes/no" 
to every new XXXXX_link road part I create or modify but I am not 
willing to check all the XXXX_links I once created, that should be a 
work for those who decided to change the definition...
PPS: the change in the definitions was made silent: no announcement on 
the MLs and forum or any good visible place so it was not noticed by a 
lot of persons for long time. That's another issue I  think about being 
a very bad behavious...