[OSM-talk] Data Model, Relationships, AND

Frederik Ramm frederik at remote.org
Fri Jul 6 20:09:44 BST 2007


    some people have used the opportunity of AND CC'ing their data to 
point out, rather smugly, that AND use a similar data model to ours - 
implicating that ours can't be the problem that I and other advocates of 
more complexity view it to be.

I'm eager to see how AND model house numbers, turn restrictions, 
multiple ways on one bridge, and so on, and I will no doubt have a very 
good look at these things once they're released.

But one thing is clear even now: You can pretty much tolerate anything 
if you operate in a closed environment, where data is either static or 
edited only by one and the same in-house proprietary software.

For example, the often-voiced idea of simply linking objects by having 
tags pointing to other objects' ids (e.g. "key=no-turn-into, 
value=segment-id-217237234" or so) would work perfectly if we wouldn't 
have to fear that someone, by moving, inserting, or replacing segments, 
breaks this connection or multiplies it and creates an invalid link. But 
since we're open, anyone can access the API with any tools he or she 
wants, and that's why we can't do that - we can never ensure that all 
editors comply to some unwritten rules as long as these are not enforced 
by the API.

I'm just writing this so that people do not make the mistake of assuming 
that anything which is sufficient for a closed, proprietary data set 
used by one company is also good for a Wiki-like world which we're building.

I would have thought that's obvious but obviously not to everyone ,-)


More information about the talk mailing list