[OSM-dev] Re: [OSM-talk] Re: Data types; indexes etc.

[Richard Fairhurst]

Steve wrote:

> [moved to dev list]

Hallelujah :)

> If you're talking about nodes, then it's because you can move a node
> independently of whatever segments it might be connected to. Remeber -
> we have history data to deal with. Also doing it this way means we can
> do routing easily - we don't have to deal with 'well this segments end
> point is near that ones start point sorta, so lets just say they are
> connected'. Our way, you _know_ if they're connected or not.

Yes, it makes sense for the wiki model.

Worth noting, though, that for some applications (e.g. non-editing map  
views) we're delivering a product of the data, not the data itself.

So if it's more efficient, we could create tables of lines as a  
derivative from the nodes/segments/ways tables, which would be updated  
only on edit of the source data. Then use these to render the maps. I  
guess it depends which apps are causing the most load.

Richard