On Fri, Feb 15, 2008 at 7:48 AM, Igor Brejc <[EMAIL PROTECTED]> wrote:
> Dermot McNally wrote: > > Cases where uploading fails midway through and can't (easily) be > > completed - due to precondition failure, for instance - suffer from > > the fact that data self-consistency is lost because of the nodes-first > > policy. In such cases, and I've often been victim to it, you end up > > with a stack of nodes and no easy way of laying the ways back down. I > > usually end up deleting the nodes and starting again. > > > > It's a pity the API can't support rollback in these cases. > > > > Dermot > > > > > Wouldn't it be good for the next (or some future) version of the OSM API > to introduce some kind of atomic updates similar to SVN commits, on "all > or nothing" basis? I guess this is not the right discussion list to ask > this, but I'm just daydreaming... > > Igor > Yes, I think changesets are becoming more and more important. They would solve a lot of these problems, in addition to a few more (such as rolling back inadvertent or malicious changes). Karl
_______________________________________________ josm-dev mailing list [email protected] http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/josm-dev
