HI, > Right now if you delete a node in potlatch it just removes itself > from relations using a bunch of SQL. Richard says this is what you > said to do.
Well he offered me the choice of either refusing to delete the object, or removing it from the relation, and I replied that I preferred the latter, adding: > We will not be able to go on like that forever because this leaves > unneeded relations lying around in the database; on the other hand > these can be identified and removed automatically, so it is not a big > problem for now. > > The long-term solution surely must be something where the user is > informed "object part of relation - (a) cancel deletion, (b) > remove object from relation, (c) remove relation altogether", with the > editor being able to make smart choices from some "type" tag. But it > would be too early and time consuming to do any work on that now > before people have even understood what relations are about. I still think that the a/b/c thing would be ideal, but if we haven't got that then the silent deletion that Richard currently does is not the worst thing. After all, if the user wants to delete a way then that's what he's going to do, no matter whether we tell him that he's breaking a NCN route by that or not? > What I think should happen is that until potlatch supports > relations it should prompt the user that you need to delete the > relation, much like you can't delete a node that is part of a way? Ideally it should ask for the a/b/c choice outlined above. > Richard doesn't want to prompt the user to use another editor and I > see that but lean towards not breaking data - for example removing > something from a turn restriction relation and suddenly the > relation makes no sense. There is a danger of breaking relations (leaving them meaningless) but we can easily find the meaningless ones in the planet and see what happened to them. So I'd say it is not a high-priority thing; in the long run I hope that Richard is going to put in an a/b/c prompt but I think of all methods not requiring a lot of work, the current method is the one least irritating (less irritating than saying "you cannot do this" and surely less irritating than deleting the object without removing it from the relation ;-). One of my to-do-things is to comb through the planet and see what relations people are using, and somehow document that and provide tailor-made editors for these kinds of relations in JOSM, this will hopefully give us a clearer picture what people want/need in terms of relation support in editors. Bye Frederik -- Frederik Ramm ## eMail [EMAIL PROTECTED] ## N49°00.09' E008°23.33' _______________________________________________ dev mailing list [email protected] http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/dev
