Hi Serge, On Fri, May 27, 2011 at 11:17 AM, Serge Wroclawski <[email protected]> wrote: > Martijn, I'm going to re-arange the sentences here to make it easier to > respond. > > On Thu, May 26, 2011 at 1:51 PM, Martijn van Exel <[email protected]> wrote: >> Hi all, >> >> Consider the following applicaton scheme: >> * a twitter user sends a geo-located tweet containing a specified >> hashtag, say #addosm and key-value pairs like "amenity:pub;name:Red >> Devil;smoking:yes" > >> This would be an easy way to add POIs on the go, and could be an >> interface for mobile applications to post new POIs. > > I see technical issues here. > > First, it seems like you'd only be able to add nodes. This would be > pretty sloppy, offer no mechanism for checking existing nodes, etc.
Yes, it would deliberately be constrained to nodes. Checking existing nodes would definitely have to be part of it, using (fuzzy) name matching within a radius of the new object. Fact remains that the location may be less accurate than desirable, due to a bad GPS fix in urban areas and more generally lower grade GPS receivers in smartphones. To me however, this does not outweigh the advantage of offering a low barrier to entry way of adding something to OSM. > Secondly, if you have a smart phone with twitter and geo-coding, why > aren't you using a real POI collector app? Because that takes away the main argument for doing something like this: offering a really easy way to add something to OSM, using existing tools that people already use. I talked about the idea at the ongoing WhereCampEU yesterday[1] and got some more feedback on it. One of the ideas was to restrict the scope even further by allowing to add only a certain type of features, doing away with the need to encode the key-value pairs. Picking pubs as an example, one would just tweet "@osmadd Bellman Bar" would add a pub named Bellman Bar[2]. > >> * a twitter scraper picks up the tweet, archives it and posts a new >> point using the twitter coordinate and the decoded k-v pairs, plus an >> additional tag source:twitter[@twitteruser] or something like that. > > This is okay, in my opinion if you've pre-authenticated this web > application to do edits on your user behalf, ie used OAuth. > >> This would not be >> totally anonymous but it's close. > > Why should it be anonymous? > > Let's image such a web application. > > It looks for twitter feeds and does what you suggest. I think it must > be per user. And if it's per user, then it can accept OAuth and act on > the user's behalf. And if it does that, then it doesn't need to be > anonymous. This would be a good scheme, something I would propose would be obligatory after doing a few contributions as an anonymous user. You would get a tweet saying 'thanks for your contributions, to continue, please link your twitter and OSM accounts here >> [url]'. > >> What do you think, is this acceptable? > > If it's taking random user data from Twitter, no way. > > If it's taking data from users who have authorized it to do so, I > think it's still a bad idea for technical reasons, but I see no issue > with it from that standpoint. > >> A similar level of anonymity is reached by WheelMap.org >> that allows anonymous OSM edits through their web site via the OSM >> account wheelmap_visitor[2]. > > This pattern has been tried before in OSM and is generally considered a > problem. > > WheelMap.org takes huge precautions but even that, I think, isn't enough. > > We're making it increasingly easy to have users sign up to OSM. We > have OAuth for external editors, and we'll likely have OpenID soon for > external authentication. If you see a way to simplify signup further, > then you should be trying to code it up and show it to people. I don't think it's going to get much simpler than OpenID. I did not know that that was already around the corner by the way. It's not the signup process that I'm worried about, it's what happens after that. The figures I showed in my talk speak volumes and I think most of us know them: more than 2/3 of all signups never continue to make any edits. That is the problem that this is contributing a tiny little bit to solving: easing people into becoming active contributors. [1] http://www.slideshare.net/mvexel/110527-osmaddwherecampeu [2] comes well recommended by the way, http://www.berlinfo.com/freetime/food-drink/bars/kreuzberg/index.htm. Already in OSM. -- Martijn van Exel http://about.me/mvexel _______________________________________________ talk mailing list [email protected] http://lists.openstreetmap.org/listinfo/talk
