Talk:Richmond-on-Thames Mapping Party 2007

From OpenStreetMap Wiki
Jump to navigation Jump to search

Mapping of Kew Gardens

I think it's cool that you're going to map Kew Gardens. Unfortunately, I'm going to the Hack Day event on the same weekend (otherwise I'd be volunteering to help), but I'll try and follow your progress. It'd be cool if you could map the main pathways, lakes and glasshouses (maybe even the pagoda?).

If it's a success, I'll try and let my Kew Gardens colleagues know - I'm sure they'll be interested...

--Frankie Roberto 20:31, 12 June 2007 (BST) (who works for the Science Museum, an institution in the same sector).

Yeah that's one particularly pleasant slice of the cake, however there is an entry fee. I think it's a fiver to get in.
In exchange for your entry fee I think they give you a map. One of those stylised tourist guide maps showing the glasseshouses, and pagodo. Personally I was most impressed by the largest compost heap in the world! ....But don't look at the map. Burn it! It's copyrighted! -- Harry Wood 13:38, 13 June 2007 (BST)

it's a bit more than a five unfortunately - quite expensive actually. but good luck! Frankie Roberto 16:11, 16 June 2007 (BST)

Yahoo! Imagery - don't actually need to use GPS units

Given that this area has Yahoo! Aerial Imagery coverage, we don't actually need to use GPS units, except as a minor additional mapping aid.

The page states "Yahoo! imagery is available for this area so you can get started ahead of the mapping party by tracing all the visible water features, green spaces and land-use"

But in fact you should also be able to use the aerial imagery to sketch in road layouts (without knowing the names or exact topology). I say "should", because this does pre-suppose that the process of sketching over imagery still works reasonably well. Since we swapped from Java Applet to Potlatch I've not been able to do it properly.

Sketching in roads over the imagery, used to be facilitated by the Java Applet's support for unwayed segments, but with Potlatch the idea is to draw everything in ways. I haven't really mastered this approach, although I've not had much time to experiment with it lately. Talk:Unwayed segments has a discussion on this.

I would have been able to sketch in road layouts for the entire Richmond region, in three or four hours perhaps, were it not for the Java Applet being taken offline. ...and also you all seem to be enjoying your GPS toys. I wouldn't want to spoil your fun!

-- Harry Wood 14:02, 13 June 2007 (BST)

Can you tell me what it is that Potlatch doesn't do that the applet did? I mean, yes, obviously Potlatch creates ways whereas the applet created segments, but the only difference there is in the db - it's not something exposed to the user, they both appear as lines on the map.
It's not a modus operandi that I'd use myself (in fact, I generally think people tracing areas they've not visited is more harmful than useful, but as you say "I wouldn't want to spoil your fun"!) so am having difficulty grasping what it is that you want Potlatch to do.
Of course, as has been said several times, there's nothing stopping anyone who's pining for the applet from reuploading it and adding a user preference option "Use Potlatch/Java Applet on the edit tab". :) --Richard 11:10, 14 June 2007 (BST)
I'd find it quicker to use the java applet, and unwayed segments, because that's what I'm used to. Looks like Potlatch is also quick and easy to work with, but I'd have to experiment a little more. To give you some specifics, I've struggled a little with the bug which causes Potlatch to recenter itself over to the 'home' location. That's actually a bug in the rails code as I understand it. But mainly I've struggled because I've been trying to use JOSM (which I'm also not familiar with) in conjuction with Potlatch to convert some of my old unwayed segments. In a freshly unmapped region I might well find Potlatch quite nifty. I'll try it some time soon.
You wouldn't use the modus operandi of a two-phase mapping approach? Why ever not? This is something I dont understand, and it's this point which I'm raising here again. It's very quick and easy to 'sketch' over the yahoo imagery right? For the second phase you have a rough map already, and you just need to go around writing down street names. So why all this focus on use of GPS? It's really this point that I wanted to make, but I also appreciate that it's a headshift for people who are planning a weekend of GPS tracking fun. So it's a bit last minute to be raising this really.
Another software issue is that (I think) Mapnik and Tiles@home layers dont show untagged ways, so after you've done your first phase sketching in Potlatch, the ways dont show up, and areas still appear to be completely unmapped. If you were to tag all your new ways in Potlatch, e.g. set them all to highway=residential then they would show up, but actually you would have the opposite problem. The area would then start to look completely mapped, even though nobody has actually been there to gather street names etc. Perhaps it's this which prompts your "more harmful than useful" comment. This is another issue which was (accidentally) solved by the old unwayed segments approach (tiles@home shows them as feint lines). Perhaps the solution would be a new special 'sketch' tag which must also be supported by the renderers, to show sketched in regions more clearly.
Hope you don't find me rude. I'm trying to nudge people in the direction of something which seems like common sense to me. Making more effective use of the aerial imagery, where it is available. But I appreciate that discussing things is infinately less helpful than making actual code changes. I do understand what you've said about unwayed segments, and I actually like the idea of hiding the 'segment' concept from the user. -- Harry Wood 12:38, 14 June 2007 (BST)
Heh, no offence taken. :) Why don't I like the two-stage approach? Well, two reasons. One is that I don't think you can adequately map somewhere without having been there. Whenever I've tried to do it, I find that the assumptions I made in initially drawing the lines - whether from GPS or aerial imagery - are very often wrong. So in the second stage, I have to spend a load of extra time correcting my mistakes first time round.
The other reason is that the two-stage approach, to me, seems to encourage 'one person does the first stage, another does the second'. Quite often I've uploaded a GPS track, and before I can get round to fully mapping it, someone comes along and traces the line. Now I'm never going to complain about people putting extra work into OSM, but again, I find I have to take a whole load of time correcting their work. Sometimes they've drawn T-junctions as bends or vice versa, most often they've drawn really rough curves with far too few points on. (I have a bit of a bee in my bonnet about that... if it's a curve, you should draw it like it looks like a curve, even at high zoom levels. We're making maps, not just supplying topological data for routing systems.)
In terms of effort, it seems more efficient, to me, to do it in one go. But that's just my opinion.
Anyway, the main issue is how do we make Potlatch better for those who like the two-stage approach. First of all there's the home location thing. Yeah, I agree, it's annoying. Basically the edit tab URL isn't always updated when it should be, so Potlatch doesn't get a lat/lon and defaults to your home location instead. We are looking at it, but for now, the workaround I use is to get a view URL with lat and long in the address bar, then manually edit it to be 'edit.html' instead (with the lat, lon and zoom parameters the same). It's a horrid kludge, I know.
On the rendering, it sounds like t@h should render untagged ways in a similar way to unwayed segments. This seems eminently sensible to me. I'm not a t@h person but, again, it would be worth punting the suggestion towards those who are, either through trac or on the t@h mailing list.
cheers --Richard 17:01, 14 June 2007 (BST)

Ongoing work in the area

I am mapping Richmond itself slowly in spare evenings after work at a rate of maybe 1-2 hours per week, so all things being equal it makes no sense to assign cake slice E during this weekend unless you're running out of other things for people to do. In particular I may be surveying on the evening of the 15th so if someone is assigned area E on the morning of the 16th there's a good chance that there is survey data still being uploaded.

--Tialaramex 07:56, 14 June 2007 (BST)