Believe me, I have been thinking this over for several months, even years by now. It’s only recently that I start speaking out about it. If mass edits take months to complete, they are actually regular mapping. Most of those bus stops have been added by yours truly, by the way. Like a fool I have been adding public_transport=platform/bus=yes, in addition to highway=bus_stop to them.
Hopefully it can serve as a wake up call that there is something fundamentally wrong with the public transport scheme as it is described at present on the wiki.
I don’t mind adding the detail. We describe the world in ever more detail and precision, that is just fine. When adding detail starts to mean that complexity increases, that’s where I disagree.
When each and every stop is represented as a node, we can add shelters as an area, platforms as ways or areas. stop_position nodes if you must. waste_baskets, benches, all you want.
But the basic premise should be and remain: map the stop on a node next to the highway/railway. That node contains details like name, ref, operator, network, route_ref, etc.
And that node is added once to the route relations (or several times if the vehicle serves it more than once on the variation represented by that route relation).
This works perfectly fine, without creating the need to add more than one object to the route relations. AND without the need for duplicating details across several objects.
Mapping of public transport should be within reach of all mappers. At the moment most say: it’s too complicated, leave it to the experts. Of course the ‘experts” don’t have enough time to mend millions of route relations around the world.
Converting from the old style to the “v2” is not a trivial exercise. Some mappers tag them as public_transport:version=2 just to make the validator shut up about them. They must not have the PT_Assistant plugin installed, as that one would make a lot of noise about them. The right thing to do is to put public_transport:version=1 on them. Or convert them, of course.
I’ll do one just to show what is involved (and how PT_Assistant can help with the task, even if it can’t do it all automatically, as those v1 relations are mostly useless except for giving a rough indication of which streets the buses use.
I created 2 screencasts with a demo of how PT_Assistant can be useful for such conversions.
Those Blacktown routes should be verified further though, that relation I started with was ‘almost’ v2. Usually a v1 has stops for both directions, this one didn’t.
Because there is this perception of some that mapping the stop on a node, with the details on that node exactly once, and having just these nodes in the route relations is somehow incomplete or even incorrect.
At present the wiki is interpreted as if the end goal is to have a platform way next to the road (for some regardless of the actual presence of such platform) and stop_position node on the highway as the only solution that is PT v2 compatible.
I can’t follow every edit happening on the wiki, but I did make sure that is still says that having a single public_transport=platform/highway=bus_stop node next to the highway and having just this node in the route relations IS fully PT v2 compliant. There is no need for futher “improvement”, when mapped that way.
Now, if there are actual platforms, it’s perfectly fine to map them as ways or areas, but they don’t need details like name, ref, etc. (Those are already present on the nodes that represent the stop). This doesn’t limit the ability to map stops in ever more detail, but doing so wouldn’t affect the route relations.
And I would like to see that we move to a solution where the end goal is to have a single object that represents a stop for each stop. Call it PT v3, if you like. The way it’s interpreted now from the wiki is too complex and hard to maintain.
There should be no need to ‘convert’ from that basic node that can be used for a simple bus stop to a way/area at any point in the lifetime of the bus stop. No need to update the route relations, or ‘augment’ them by adding stop_position nodes as well.
A single object for each and every bus or tram stop on a node next to the highway/railway. That’s what I’m saying.
Wonderful! I think I’d consider validating that data and add it to Wikidata while doing so. Would that be alright?
¡Genial! Estoy considerando de hacer el paso de la validación y al mismo tiempo agregar los datos a Wikidata. ¿Eso les parece buena idea?
User Weide is proposing something along those lines. For buses just the stops is still ambiguous though, so he wants to add nodes at crossings along the way, to give an indication of the route the bus follows.
I don’t like that too much, because it moves the burden of calculating and recalculating the
which ways belong to the routes to the renderers over and over again.
This means it won’t be possible to fetch itineraries in a convenient way.
It may be a bit more practical for mappers, but it’s a lot less useful for renderers.
There hasn’t. It would be very hard to accomplish technically and most mappers’ opinion is that timetables don’t belong in OpenStreetMap. Even if we would come up with a scheme, it would be really hard for contributors to keep those timetables up to date in OSM.
The version number isn’t all that important. I’m indeed trying to get everybody to map the same way, but I’d like to see it done in a scalable mapping scheme that is simple enough for anybody to understand, but expressive enough to go very deep in detail where needed/wanted.
Full disclosure, the HOT mailing list is one of the few that I’m absent from. I also think the iD editor is pretty good for beginner level mapping. Unfortunately one major thing it’s lacking is an easy tool for adding rectangular buildings that don’t end up as meaningless area=yes closedways.
My personal preference is to get people started on JOSM right away, but having tried that, I know it’s not easy either and mapper retention is definitely not higher.
So I was not reacting to the whole preceding discussion on that particular mailing list and I’m definitely not trying to attack Bryan personally in any way. I think he did an amazing job developing iD further.
Hi Komяpa, Bryan is already used to being upset by me. :-) I left him alone for many months now.
There actually are developers, but they are spread thin, as OSM is very broad.
-karlos- I don’t use iD either, but I am, occasionally, presented with the hubris in its wake.
There is indeed a function to square an area, if it’s not too irregular, at least. There is also an extensive tutorial explaining what’s the shortest way to developing RSI or a tennis elbow for people wanting to use it to draw more than a handful buildings. It is indeed annoying that one has to deselect and select again for ‘s’ to work, I’d also call that a bug (but not out loud, don’t want to upset Bryan TOO much either now).
Fixing that issue is not the solution to this problem though. A whole new map mode dedicated to drawing rectangular buildings is. I fail to understand why it has to take many years to actually develop this feature. I’m not much of a programmer myself though, so maybe I’m severely underestimating the required mathematics and analysis involved…
What is really needed is a new button next to point, line and area that says: building. The areas it draws would be rectangular right away, and tagged building=yes.
And this is not a problem that only regards HOT mapping, so unsubscribing from that mailing list won’t help much, if anything. The reason why the problem becomes apparent to HOT mappers is that there is an extra step involved in mapping for HOT projects, performed by validatorsj, a species endangered by extinction.
You know what would be very useful too? There are projects on tasks.hotosm.org in Bangladesh too, like this one:
You can find them by typing a search term. The advantage you have is that you would have a very good idea what you’re looking at, when on familiar terrain.
If you like, I can show you how to map more efficiently with JOSM by having a Google hangout
or Skype talk.
What do you think of this list?
Maybe you also want to add to it. I might try to program some of those myself, or I’ll use the easier ones to test candidates for next summer’s GSoC.
I think there is definitely still room for improvement. One issue is that it should be easier to verify what will be done when using the fix button. Visualisation, but that’s a hard nut to crack.
For the time being you can start to learn how to create models and create some. You might also want to document your “learning path”, so others can follow more easily. (Myself included, I have no idea how to get started with Blender for drawing models of buildings and structures either)
I realise it’s less attractive to do so, if the models aren’t readily accessible, but it’s good to know that that’s in the pipeline.
Google has a project where they pay a stipend to students to improve Free Software projects.
So our software is improved and Google is paying for it, so that’s nice of them.
You add the OBJ to the website and declare there to which OSM object it belongs. Pointing in the other direction comes later, if I understood correctly.
Don’t know the answer to question 3. It’s all fairly ‘cutting edge’ :-)
That’s odd. It might be related to having Expert Mode switched on.
In a previous diary entry n42k mention he used https://www.blender.org/ to convert the model of the Eiffel tower to Wavefront OBJ format.
I guess it would be possible to use it directly, but most of us probably need some tutorials first, before we can do anything useful.
Mateusz, is there a SPARQL query we can use ourselves in JOSM to download this for an area? I realise we would miss the descriptions in that case.
I must admit I didn’t follow this project closely, so maybe this is a dumb question…
Is there a tag on the OSM side that indicates a model is available for a feature on the map?
I’m not much of a programmer. Next summer is still rather far away, but if it’s possible to define what needs to be done, maybe we can formulate it as a GSoC task for 2018?
Or maybe it’s something for the EWG to tackle?