Tschirl
Here we go.
email posting
edit[...]
Hi, we are working on it, but there is sort of not much coming back. It seems like things have to resolved within Wiki and if it means to reinvent the wheel, well lets reinvent the wheel.
[...]
It should be as easy to enter new data into a Wikimap.
BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
For now, I'll just reply to everyone, creating huge amounts of mail traffic :-)
Following is my case for my own wikimaps demo implementation.
As far as I know, we all agree on one point: using a wiki as the (meta-)data source. -> That is just one part of the story. What is a lot more important is that people can see a map. You will need some kind of interface.
That means using a syntax that is both human- and machine-readable. Which pretty much disqualifies the existing GIS standards. -> Thats the point. Wikipedians usually do not edit texts using SQL, they use this interface. You will need an interface to build maps.
My demo at [1] introduces such a format, which uses wiki text alone for all map data. Shapes are pased on poly-lines, which can be grouped together as borders to form a polygon. By using a single polyline for multiple objects (like two objects bordering, or two objects with one "including" the other like a country includes a state), mismatching borders can be avoided. -> What about enclaves and excalves? What about complex objects. What happens when you delete one object that shares a border with another line.
Storing raw data as wiki text seems insane at first glance, but:
- Inserting is easy (paste)
- Editing is easy
- No need to write additional software for data management
- Data and metadata coexist in the same system, maybe even on the same page
-> It is not at all insane. But is is insane to implement the format again. It has been deone 1000 times already. The people who invented those 1000 formats one day gathered and decided to just do it in a standard way. I see no reason whatsoever (except insanity) why Wikipedia should not store its geometries in the standard format. There must be an algorithm which parses the coordinate and stores it in a geometry field. Best possible database currently is PostgreSQL with the PostGIS langext.
And, IMHO, speed of data access is *not* an issue here. Maps would not be generated on-the-fly, but rather cached as vector/bitmap files and updated as the need arises. -> Speed of geographic data is an issue. You will not want to keep people zoom in and out in intervals, this does not interoperate with any other geodata system.
If this turns out to be a problem, the raw data can be copied into a "real" database upon saving. By making the wiki text machine-readble one can switch to an alternate data model at any time. -> This is sort of developing into the right direction.
The syntax I use is currently limited, but can be extended with new keywords. This also means that it is possible to start with "plain maps" right now, and add things like points of interest, historical maps, etc. later.
For map display, I currently have a SVG generator. -> Please do not use SVG. It is closed source. Additionally it is not very useful once you start to create more data and want to srve larger chunks of maps. Use standardized software which has been developed by people wo have done nothing but this for years.
I couldn't get imagemagick to work at my web hosting service, so you'll need a plugin or Konqueror to view the maps "on site". Other rendering engines are, of course, possible. -> I did not understand this one. We are talking about Wikimaps not about whether this has to be installed by anybody. This will be a fairly large and powerful infrastructure and not many would want to duplicate it.
offline comment
edit[...]
Storing point data directly in a wiki page, making it completely unshareable with anyone, not able to use any exisiting gis data? Argh! I don't have the time, but someone needs to explain the joys of interoperability. But really, more than that, we just need a really nice demo. I doubt we'll make progress with him, by telling about how much better specifications and interoperability and all is, since he is convinced that he has a smart way to do it. The key will be an impressive demo. It should link to wiki pages by clicking on it, and should be able to zoom in, with finer levels of detail, and ideally more wiki links. Once we get that together we should talk to him again. It'll be much more effective than the unfortunate normal ogc arguments of just saying how much better it could be. But do point out that others can then access the work that goes into wikipedia maps through other interfaces, and how they can pick up other's changes.