On Fri, 08 Oct 2004 02:47:22 +0200, Anthere anthere9@yahoo.com wrote:

Some directions of thought may be added here (these are only flying thoughts) : http://meta.wikimedia.org/wiki/Newsletter_poll

Flying thoughts, but good ones. Thanks, ant :)

On Oct 8, 2004, at 6:30 AM, Stirling Newberry wrote:

Using Grass and a postgres SQL database, it would be possible to create an open mapping repository. Since a large fraction of all wiki entries are locations, this would allow maps to be made for most of them, perhaps even by a bot process eventually.

Is there any such project in process? Is there any interest if there isn't such a project?

See: http://meta.wikimedia.org/wiki/Maps http://meta.wikimedia.org/wiki/Wikipediatlas http://meta.wikimedia.org/wiki/Wikimaps http://meta.wikimedia.org/wiki/Map_generator

AFAIK no current activity.

-- brion vibber (brion @ pobox.com)

See also: http://mail.wikimedia.org/pipermail/wikipedia-l/2004-July/033704.html http://meta.wikimedia.org/wiki/GALILEO_Masters_2004

And the discussions surrounding the galileo-contest application. There are certainly many people interested in these ideas.

~sj

On Fri, 8 Oct 2004 08:19:01 -0700, Brion Vibber brion@pobox.com wrote:

On Oct 8, 2004, at 6:30 AM, Stirling Newberry wrote:

...

Using Grass and a postgres SQL database, it would be possible to create an open mapping repository. Since a large fraction of all wiki entries are locations, this would allow maps to be made for most of them, perhaps even by a bot process eventually.

Is there any such project in process? Is there any interest if there isn't such a project?

See: http://meta.wikimedia.org/wiki/Maps http://meta.wikimedia.org/wiki/Wikipediatlas http://meta.wikimedia.org/wiki/Wikimaps http://meta.wikimedia.org/wiki/Map_generator

AFAIK no current activity.

-- brion vibber (brion @ pobox.com)

---

Wikipedia-l mailing list Wikipedia-l@Wikimedia.org http://mail.wikipedia.org/mailman/listinfo/wikipedia-l

On Oct 8, 2004, at 4:49 PM, Delirium wrote:

Stirling Newberry wrote:

...

Using Grass and a postgres SQL database, it would be possible to create an open mapping repository. Since a large fraction of all wiki entries are locations, this would allow maps to be made for most of them, perhaps even by a bot process eventually.

Is there any such project in process? Is there any interest if there isn't such a project?

The last time I looked at it, the maps Grass produced looked _really_ bad, along the lines of the quality of maps I could view on my Apple ][ in the mid-1980s. Simple color schemes, no shading, lots of aliasing, etc. There are a few such auto-generated maps in Wikipedia, and they generally look pretty bad too---much worse than the maps we have from other sources, including some Wikipedian-created maps.

Has this gotten any better? Ideally I'd love such a solution, but it'd have to produce results that looked reasonably good.

-Mark

The quality of inputs equals the quality of outputs - grass is perfectly capable of serving good maps from material that is good quality. It isn't the tool to produce many of those rasterizations, but there are other tools that can, which don't need to be used in wikipedia. Nor is it necessary to be committed to one GIS server.

Daniel Mayer wrote:

Right now I'm taking an Internet-based GIS class.

You probably know more about this than I do, so this is quite possibly a naive question. Is it possible to leverage any of the existing internet-based GIS stuff for our purposes? The Open GIS Consortium has a number of protocols that seem to do things similar to what we want, with software already built to do it --- the most relevant seems to be the Web Map Server (WMS), which worked pretty nicely the last time I played around with one. Basically generates raster-image views of an internal GIS-based map. There's also a Web Feature Server (WFS) if you want feature-type information rather than rasterized renderings to be served.

-Mark

Daniel Mayer wrote:

PS - the whole idea of copyrights on spatial data seem to be absurd to me; how could anybody claim to *own* any exact digital reproduction of the street network for Paris or the exact location and shape of the Zaire River? And yet this type of thing is *very* common in the GIS world (but I also think it is absurd for people to copyright *natural* DNA sequences and patent their *natural* products ; I guess I'm weird that way). Argh!

While I basically agree with both notions, I can see that someone investing money in getting to Zaire and taking precise measurements, or in sequencing some DNA and determining what it does (broadly speaking), should get something in return. If the question is no information vs. information to pay for, the latter is preferable, IMHO. But, patents are likely the wrong way, and Wikipedians traveling to Zaire and taking down coordinates would certainly be best :-)

Magnus

Ray Saintonge wrote:

It seems to me that before we get people into the Zaire jungles taking measurements we should probably take full advantage of available aerial photographs. :-)

No way! We need to get the coordinates down to the micron! Otherwise, imagine the NPOV disputes when it comes to national borders ("That hill is *OURS*!!") :-)

Magnus

On Mon, Oct 11, 2004 at 10:47:50AM -0700, Daniel Mayer wrote:

--- Magnus Manske magnus.manske@web.de wrote:

...

While I basically agree with both notions, I can see that someone investing money in getting to Zaire and taking precise measurements, or in sequencing some DNA and determining what it does (broadly speaking), should get something in return. If the question is no information vs. information to pay for, the latter is preferable, IMHO. But, patents are likely the wrong way, and Wikipedians traveling to Zaire and taking down coordinates would certainly be best :-)

It also takes effort and money to obtain a photograph of a painting whose copyright is in the public domain. Yet the resulting photograph is not protected by copyright.

Creating a map is a creative work while making an almost exact copy of a picure is not. That is what matters regarding copyright.

(I don't know how much work is needed to create a map - probably it's not that scary nowadays, being able to trace aerial photos - but I guess it's many magnitude more work than going to British Museum and taking a photo. I wouldn't debate that it contains creative work, and lots of it.)

my 2 'cents, grin

On Oct 12, 2004, at 11:37 AM, Andre Engels wrote:

However, the creative work in creating a map is in the search of map symbols - the colour and thickness of the lines, the precision in which they are drawn, etcetera. The spatial data on which the map is based are not part of the 'creative' work on the map.

Andre Engels

Grin wrote:

Collecting and creating spatial data is indeed creative work, however, most of the spatial data needed are not copyright - but are generated by governments (USGS quads for example) and can be turned into maps. This is a fundamental wiki process - and indeed a fundamental open source process: isolate the creative component which "protects" information, that is, withholds it from the public. Duplicate that component, and include the public domain information which has been "fenced in" by the small amount of copyrightable material.

In the case of mapping, that's conerting PD spatial data where it exists into visual maps. Where there is no PD spatial data, then we have to do more. But much of what people want can be accomplished by the above mechanism. We should not attempt to appropriate copyright material on the argument that it isn't very copyright, on the other hand, we should be very aggressive in making sure that every particle of what is in the intellectual commons is accessible.

Thanks for the pointers people, I will draft a proposal to the people working on the problem.

On Oct 12, 2004, at 12:14 PM, Daniel Mayer wrote:

--- Stirling Newberry stirling.newberry@xigenics.net wrote:

...

Collecting and creating spatial data is indeed creative work,

Making a digital outline of a river or road from an aerial photo is slavish, not creative work. The only creativity possible is in making mistakes but the Supremes have ruled that that cannot be used to justify a copyright defense. The problem is that this concept has not be tested yet for GIS data.

-- mav

Translating from one medium to another is defined as creativity. Just as making a translation is copyrightable (though derrivative of the original work). However, taking a PD aerial photograph and taking a program to reduce it down to a line and releasing it under GFDL puts the material in the intellectual commons. The change is that in 1980 you had to hire a person to do the work, and companies successfully argued that if they could not make enough to pay that person, it would not be done.

We shouldn't be trying to rationalize skirting current law. We should be working very hard to make sure that everything that is available under current law is accessible and usuable. This is my last online comment on the IP issues of mapping, if someone has a comment, objection or question about IP issues involved here, email me privately and I'll be happy to discuss it - I just don't want to be taking up community bandwidth on the issue.

Daniel Mayer wrote:

Translating is one thing, literally tracing is very different. Tracing does not require any creativity at all. But as I said, the particulars in regard to GIS data have not been tested yet (AFAIK). The Court may find there is a compelling interest in giving that type of GIS data copyright protection. Thus we must play it safe and create our own whenever PD data is not available.

Well, I don't know much about GIS data, or mapping, and I'll prove that through my following proposal :-)

Looking at a map, I see three possible structures: * Points (coordinates and altitde), like in mountain tops, cities (their center, at least), etc. * Open polygons ("polyline", made from points as described above), like rivers * Closed polygons, for regions, states, counties, lakes, continents, etc.

A world map can be described by everything belonging to another object. "Lake Superior" would thus belong to the next larger units, namely Ontario, Minnesota, Wisconsin, and Michigan. Ontario belongs to Canada, whereas the others belong to the US. Both Canada and the US belong to the continent America, which belongs to the world.

The outer shape of, say, Wisconsin, can now be calculated by merging all objects that belong to Wisconsin. Likewise, the outer shape of the NATO can be calculated by merging all countries that belong to the NATO. Drawing a map of a region with given coordinates can thus easily exclude irrelevant parts by going down from world view; if Europe is not visible on the given part of the map, it can be excluded from drawing as a whole, speeding up map drawing. Likewise, one could just require to show a bunch of objects ("Lake Superior", "New York", "Ottawa"), and a rectangular view of the world map is selected that includes all these objects; that map section is then piped into the process above.

But how to create such a complex structure? After all, an object *might* consist of a bunch of multiple polygons (can't think of an example right now); it probably belongs to several objects; it might be of different "types" (for example, "Hamburg, Germany" is both a county and a city).

I believe the best way to do this is (surprise!) through a wiki. *But*, one that will, upon saving, analyze (parse) a page and put the contents into a separate database, which is then used to draw the maps and do other things (like click on a map, and it takes you to the object you clicked on, thus a reverse-lookup). This has the advantage to be open to practical structures. For example, it might be a good idea to list most of the cities on the same page as the county they're in, instead of creating thousands of one-line "darticles" (from "data-article"). Exceptions can, of course, be made, like "Kansas City", which belongs to both Kansas and Missouri. The database would remember where that data came from, so one can always find the source "darticle". [Note that this is a little different from the wikidata concept, as there is no from to enter your data, rather a markup which then gets parsed.]

As there are various types of data to enter, I propose a HTML/XML-like structure (which would make parsing a breeze once the XML-parser is up and running :-): <mapdata> <name>Lake Superior</name> <type magnitude=5>lake</type> <polygon>x,y,a;x,y,a;x,y,a;x,y,a;x,y,a</polygon> <belongsto>Ontario</belongsto> <belongsto>Minnesota</belongsto> <belongsto>Wisconsin</belongsto> <belongsto>Canada</belongsto> </mapdata> with x,y=coordinates, a=altitude, magnitude between 1-5, where 1 is smallest and 5 is largest object category of that type

This schema could also work for multiple languages: <name language="de">...</name> For keeping data consistent, I propose that we require one <name> without language attribute (or language="en"), which is both the English name and the identifier in the database. Alternate names could then be used in map drawing in different wikipedias.

For those of you concerned about Even More HTML, and that being to complicated for newbies: I strongly believe that anyone who can take down GPS coordinates or calculate them from a map, and go to wikipedia to enter them, should be able to make at least the basic entry in that manner when shown a good example first.

The final piece missing, the actual "drawing syntax" in the actual wikipedia article, is another thing entirely, and should be discussed separately IMHO. However, as the above schema uses labels for objects, it would easily be possible to implement something like "show the US and Canada; 300x300 pixel; outline the Greate Lakes in red, 3px pen; show the US in brown; show Canada in green; show all other countries in pink; show country borders; show rivers of magnitude 4 and 5; show mountains of magnitude 4 and 5".

But I'm sure the mapping pros would do something entirely different.

Magnus