http://scienceblogs.com/commonknowledge/2010/06/open_data_an...
as well as the Panton Principles
In general, like with certain software tools and libraries a copy left license might not be the best choice, which is part of where John comes from.
Facts are not copyrightable, but the arrangement or presentation can be. The major case law on this is Feist vs. Rural Telephone. Wikipedia has a good writeup at the moment: http://en.wikipedia.org/wiki/Feist_Publications_v._Rural_Tel....
Copyright doesn't apply in the same way, educate yourself a bit before trying too hard; you may find out you're not going to be able to do what you want to do:
http://www.pddoc.com/copyright/compilation.htm
http://www.bitlaw.com/copyright/database.html
http://www.bitlaw.com/copyright/obtaining.html
http://en.wikipedia.org/wiki/United_States_copyright_law#Com...
Obviously we know that (unfortunately) that includes software, but data would be a stretch.
Also here: http://sciencecommons.org/old/databases/#dbcopyright The important quote is: "...whether the data itself is copyrightable, depends on what it is. To the extent it consists of factual information, it will not be copyrightable."
If it's not copyrighted, you don't need the GPL or anything similar.
How is it unfortunate that copyright covers software? Without copyright there would be no such thing as Free Software as it would be impossible to require people to distribute source with binaries or to distribute their changes.
Closed source could survive somewhat in a world without copyright on software since it doesn't come with source in the first place and can at least attempt to use DRM or other technological means to stop people from redistributing.
IANAL, but this seems to be at least close to what you are looking for. I do not know exactly what constitutes the difference between a "document" and a "data set" (outside of the physical format"), but I would think that the license could apply to any data set printed on paper, so it could also apply to one in digital format.