[Foundation-l] Problems with the new license TOS

[Tisza Gergő]

I found a few apparent legal problems while translating the license
update documents. Apologies if these have already been discussed to
death - I didn't follow earlier debates, and the archives are mostly
useless as a knowledge base.

== revision not specified ==

The TOS says that reusers have to attribute the authors by linking to
the article. The problem is that such a link will actually point to a
different article after each edit (that is, the text and author list
will have been changed). If you find a text copied from Wikipedia on
the net, and there is no date information, it is very hard to find out
which version of the article it is (and thus who the authors are); if
the text is a derivative work from a Wikipedia article, then it's
practically impossible.

Even if one argues that attributing bogus authors is not a problem as
long as the real ones all appear on the list, the author list can
change arbitrarily when the article is renamed or deleted and
rewritten. (Neither of which is apparent even if one looks at the page
history.)

A few possible solutions to that:
- require reusers to permalink to the revision they used; change the
totally unhelpful error message that is shown when one follows a link
to a deleted version. (Probably not a very good idea as it messes up
caching. Also, bad usability: most of the people who click such a link
don't care about authors and original version one bit, and just want
to see/edit the current version of the article.)
- develop some syntax that shows the current version of the article,
but with a little message on top saying "you have followed a link from
a page reusing an older version of this article. You can see the most
recent version of the article; if you want to see the original click
here." (Maybe through some fragment id trick and javascript so it can
go through the cache?) We would still have to address links to deleted
versions.
- require reusers to give date/revision of the page along with the
url. Make some sort of search interface to find the text and/or author
set of an article based on that information.

== CC version incompatibilities ==

Copyright policy now says "You may import any text from other sources
that is available under the CC-BY-SA license", which is incorrect for
to reasons. First, CC-BY-SA-1.0 (used, for example, by Wikitravel) is
not compatible with anything but itself (as they forgot to include the
("or any later version" part). Second, different versions and
jurisdictions of CC are not quite compatible: for example if a wiki
has an article under CC-BY-SA-3.0-US, then uploading that to Wikipedia
(which will use CC-BY-SA-3.0 unported) is actually a breach of the
license. You could change the version or jurisdiction when you create
an adaptation (that is, you make changes significant enough to be
considered on of the authors), but not when you just redistribute the
work. (I doubt anything could be done about this beyond prodding CC to
release a saner version of their license soon.)

== edit summary cannot contain links ==

The currently proposed editing policy says:

"If you import text under the CC-BY-SA license, you must abide by the
terms of the license; specifically, you must, in a reasonable fashion,
credit the author(s). Where such credit is commonly given through page
histories (such as wiki-to-wiki copying), it is sufficient to give
attribution in the edit summary, which is recorded in the page
history, when importing the content."

(which BTW should be rephrased more clearly - does it mean you can use
the edit summary if you import text from another wiki, but not when
you do it from any other web page?)
The problem is that the edit summary does not allow external links:
they will show as plain text, and it would be hard to argue that that
is reasonable to the medium. (This one is easy to fix: allow them, and
rely on rev_delete and capctha to stop edit summary spam instead.)
Furthermore, a long link does not necessarily fit into the summary
(which is 255 bytes long, and there are a number of web pages that use
ugly links with loads GET parameters that are longer than that), so
some sort of separate attribution log might be more reasonable.