I think checking line length for the license text block would be much more discriminative: approx 9% of the 317 currently matching packages would trigger the test (some more than once). I attempted to simulate this change in Lintian with a totally separate hacky script, manually checking over the matches from *my* script produced the following bad packages (including the 1/9 I found by-hand in my random-sampling)
audacity bibletime gnome-devel-docs kbounce ksirk ktuberling marble mrpt ogre opensesame openscad-mcad (including: "it is licensed under the Creative Commons - GNU LGPL 2.1 license.") phabricator pioneers plm python-odf recommonmark starfighter ("CC-BY-SA-3.0+ Identical to CC-BY-SA-3.0, except you may also chose to use any later version of this license.") supercollider ublock-origin xia My hacky script had 3-4 false positives too, but manual inspection of them show that to be due to problems in my hacky script, and not in the notion of line-length checking the license. Special handling of CC0 is indeed necessary for the line length check. I've seen it referred to as "CC-Zero" (haskell-network-multicast) as well as the more common CC0. So I support adjusting the regex to avoid CC0/Zero and doing a simpler line length check, 20 seems reasonable. What I haven't done is look at the other 297 packages currently matching and determine whether they are also incorrect. But let's assume not. -- Jonathan Dowland