The GPL version 2 was written back in 1991, in some sort of "plain english". At least the intention was to write a clear document that allows recipients of a copyrighted work (e.g. a compute program in source code form) the four freedoms,

  • 0: The reedom to use, for any purpose;
  • 1: The Freedom to study the program;
  • 2: The Freedom to make modifications to the program;
  • 3: The Freedom to distribute the program, either in modified form or verbatim, either as source or as a compiled object./li>

(This is not the canonical form of the four freedoms, heck no). There are restrictions on when you may exercise those freedoms. In particular, when you distribute the program, you need to give the recipient the source code. If we boil it down to its syrupy goodness, this becomes "you can have this to do what you like, but anyone you give this to gets that same right."

Well, that's the intention. And under normal use, this is how it works. The GPL gives you permission to use the software (you must have a license to even run a piece of software you have, because of the way copyright law interacts with software). If you violate the terms of the GPL, then you can't use the software. Simple.

The GPL version 2 has some extra text outside of the legal parts; for instance, one bit tries to clarify the intention of the license:

This General Public License does not permit incorporating your program into proprietary programs. If your program is a subroutine library, you may consider it more useful to permit linking proprietary applications with the library. If this is what you want to do, use the GNU Library General Public License instead of this License.

However, intentions come into play to only a limited extent in licenses. There is the text of the license, which is .. well, suffice to say it was written in 1991 with plain English in mind.

The Register is reporting on a webcast hosted by Black Duck Software with Karen Copenhaver and Mark Radcliffe. The Register article starts out with the misleading paragraph:

Two prominent IP lawyers have warned that the all-pervasive General Public License version 2 (GPLv2) is legally unsound.

Unsound doesn't mean broken, and unsound doesn't mean that the main use of the GPL version 2 is unsound. There's a great deal of ambiguity in the license; I saw a talk by Sean Hogle at OSiMWorld with similar points. In particular this ambiguity exists around "derivative work", although "distribution" is also not watertight. One illustration that "distribution" doesn't cover everything that might be intended is the existence of the Affero GPL (AGPL).

Note that the analysis presented (in the webcast and summarized on the Register and then summary-summarized here) applies to the GPL version 2 only, and the GPLv3 is a great deal clearer (from a legal point of view, although it's a lot more words).

As far as the Register article goes, the first comment finishes with "Rocket science it is not." No, it's not rocket science, but the gap between what you want (or what you have been led to believe) and what the text actually says -- let alone what it does when subjected to scrutiny -- may be very great. And that's the different between landing on the moon, crashing into the moon, and exploding on the launch pad (which is AGPLv3, BTW).