On Mon, 14 Feb 2000 13:03:38 -0500, Raul Miller <[EMAIL PROTECTED]> wrote:
> > > Are you now claiming that it's legal to distribute kghostscript? > > On Mon, Feb 14, 2000 at 03:21:44PM +0100, Marc van Leeuwen wrote: > Yes, > definitely, if you are distributing sources; from your remarks I > conclude > that even you would agree to this. One could include scripts to > compile > and link (say statically, just to get the worst case) those > sources onto a > complete executable; since all the tools necessary to do > that could also > be (and are in fact) legally distributed by the same > distributor, the > effect is that this executable will be identical, bit by > bit, to an > executable that the distributor could assemble. So the exact > same effect > is achieved, although slightly less efficiently, as if the > distributor had > directly distributed that executable file. But, we agree, > the latter would > certainly not be permitted. > > If the exact same affect is achieved -- if no user intervention is required > -- then there's no legal protection gained by automatically building the > executables for the customer on their machine. You'd still be distributing > executables, you'd just be using a different technology to deliver them. > > [...] There's a difference between simply distributing GPLed sources and > distributing the sources as a part of a system which automatically builds > working executables. In the latter case it's pretty obvious that you are > distributing executables. In fact I wasn't necessarily thinking about generating the executables in a 100% automated way (although that might be a possibility), just about ensuring that the executables will be 100% identical to what the distributor has. But your remark reveals an interesting line of thought, one that would never have occurred to me. It considers any inconvenience, caused to the recipients by having to distribute sources, not as an inevitable by-product of having to abide by the conditions of the licence, but rather as a condition that in and by itself is necessary for the legality of the distribution. This raises the further question as to how much inconvenience is necessary for that legality. Obviously having to wait a couple of minutes for gcc to complete doing its thing is not sufficient. Why, the user might even be enjoying a cappucino at the same time! Would it suffice to have to click on a button labelled "Compile and install the program" (some consider something similar sufficient to enter into a legally binding contract, so why not)? Nah, too easy! Although you would have to come dangerously close to your computer, you could manage to do that without even putting down your cappucino. To have to type "Y" to a prompt "Do you want to install the program [y/N]? ". Same difference! Having to type "make"? Here I would personally put down my cup, but I guess with some practice, it can still be done without; it's still too easy! Having to read through a README file to find that the proper incantation is to type: "I would like to compile and install this program now, please." (exact spelling and punctuation required, the README could be in pdf to prevent cheating by cut-and-paste)? That ought to do it: even if you do manage to do that while holding on to your cup, it would be cold by the time you finished. Of course I'm being silly here. The real measure of inconvenience has nothing to do with coffee. The inconvenience just should be great enough to allow the conclusion "therefore there is just no way that we can include KDE in Debian". As long as conclusions are being used to find the arguments, no wonder people can't ever seem to agree on those arguments. > > Now apart from distributing sources and distributing a statically linked > > executable, there are other methods to give users access to an executable, > > varying in the amount of work do be done at the recipients end. Offhand I > > can think of distributing compiled but unlinked object files, or > > distributing dynamically linked object files; maybe other possibilities > > exist as well. > > Please take a look at > http://www.debian.org/Lists-Archives/debian-legal-0002/msg00211.html for a > historical example of shipping unlinked object files. I had almost cited that case (of NeXT and Objective C/gcc) in my original message to show how opinions can vary in these matters, but had decided not to do so because that case is dual to KDE/Qt (non-GPL on top of GPL rather than GPL on top of non-GPL), and therefore does not quite match the example. Marc van Leeuwen Universite de Poitiers http://wwwmathlabo.univ-poitiers.fr/~maavl/
