Rationale for the amendment
I believe this is a better option to be voting on now than Raul's proposals as it makes the forthcoming ballot a simple decision on whether to keep non-free or not; without either requiring any later decisions to be made whatever the outcome here, nor tying the outcome on that question with other independent changes that people may or may not support.
I believe keeping non-free continues to be useful in a number of ways.
First, it allows us to provide useful packages that we could not otherwise provide. In the future, we can probably expect non-free to be the only way we can distribute such things as the GNU Manuals and Centrino wireless drivers. In the past it's allowed us to distribute such things as Qmail, Qt, netscape, ncftp and Java.
Second, it allows us to ensure that our operating system works well with popular pieces of non-free software, and vice-versa; software that's in non-free can be maintained with all the usual tools we have for the main Debian distribution: dependency analysis, autobuilders, even security support. Without that software in the Debian archive, it becomes significantly more difficult for developers of free software to reproduce reports of bugs in their packages that show up when used with particular non-free applications.
Third, it allows us to establish productive relationships with upstream authors of non-free software, which gives the free software community an effective channel for communicating their needs and desires. In some cases this results in the package being relicensed under a free license, in others it gives us the opportunity to get increased compatability across apps, in others it simply makes it clear that the appropriate arguments have been presented to upstream but won't be accepted.
While none of those arguments can be proven in any sense, they all have supporting evidence — the fact that people bother to maintain it and install it is evidence that some non-free software continues to be useful, there are numerous examples of packages in non-free being relicensed to be free (and far more examples of that than the converse, despite there being far more packages in main), numerous non-free packages are better integrated into Debian by us than they are by upstream, and so forth.
By contrast there is not, to the best of my knowledge, any evidence at all to support the claims that supporting non-free costs as anything notable. The costs in resources of non-free are trivial: the amount of diskspace and bandwidth it uses up in its entirety are less than the number of uploads we get to main most days. The costs in manpower are also fairly small: all the ongoing support is a freebie from supporting software in main (and contrib); and the setup support is (by my estimation) trivial, and having already been spent isn't able to be recovered. Similarly, there have been claims that without non-free, we'll have a bigger incentive to encourage people to relicense their software freely: that if they don't, we won't distribute it, but those claims haven't been supported by any evidence at all, anecdotal or otherwise.
On the other hand there are reasonably measurable potential costs to removing non-free. It's been speculated that if Debian's infrastructure really is valuable for non-free software, then if we stop providing it, someone else will set it up and maintain it. This is, IMO, a significant cost: maintaining the Debian infrastructure isn't trivial, even if you exclude all the development: ensuring the servers you run are secure, making sure the appropriate people have upload access, avoiding spam, and making sure the system keeps working all take time, and the only people that seem likely to invest that time seem to be people who would have otherwise invested it in doing things beneficial to Debian — either doing a better job maintaining non-free software for Debian users, or more likely working on free software or Debian infrastructure that's useful for free software users.
Finally, I believe that making a resolution to keep non-free is a better outcome than further discussion so that we can ensure that it's clear to all our members and all our users what our intentions are on this controversial issue for the foreseeable future.
Manoj Srivastava