I feel like I'm still confused about the underlying goal, so it's hard to recommend a solid "business strategy". A business strategy is only necessary if you want a business! If you need a "recompense strategy" it doesn't necessarily involve a *business*, so that's a key distinction. If all you want is not to pay out of pocket for things that cost real money, but don't need a "salary" as such (payment for your time input), then I think things get a lot simpler and maybe more realistic. If you actually want to create a business around this, even if only to cover "time and materials", then I think the only sensible way forward is a completely GPL-free commercial release and, yes, a well thought out business model.
I will proceed on the assumption that profit is not a motive and that cost of *materials* (not time) is what is sought, as that's essentially my understanding from previous CD discussions.
So, as far as I understand it there have been 2 specific costs discussed here and in the other thread.
1: Cost of development environment, Visual Studio. This is a fixed and specific cost, one that hopefully won't reoccur any time soon. As such it could easily be treated like the LCD donation was. Set a specific needed donation amount and ask for it until you get it. Steve did a great job of rallying people on that cause IMO and raised $150 in just a few days.
Also seen in that thread was evidence of a persistent desire to *avoid* the donation request (a perspective I know Mouser also tends to share), and that's fine. But it's not good to let that be the policy while behind the scenes it's causing major harm. I for one had no idea things were getting so bad until the announcement of CD2 and the new licensing scheme. I can't help thinking that if there had been more information shared, more direct and specific requests for donation, and more clear understanding of the costs involved, we could have avoided most of this.
2: Cost of hosting/bandwidth for CD downloads. It's pretty awesome to hear that CD downloads take up 1TB a month. That's incredible to me. Now it's true, that's a lot of bandwidth, and can be costly, but there are potential solutions, and frankly it seems like several have either not been fully investigated/considered, or not considered at all. This may be a lack of knowledge on my part about what you've actually done, so I hope I don't seem presumptuous, but for the sake of outlining some potentially new ideas I'll assume what you've stated publicly is the general limit of what you've explored.
Actually purchasing that amount of bandwidth monthly outright is not necessarily a bank breaker. VPS.net for example will give you 1TB for $50/mo with their CDN service. Add 1 VPS node for the server to tie it to (which gives you an additional 250GB of bandwidth) and your monthly cost is $70. Not cheap, but if you can manage to get $30 or $40/mo in donations, it could be very affordable. I realize it's a lot more than the 0 you may be paying now (using the DC server), but it does seem manageable, or at least a lot less risky to try than a total conversion of licensing terms, etc.
Then there's the obvious torrent idea that was mentioned. Many people do not want to deal with torrents, but it's a legitimate option for some and can definitely help take at least some of the load off. One way to make the torrent approach more effective and utilized too would be to take advantage of the separate component nature of CD and *only* offer the larger CD components via torrent, leaving the core package as a normal download.
I'm really not understanding why CNET, Brothersoft, and similar sites aren't an option as I've seen very large packages on all of them. The componentized nature of CD was, as I understood it, partly a reaction to the large bandwidth needs and a desire to unbundle optional stuff and reduce size of the core download. But if you don't pay for the bandwdith, why worry? Stick the full CD bundled package up on CNET, etc. and then keep the component versions available separately on the main site or other mirrors.
Something else that didn't seem to be mentioned was soliciting for additional mirrors. You can setup a simple round-robin type download system and get a bunch of smaller mirrors and spread the load out quite effectively. Many shared hosts come with surprising amounts of bandwidth these days and people can easily toss up a file and share their bandwidth with you. I'd be glad to do so myself, and could probably donate 100GB/mo of transfer comfortably. With a sophisticated enough download distribution script (which there are some available I believe), you could even limit each mirror to a specific amount of transfer (theoretical, based on number of requested downloads, but still). Get 10 people donating 100GB and you have your whole bandwidth for a month taken care of. There are tons of apps that have many or even just a few mirrors, very popular apps at that surely have similar bandwidth requirements. I would think that with the popularity of CD, you could find some people willing to donate bandwidth...
And why did SourceForge not work out? Is CD not licensed in the right way to comply with their terms?
Now I realize I may be displaying a good deal of ignorance as to the history and current status of CD as well as what has been looked into in trying to solve this problem. I have a lot of faith in Steve, so it's hard to imagine he hasn't fully considered all the above options. Yet I still am not understanding why some would not work, or at least help. I have no right to expect an explanation, heck I'm not even a CD user (but I have donated), but hopefully it's not presumptuous to ask for more details on what has been considered and the reasons why these options have been decided against in favor of a seemingly more complex and risky approach.