Duplicati looks very interesting, especially in conjunction with Amazon's new(er) Glacier long-term storage service that is relatively cheap. For my 2TB data set, it looks like I'd pay about $30/mo for Glacier, far more than I pay for CrashPlan, but still worth it. It would be awesome if Duplicati could somehow support the CrashPan back-end, but it being proprietary (CrashPlan I mean), that seems unlikely, hehe.
I'm also curious about the cross-platform nature of Duplicati. It says it's primarily programmed in C# and .NET. It also sounds like the dev reimplemented e.g. Rsync and some other things, but I get the impression the Duplicity Python back-end is still being used in at least some way (which would explain part of its cross platform capability). I suppose my question is how the GUI is handled on each platform. I ask because I'm inherently a bit wary of interpreted languages due to overhead and inefficiency, and both Python and Java that CrashPlan is programmed in have this potential issue. That being said it's a much easier way to get cross platform code...
Unfortunately I don't have an easy way to test Duplicati on my full backup data set, but I'm quite curious how it would compare to CrashPlan in terms of memory use. I'll see if I can get a chance to test it at some point.