I think the glaringly obvious thing for me is that (AFAIK), in order to have a true ‘darknet’ mode in Freenet, you need to be able to ‘trust’ another node. That is, you need to know exactly who you are communicating with to ensure that they aren’t law enforcement or another bad actor. If it works the way I understand it, then every leaf of a ‘trusted node tree’ is exposed to potential law enforcement or NSA hijacking.
SAFE doesn’t work that way. No trust is required; the system self heals and punishes bad node behaviour.
The second issue for me is the speed. I think that SAFE will be faster, and this is based purely on @dirvine 's assertions that we may see ‘faster than disk access’ comms. Freenet is woefully slow compared to the general internet (I speak from experience). We still have to see how SAFE goes on this issue, though.
A lesson we could learn from Freenet is the appearance of child pornography. I’m no advocate of censorship, and will argue against any kind of infrastructure-level limiting of the type of information that appears on the network. But the fact is, it’s actually quite difficult to access Freenet (at least through Frost), without coming across people asking about child porn and informing each other how to get it. It’s ok for me, I can ignore it. But I wouldn’t allow my children to use it; it’s not a SAFE environment. And even though you still need to follow their links to find the boards that contain child porn, kids may get curious, and then potentially fall prey to these people themselves. The future internet must be configurable (by parents) to be SAFE for kids.
I think the solution to this is built into SAFE, at least potentially. The app-focused approach should allow app devs to be able to screen out this sort of thing, effectively removing it from system-wide visibility. It’ll still be there, but it will be hard(er) for children to come across it by accident. We shouldn’t ‘police’ everything, but I think app devs will be able to mitigate this kind of thing quite easily.