Non-realtime publishing for censorship resistance

One part that the author doesn't go into in detail is that low-latency interaction makes it much harder to be location-anonymous when an adversary can experiment on your service. Of the three examples given in the article, only Silk Road has tried to be location-anonymous by being exclusively available with Tor onion services. But for Silk Road and several other dark markets, it didn't work out because investigators were eventually able to locate the infrastructure.

I haven't looked into how that happened in detail, and I think it was quite varied in practice in each case, but there are known methods to take advantage of the low-latency thing to experiment on services and draw conclusions about where their infrastructure is located. One example is to try to attack different parts of the Internet (different links or different colocation facilities, for example, maybe by overloading them with a DDoS or trying to partition them from the Internet with a BGP routing attack) and see which ones cause the service's availability to suffer when they're under attack. When a service is available in real time, this kind of experiment is much more practical because it's possible to see directly whether each individual attack has had an impact.

I think Freenet has something along the lines of the author's zine concept, where people can publish signed pointers to the latest revision of a document or set of documents.

I'm having a tough time following the logic exactly. Can someone ELI5? I would think the question of centralized vs decentralized would be the dispositive issue. I think I'm missing something basic.

Wouldn't ipns and ipfs be enough to achieve most of this?