The project's main differentiating factor seems to be not propagating the actual content to the nodes but instead save disk space by only distributing hashes of content.
However, having a "p2p" decentralized network of hashes doesn't solve the "discovery" problem. The blog lists the following bullet points of metadata but that's not enough to facilitate "content discovery":
>However it could be possible to build a scalable and fast decentralized infrastructure if instances only kept references to hosted content.
>Let’s define what could be the absolute minimum structure of decentralized content unit:
>- Reference to your content — a URL
>- User ID — A way to identify who posted the content (domain name)
>- Signature — A way to verify that the user is the actual owner
>- Content hash — A way to identify if content was changed after publishing
>- Post time — A way to know when the post was submitted to the platform
>It is not unreasonable to expect that all this information could fit into roughly 100 bytes.
Those minimal 5 fields of metadata (url+userid+sig+hash+time) are not enough to facilitate content discovery.
Content discovery of reducing the infinite internet down to a manageable subset requires a lot more metadata. That extra metadata requires scanning the actual content instead of the hashes. This extra metadata based on actual content (e.g. Google's "search index", Twitter's tweets & hashtags, etc) -- is one of the factors that acts as unescapable gravity pulling users towards centralization.
To the author, what algorithm did you have in mind for decentralized content discovery?