Folks have you seen this project: https://infocentral.org/ One excerpt:

The future is app-free computing – fully-integrated, composeable, and adaptive software functionality that comes alongside neutral information rather than creating artificial boundaries.

Another:

InfoCentral is a next-generation internet engineering project and proposal. It combines Information-Centric Networking, persistent graph data models, declarative programming, and the best elements of the Semantic Web into a new software and internet architecture – one that is fundamentally decentralized and distribut_able_, while also easier to secure.

Hi @shalabh thank you for sharing this, I just became aware of via the newsletter by @Mariano Guerra . I find it always interesting how many different ways to seemingly similar topics exist. So in one way it encourages me that I see more and more people resonating with the vision of a decentralised information architecture. On the other hand it also seems a bit strange (at least to me) when a single person comes to great lengths and depths without collaboration (that is without active collaboration) from other people. I think there are many such attempts and many failed but. some succeeded to a certain degree. For me personally the most interesting effort with certainly a lot of traction already is web3. I hope to see lots of interesting experiments in that space. I would love to see some form of collaboration between infocentral and other FoC single person efforts in a more direct manner. It seems to me that everyone can think up the brightest future scenarios but oh god beware if contact to other human beings with different opinions is needed in the processes of creating something better for the future 😅 so I’m looking forward in engaging in discussion on how things could further proceed from here. There is so much room for improvement:)

Interesting @curious_reader - I knew about IPFS but was not following the greater initiative around web3.

I think That’s because there is no “central” web3 Initiative but only decentralised protocols. The key is interoperability and a strong focus on privacy/sovereignty.

Finally got around to look at this in detail. My conclusion is almost exactly what @curious_reader already wrote. This looks like someone has thought long and hard about organizing information systems, found all existing similar initiatives imperfect, and decided to do it right all on his own. With little chance of ever getting a viable implementation running, considering how much effort something like IPFS represents. My guess is that he could implement his ideas by adding a new codec to IPFS, and I'd even expect the IPFS community to be quite open to his ideas. On a more fundamental level, there is one aspect of this project that I am dubious about: the complete absence of named entities. I think there is a real need for concepts such as "the current state of Amazon's Web store", implemented in a predictable way. In his project, "Amazon" would just be an emerging concept constructed from references to an unnamed root entity. I don't believe that will work before I see it working.

@Konrad Hinsen Question about your skepticism regarding naming — not sure I understand you fully: doesn't IPNS fill the naming gap and if so, doesn't it work pretty much like DNS, which would be a naming system that works and we can see it working? In other words, do you think of "naming" as more than just a mapping of a human-readable token to a unique identifier, where the token is updated less often than the identifier?

@Stefan IPNS defines mutable references via a public/private key pair. The "name" of such a reference is the public key, and only the holder of the private key can change the value of the reference. So IPNS defines its own global name space, but the names are just as ugly as IPFS multihashes. There is also a mechanism to map standard domain names to IPNS public keys (I don't remember how it works exactly), but that's considered a temporary workaround, since relying on DNS is not compatible with the goals of the DWeb. My impression is that ENS (Ethereum Name Service) has become a more popular alternative for named entry points to IPFS data. The author of InfoCentral criticizes IPFS for its IPNS layer, which doesn't fit his idea of a complete absence of mutable references. What he doesn't seem to realized is that IPNS is an optional layer; he could build his system on the fully immutable basic layer and simply ignore IPNS.

Ok I should probably look into this stuff a bit more. My reluctance look into erethreum has mainly been about how I'm a bit skeptical about blockchain based stuff in general and "programmed/smart contracts", which I don't think are going to be that useful. I'm very interested in distributed apps, but words like "runs exactly as programmed" always leaves me a bit confused (it says on their home page). I mean, instead of "write smart contracts" how about just "write small, easily distributed apps, auto published on a decentralized network".

@shalabh I share your views on Ethereum, but it's a vast ecosystem by now and the originally somewhat fundamentalist idea of trust-free transactions (money and other "smart contracts") has been complemented by more pragmatic views and experiments. Ethereum Name Service is probably the most useful applications of blockchains I have seen so far. Filecoin is another interesting use of blockchains: a cryptocurrency used for paying for data storage, with verifiable storage replacing the "proof of work" of bitcoin. As for hash collisions, you never seem to find another reply than the theoretical estimations of their very low probability. I doubt anyone knows how likely they are in practice for real data, but IPFS is a real-life experiment to find out. I also wonder about the possibility of hash-collision-based DOS attacks, such a massively injecting random data into IPFS. For now I am looking at IPFS merely out of curiosity, so it's not a reason to worry. One advantage of IPFS is that its hash algorithm is not baked in. If SHA256 turns out to be insufficient, they can switch to something else without abandoning existing data.

One very appealing thing to me here is a global namespace pointing to stable objects (e.g. IPFS address points to a byte chunk) and other names (IPNS?) pointing to evolving objects. It gets more interesting when these things point to ensemble of objects (could do this with IPNS) and when those objects can be more than byte chunks (is this where erethreum comes in?). Because now we have a global, shared namespace of some stable and some evolving 'links' that very precisely point to a collection of data or computation or ensembles of both (~apps) and anyone can add their own objects to this worldwide space. The part where I'm not that optimistic about is things like program enforced contracts, including program enforced storage. First it seems like a costly way of doing storage (constantly proving that you indeed have the data). Second how do we trust these programs anyway? We have a hard time writing correct programs already - and these programs seem particularly complicated.

IPFS' lowest level is byte chunks, but it defines a few codecs for building data structures on top of that. The most flexible one is CBOR which is basically binary JSON. You can thus refer to arbitrarily complex data structures using a single IPFS hash, and store that under an IPNS key. Ethererum Name Service comes in for adding human-readable names, the blockchain making sure that everybody agrees on a single meaning for each name. As for your comments on Filecoin, I agree it seems expensive, but I don't have a better idea for decentralized and anonymous storage. But then, that's be basic issue with all blockchain applications: anonymous trust ends up being expensive.