[pnathan]

Nearly all of these relate to blockchain tech. Which is interesting. I do ask: why blockchain? Are modern blockchains still cycling away on silicon to find hashes? Seems a terrifically ineffecient way to build a system. I can understand the distributed hash table / transaction tree idea, but I struggle to grasp the rationale for proof of work systems.

[wmf]

Proof of work is the only known way to achieve consensus among entities that don't trust anyone. It also provides a way to "fairly" distribute new currency among entities who don't trust anyone. It's not clear to me that the decentralized Web needs consensus, its own currency, or total lack of trust.

[jpetso]

Which is why I like layered approaches like IPFS that separate the merkle-tree layer (hashed content) from the naming layer (IPNS, DNS, Namecoin, etc.).

There are definitely valid uses of blockchain tech but requiring it as central part of the architecture is a recipe for (re)centralization because eventually it will always grow beyond home-computer scale.

[pnathan]

> It's not clear to me that the decentralized Web needs consensus, its own currency, or total lack of trust.

I'm pretty sure I would say that the above is not how things work in human society. Limited trust, fiat currency, and regular disagreements is the norm.

[wyager]

There is no other known effective defense against Sybil attacks in fully decentralized systems.

[uudecode]

Sybil attacks do not work in small communities that members may choose to form where the members already know each other.

If a system forces all users to be part of some large, Borg-like, distributed hash table, or ledger, then by my definition it's not "fully decentralized".

[wyager]

Indeed, if you don't plan on writing distributed systems that work for more than a few people, you don't need to worry about Sybil attacks. However, the nice thing about the internet is that it connects billions of people, so here we are.

[uudecode]

I think there's a lot of historical evidence over the last few thousand years that people naturally form small communities, or at least small groups within large communities.

Today, people can, in theory, choose from among billions of peers to form these small groups. And the groups can if they so choose connect with each other, via a network of networks.

This internet "connects billions of people". True. But your company's LAN probably does not connect that many.

If a user started creating numerous fake identities on the LAN, then it's likely she would be detected.

Is it possible to create distributed "LANs" over the internet? (rhetorical question)

Another commenter questioned why a distributed Web needs "lack of trust".

People in small groups can and do trust each other. No computers are needed to make this happen.

[wyager]

Unfortunately, I'd like to engage in transactions (both economical and informational) with more than my immediate social group.

[uudecode]

Fortunately the two approaches are not mutually exclusive. There are no rules about how the "distributed Web" must be constructed. As the old saying goes, there's more than one way to do it.