> Pubky's idea seems to be that centralization content distribution is unavoidable, so they aren't even trying. The idea of Nostr is that such thing isn't unavoidable, so we are trying.
This is a fair assessment, though some nuances are worth highlighting.
Firstly, indexer centralization primarily becomes necessary in Pubky if your application requires a comprehensive, network-wide view of all homeservers—this is in fact something essential for pubky.app’s social functionalities. Features like search, semantic social graph inferences, and others inherently demand centralization due to the resource intensity of crawling the entire Pubky ecosystem, much like Google indexing the internet. I'm not uptodate on Nostr developments, but I believe it might face similar challenges in this regard, although I may stand corrected.
Importantly, an indexer in Pubky doesn’t necessarily need to handle content distribution; it only needs to guide users to content locations. The verification of content provenance still happens at the homeserver level. Indexers cloning data and serving directly, however, can enhance user experience by improving responsiveness, and I anticipate the emergence of both lightweight and full-content indexers. We are building Pubky Nexus, a full-content indexer, but it can be strip down to become lightweight as well.
We envision multiple competing indexers evolving, akin to the variety seen in web search engines today, despite Google’s dominance. While fully decentralized content distribution may have limitations, I envision (and want to dedicate effort to it) niche users with sufficient resources and interests could potentially run their own indexers, though they naturally only index a partial view of the network. For what is worth, I would like to run one at home.