nostr:nprofile1qy2hwumn8ghj7un9d3shjtnddaehgu3wwp6kyqpqcu4r6pt94plck34vtw9t3zj974k0vl0z8x9gew5vdp2ukw3ekr4qct2mzm At least half of the complaints I see people make where specific services are mentione...
@nprofile1q... There is no good option besides rate-limiting unauthenticated accesses for interactions that cannot be backed solely by content-addressed mechanisms.
The tarball would be a single linear read, that's probably the single least expensive interaction those misbehaving scrapers have with the service.
Wikipedia could be hosted on IPFS (or other content-addressed networks) without any issue. Update messages & transactions require authentication already (at least if accessed over Tor) and can be rate-limited if needed. Replication of successful transactions to content-addressed backing is easy.
What we are seeing is long-coming death of server-centric networking as the dominant paradigm. The c10k & c1M problems were part warnings. Past a point it doesn't work without absurd tradeoffs (such as ridiculous overprovisioning and power consumption).
Offline-first clients & user-agents should be a priority for either server-centric or content-addressed data access, because it enables for mitigating the downsides of either the best. Low-latency high-bandwidth unauthenticated server-centric access is simply not going to be a thing anymore because it is untenable. Sure, in the meantime we'll see very stupid doxing-based schemes but that won't work.
On a non-technical, social level, the economic incentives & properties that even allow the commercial-focus misbehaving scrapers to operate as they do must be destroyed. Their misbehavior isn't free and they have to fund it.
The tarball would be a single linear read, that's probably the single least expensive interaction those misbehaving scrapers have with the service.
Wikipedia could be hosted on IPFS (or other content-addressed networks) without any issue. Update messages & transactions require authentication already (at least if accessed over Tor) and can be rate-limited if needed. Replication of successful transactions to content-addressed backing is easy.
What we are seeing is long-coming death of server-centric networking as the dominant paradigm. The c10k & c1M problems were part warnings. Past a point it doesn't work without absurd tradeoffs (such as ridiculous overprovisioning and power consumption).
Offline-first clients & user-agents should be a priority for either server-centric or content-addressed data access, because it enables for mitigating the downsides of either the best. Low-latency high-bandwidth unauthenticated server-centric access is simply not going to be a thing anymore because it is untenable. Sure, in the meantime we'll see very stupid doxing-based schemes but that won't work.
On a non-technical, social level, the economic incentives & properties that even allow the commercial-focus misbehaving scrapers to operate as they do must be destroyed. Their misbehavior isn't free and they have to fund it.
2