Patches Carved into Developer Sigchains

https://people.kernel.org/monsieuricon/patches-carved-into-developer-sigchains

by RyJones • 7 years ago

41 4 7 years ago

people.kernel.org

cjbprime2• 7 years ago

This doesn't actually work. I mean, the exact example does, but reasonable inferences about what else you could do with it don't work. Specifically, it doesn't work for kernel development.

Secure Scuttlebutt is a gossip network. You only see new activity from people you already follow. But kernel development involves anyone in the world being able to ask a question, post a patch, report an issue, without having to ask you to follow them first.

And if you did choose to enlarge the graph to encompass everyone, it would fall over from bandwidth multiplication, as gossip networks do.

SSB is great! But not for receiving activity from people you don't already have a social graph connection to.

Confiks• 7 years ago

I really don't understand how you can flatly conclude that "it doesn't work", based only on the gossiping nature of SSB.

Gossip networks indeed don't scale very well when not all participants want to communicate with each other (perhaps simplified as "don't already have a social graph connection"). But nothing in the example of kernel development necessitates to involve the entirety of the world.

Gossiping can be done over a medium that inherently reaches only an interested subset of people, either by self-selection or moderation. Trivially, you could start a verifiable gossip network on top of the kernel development mailing list, where most interested people have subscribed themselves.

SSB instead implements pub servers, which can be designated to contain gossip with a common topic. If and when gossiping becomes inefficient, you simply shouldn't continue to gossip indiscriminately; centralize and federate a bit, but preferably don't become reliant on a central logistical point.

cjbprime2• 7 years ago

The post talks about replacing the Linux kernel's use of email (so, e.g. the linux-kernel mailing list) with this new solution. linux-kernel has hundreds of thousands of subscribers and thousands of posters, all of whom are being given (according to this post) a full and complete copy of everything non-code related to its development.

I don't think it would work, at least while maintaining the same "anyone can post to the list" guarantees we have today. And I think it's important to say that it wouldn't work, so that people don't think the job of decentralizing open source communities has a substrate solution in-hand in SSB.

pcr910303• 7 years ago

For people who are interested in the decentralized web:

Technical explanation of the Scuttlebutt protocol they are talking about: https://ssbc.github.io/scuttlebutt-protocol-guide/

The scuttlebutt social network built on the scuttlebutt protocol: https://www.scuttlebutt.nz/

A peer-to-peer log store built on top of the protocol: http://scuttlebot.io/

Some other protocols that are aiming for the decentralized web:

Dat protocol: https://dat.foundation/

Technical explanation of the Dat protocol: https://datprotocol.github.io/how-dat-works/

IPFS: https://ipfs.io/

Also see my comment about the use of DHT(Distributed Hash Tables) in these decentralized web/protocols so that things like .org doesn’t happen. https://news.ycombinator.com/item?id=21612307

Patches Carved into Developer Sigchains

Related Stories