3 ms·

I don't see why decentralization would lead to a spam problem. Presumably you would hardly see messages that haven't been approved by someone who has been tran

I don't see why decentralization would lead to a spam problem. Presumably you would hardly see messages that haven't been approved by someone who has been transitively approved by you. Decentralized systems are usually networks, and if you got spam or uninteresting messages you would reduce the weight of that part of the network. You can't trust a centralized party to censor for you, because then that entity has power over you through what you see. It has to be your secretary, i.e. it has to actually work for you and be a reliable servant. What better to use as a secretary than network of people connected by who is interested in the output stream from whom?

zinekeller 4y ago
looks at e-mail and SMS Sure you could install filters, but decentralised networks in practice has spam (the only ones that don't have any spam and don't have any filters are those too niche that spammers won't bother).
- actionfromafar 4y ago
  Then we have to checks notes - make every network niche! I quip, but I think there are solutions in that kind of thinking. Not everything needs to connect to everything.
- impossiblefork 4y ago
  Yes, but in both the case pf e-mail and SMS you are not looking at public streams, and neither uses a disyributed system. It uses a centralized system with a small number of telephone companies that are slow to respond to reports of spam. In a distributed system of the kind I propsed above, if you got a spam message you wpuld reduce your trust in those who transitively approved the source of the message, immediately. Someone spamming would be cut off quick, as would anyone with high trust in a spammer.
  - zinekeller 4y ago
    > In a distributed system of the kind I propsed above, if you got a spam message you wpuld reduce your trust in those who transitively approved the source of the message, immediately. I won't think too deeply if your solution works theoretically, but in reality in our imperfect world most people would just choose to outsource the job to the point that we're back to the email problem. Majority of people liked Gmail - majority chosed Gmail and spammers galore we have. Even Mastodon is not resistant to centralization despite their efforts to be decentralized - most wouldn't bother to set up their server and so a lot ends up signing up to mastodon.social.
    - TakeBlaster16 4y ago
      Parent's idea is called Web of Trust and it does work, but it doesn't have wide adoption for exactly this reason. For the average person it's too much work to curate their network, or to "debug" the chain of trust that let some spam slip by. https://en.wikipedia.org/wiki/Web_of_trust https://en.wikipedia.org/wiki/Web_of_trust
      - impossiblefork 4y ago
        Yes, although I imagine continuous connection weights and for the trust to be based on upvotes and downvotes on the public comments of others, not something for cryptography or private messages.
  - bcrl 4y ago
    Email and SMS are very much distributed systems. There are many mail servers, and many telephone companies, cell phone companies and VoIP providers that can send SMS messages. With SMS messages the problem is quite similar to e-mail in that the protocol the network is built on (SS7) was designed in a time when all carriers that had access to the phone network were trusted. You've fallen into the naive assumption that spammers use the same identity for each piece of spam being sent out, which is simply not true. I get plenty of spam phone calls with fake phone numbers despite the fact that STIR/SHAKEN has supposedly been implemented. Forged digital identities are cheap. And if forged identities are cheap, there's no trust mechanism that can work. Presumably you'll counter by saying that you can simply assign a negative trust to unknown identities. In the real world there are plenty of us that have to communicate with people we haven't communicated with before (my business can't afford to turn away customers I don't already know), so we can't simply drop unknown identities. The spam problem is hard because identities are hard, and digital identities are doubly so. Time and time again has proven that simple hand wavey solutions do not work.
    - impossiblefork 4y ago
      No, I haven't. A message from an entity which has not yet sent a spam message does indeed have the save priorization characteristics as one from any other entity withour connections to anything, but presumably a new user is invited by someone. Furthermore, a new user could presumably say things on the spam-filled rubbish subforums, until he gets an upvote from someone who is trusted by somebody, and then after a while he becomes a participant. The readon you see problems is, I think, that you imagine a system of private, individual, communication, while I imagine an enormous public forum.
      - bcrl 4y ago
        You are failing to address a use-case that matters. As a result, your solution can't be adopted by everyone, and will therefore fail. You need to consider privacy concerns, as not everyone in the world wants to see their web of associations made public. Look at privacy laws in Europe and other countries to get a better idea of what is required. Writing code is dead easy if you remove all the requirements from users.
varispeed 4y ago
The spam may not be as dangerous as other types of attack - that is when spammer gets frustrated that they get blocked, they may go nuclear and post illegal content to the site and then report it to hosting provider and / or authorities causing problems to someone running the node. And if you don't have good measures to filter or prevent such content from being uploaded you are toast.
- impossiblefork 4y ago
  Yes, although I don't think a site as such is a good idea. Instead I imagine a WebRTC application connecting to many peers and exchanging text with them. The main website then has little to do with the content. Furyhermore, it's not like people can't post illegal content even if they can't spam. In a way, your readoning regarding it is like responding to a threat as if the person in question can't realize it even if his conditions are met.

zinekeller 4y ago

looks at e-mail and SMS Sure you could install filters, but decentralised networks in practice has spam (the only ones that don't have any spam and don't have any filters are those too niche that spammers won't bother).

actionfromafar 4y ago
Then we have to checks notes - make every network niche! I quip, but I think there are solutions in that kind of thinking. Not everything needs to connect to everything.
impossiblefork 4y ago
Yes, but in both the case pf e-mail and SMS you are not looking at public streams, and neither uses a disyributed system. It uses a centralized system with a small number of telephone companies that are slow to respond to reports of spam. In a distributed system of the kind I propsed above, if you got a spam message you wpuld reduce your trust in those who transitively approved the source of the message, immediately. Someone spamming would be cut off quick, as would anyone with high trust in a spammer.
- zinekeller 4y ago
  > In a distributed system of the kind I propsed above, if you got a spam message you wpuld reduce your trust in those who transitively approved the source of the message, immediately. I won't think too deeply if your solution works theoretically, but in reality in our imperfect world most people would just choose to outsource the job to the point that we're back to the email problem. Majority of people liked Gmail - majority chosed Gmail and spammers galore we have. Even Mastodon is not resistant to centralization despite their efforts to be decentralized - most wouldn't bother to set up their server and so a lot ends up signing up to mastodon.social.
  - TakeBlaster16 4y ago
    Parent's idea is called Web of Trust and it does work, but it doesn't have wide adoption for exactly this reason. For the average person it's too much work to curate their network, or to "debug" the chain of trust that let some spam slip by. https://en.wikipedia.org/wiki/Web_of_trust https://en.wikipedia.org/wiki/Web_of_trust
    - impossiblefork 4y ago
      Yes, although I imagine continuous connection weights and for the trust to be based on upvotes and downvotes on the public comments of others, not something for cryptography or private messages.
- bcrl 4y ago
  Email and SMS are very much distributed systems. There are many mail servers, and many telephone companies, cell phone companies and VoIP providers that can send SMS messages. With SMS messages the problem is quite similar to e-mail in that the protocol the network is built on (SS7) was designed in a time when all carriers that had access to the phone network were trusted. You've fallen into the naive assumption that spammers use the same identity for each piece of spam being sent out, which is simply not true. I get plenty of spam phone calls with fake phone numbers despite the fact that STIR/SHAKEN has supposedly been implemented. Forged digital identities are cheap. And if forged identities are cheap, there's no trust mechanism that can work. Presumably you'll counter by saying that you can simply assign a negative trust to unknown identities. In the real world there are plenty of us that have to communicate with people we haven't communicated with before (my business can't afford to turn away customers I don't already know), so we can't simply drop unknown identities. The spam problem is hard because identities are hard, and digital identities are doubly so. Time and time again has proven that simple hand wavey solutions do not work.
  - impossiblefork 4y ago
    No, I haven't. A message from an entity which has not yet sent a spam message does indeed have the save priorization characteristics as one from any other entity withour connections to anything, but presumably a new user is invited by someone. Furthermore, a new user could presumably say things on the spam-filled rubbish subforums, until he gets an upvote from someone who is trusted by somebody, and then after a while he becomes a participant. The readon you see problems is, I think, that you imagine a system of private, individual, communication, while I imagine an enormous public forum.
    - bcrl 4y ago
      You are failing to address a use-case that matters. As a result, your solution can't be adopted by everyone, and will therefore fail. You need to consider privacy concerns, as not everyone in the world wants to see their web of associations made public. Look at privacy laws in Europe and other countries to get a better idea of what is required. Writing code is dead easy if you remove all the requirements from users.

varispeed 4y ago

The spam may not be as dangerous as other types of attack - that is when spammer gets frustrated that they get blocked, they may go nuclear and post illegal content to the site and then report it to hosting provider and / or authorities causing problems to someone running the node. And if you don't have good measures to filter or prevent such content from being uploaded you are toast.

impossiblefork 4y ago
Yes, although I don't think a site as such is a good idea. Instead I imagine a WebRTC application connecting to many peers and exchanging text with them. The main website then has little to do with the content. Furyhermore, it's not like people can't post illegal content even if they can't spam. In a way, your readoning regarding it is like responding to a threat as if the person in question can't realize it even if his conditions are met.