6 ms·

Let's say you want to send a multimedia collection from your co-lo to someone in a different continent, via a relay. You might be able to relay it via (someone

by ucee054 13y ago

Let's say you want to send a multimedia collection from your co-lo to someone in a different continent, via a relay. You might be able to relay it via (someone like) me, because I get megabits per second download, but I get much less upload, so your target is in for a tedious wait. But, if you send it via Amazon web services, you might be able to direct connect to their hosting centre near you. From there, you could get a gigabit per second over Amazon's own network to a centre near your target where he/she can download from - instantaneous. And we're not even talking about Amazon's datacentre internal networks, where you might be able to arrange for 10 gigabits. Centralized beats decentralized by orders of magnitude. Sorry.

jokoon 13y ago
> multimedia collection How many people really want to do that ? You just better send it through mail honestly. Most people care about text data, which is very small if you handle it well.
- ucee054 13y ago
  How many? I don't know, but I do know that Grandma wants to send the photos from last night's birthday party to Aunt Tillie back in the old country. And neither is on bittorrent. I think the mistake you are making is that you think "most people" = you, whereas in reality "most people" = Grandma. If you are the guy responsible for building some internet service for Grandma to use, you will always go centralized, because that's where the bandwidth is. p2p only wins in some weird edge cases - in file sharing it's in part not to enable copyright holders to shut the system down through legal action against single points of failure.
  - jokoon 13y ago
    > I think the mistake you are making is that you think "most people" = you, whereas in reality "most people" = Grandma. I'm trying to make you understand p2p techs can go mainstream, I'm not trying to argue about making a tech only for tech aware people. Bittorrent is already used in a mainstream way. Please don't argue about a tech that appears too complex it can't suit old people. Very few people understand kernels, yet many use computers everyday without minding. With that mindset, p2p will never become mainstream. If you want innovation to suit 95% of what people want, you're screwed, because people don't know what technologies can do for them. Sometimes you should let engineers or tech-savvy marketers propose a little more that what people will always want to try. There are many people out there who like to be surprised with technologies, and that's also a market.
    - ucee054 13y ago
      Look, I have actually studied p2p. I'm sure you have not. I doubt, for example, that you could quote off-the-top-of-your-head the research papers relevant to p2p. If you had done your homework, you'd realize that p2p has been around since the nineties, and its only successes have been some weird military-type applications, file sharing, file distribution and Skype. So p2p is not a new, superior paradigm about to burst into the mainstream, only being held back because of Moaning Minnies like me not pushing it hard enough. It is actually an old, tired, second-hand, inferior technology that has been failing to go mainstream for twenty years. "Please don't argue about a tech that appears too complex it can't suit old people" You actually have to read what I wrote. Part 1: I explained to you how technology is deployed to suit people who can't deal with complexity: an engineer builds it into a product/service for them - making the service simple enough to use. Part 2: I explained to you why said engineer would never use p2p for said service: because all the bandwidth is available in the centre. You know, to make a p2p system that can even approach the ballpark performance of a centralized system, you have to deal with connection asymmetry, connection heterogeneity, overlay construction and churn tolerance. You'll have a devil of a time just deploying and testing the fucking thing, forget about development. And you still get a less performant system overall. The tech doesn't "appear" complex, it IS complex. For Grandma to be able to use it, the poor sods in Engineering have to make it "appear" simple, which makes the complexity THEIR problem. This is NOT something any sane engineer would want if he could avoid it by getting Grandma what she wants by using a simpler alternative. Like a central server, for example.
      - jokoon 13y ago
        > ballpark performance of a centralized system Performance in this case is overrated. You forget people want to have independent system which can't be shut down or controlled, that's the point of a network. > You'll have a devil of a time just deploying and testing the fucking thing, forget about development. Who cares about that ? Engineering is about results, not ease of work. Once you get results in engineering, you have them for life. Business wants easy solution, that's understandable, but the engineer's job is also to bring new products not done before. Consumers and businessmen rarely or never know about what can be done. > This is NOT something any sane engineer would want if he could avoid it by getting Grandma what she wants by using a simpler alternative. If the "simpler alternative" is now the majority, I can't understand how you can call this an alternative. At the beginning, networking was designed as in anybody can send to anybody. Most of the time engineers deliver solution to problems people are complaining about, but sometimes there are inherent problems only the engineers are aware of, in which the solution can actually create true value.
jokoon 13y ago
You're saying this like nobody uses bittorrent. Decentralized completely solves your upload problem. You're comparing planes and bicycles. Many people are content with bicycles, and that's why they are popular, but very few people actually need and use airplanes.

by ucee054 13y ago

jokoon 13y ago

> multimedia collection How many people really want to do that ? You just better send it through mail honestly. Most people care about text data, which is very small if you handle it well.

ucee054 13y ago
How many? I don't know, but I do know that Grandma wants to send the photos from last night's birthday party to Aunt Tillie back in the old country. And neither is on bittorrent. I think the mistake you are making is that you think "most people" = you, whereas in reality "most people" = Grandma. If you are the guy responsible for building some internet service for Grandma to use, you will always go centralized, because that's where the bandwidth is. p2p only wins in some weird edge cases - in file sharing it's in part not to enable copyright holders to shut the system down through legal action against single points of failure.
- jokoon 13y ago
  > I think the mistake you are making is that you think "most people" = you, whereas in reality "most people" = Grandma. I'm trying to make you understand p2p techs can go mainstream, I'm not trying to argue about making a tech only for tech aware people. Bittorrent is already used in a mainstream way. Please don't argue about a tech that appears too complex it can't suit old people. Very few people understand kernels, yet many use computers everyday without minding. With that mindset, p2p will never become mainstream. If you want innovation to suit 95% of what people want, you're screwed, because people don't know what technologies can do for them. Sometimes you should let engineers or tech-savvy marketers propose a little more that what people will always want to try. There are many people out there who like to be surprised with technologies, and that's also a market.
  - ucee054 13y ago
    Look, I have actually studied p2p. I'm sure you have not. I doubt, for example, that you could quote off-the-top-of-your-head the research papers relevant to p2p. If you had done your homework, you'd realize that p2p has been around since the nineties, and its only successes have been some weird military-type applications, file sharing, file distribution and Skype. So p2p is not a new, superior paradigm about to burst into the mainstream, only being held back because of Moaning Minnies like me not pushing it hard enough. It is actually an old, tired, second-hand, inferior technology that has been failing to go mainstream for twenty years. "Please don't argue about a tech that appears too complex it can't suit old people" You actually have to read what I wrote. Part 1: I explained to you how technology is deployed to suit people who can't deal with complexity: an engineer builds it into a product/service for them - making the service simple enough to use. Part 2: I explained to you why said engineer would never use p2p for said service: because all the bandwidth is available in the centre. You know, to make a p2p system that can even approach the ballpark performance of a centralized system, you have to deal with connection asymmetry, connection heterogeneity, overlay construction and churn tolerance. You'll have a devil of a time just deploying and testing the fucking thing, forget about development. And you still get a less performant system overall. The tech doesn't "appear" complex, it IS complex. For Grandma to be able to use it, the poor sods in Engineering have to make it "appear" simple, which makes the complexity THEIR problem. This is NOT something any sane engineer would want if he could avoid it by getting Grandma what she wants by using a simpler alternative. Like a central server, for example.
    - jokoon 13y ago
      > ballpark performance of a centralized system Performance in this case is overrated. You forget people want to have independent system which can't be shut down or controlled, that's the point of a network. > You'll have a devil of a time just deploying and testing the fucking thing, forget about development. Who cares about that ? Engineering is about results, not ease of work. Once you get results in engineering, you have them for life. Business wants easy solution, that's understandable, but the engineer's job is also to bring new products not done before. Consumers and businessmen rarely or never know about what can be done. > This is NOT something any sane engineer would want if he could avoid it by getting Grandma what she wants by using a simpler alternative. If the "simpler alternative" is now the majority, I can't understand how you can call this an alternative. At the beginning, networking was designed as in anybody can send to anybody. Most of the time engineers deliver solution to problems people are complaining about, but sometimes there are inherent problems only the engineers are aware of, in which the solution can actually create true value.

jokoon 13y ago

You're saying this like nobody uses bittorrent. Decentralized completely solves your upload problem. You're comparing planes and bicycles. Many people are content with bicycles, and that's why they are popular, but very few people actually need and use airplanes.