Comment by jokoon | Hacker News Reader

Comment by jokoon | Hacker News Reader

Parent Full thread

First, you're talking about websites, which will always require bandwidth. Why ? Because websites are narcissistic, expensive, and not designed for a dynamic experience. Websites are synchronous: the server generates a 500kB text blob file each time you want to communicate, it's zipped, unzipped, and parsed again, it's totally inefficient. Why ? because a webpage is just a string blob.

I live in france, and here, youtube is barely usable anymore, because everybody uses it. Streaming sucks, because video on demand is the worst pitfall of the internet architecture.

Websites are great for static pages. It's also great to advertise things, because html is easy, and quickly doable. But for other things, I don't believe it's really worth it. I dare you to find a html5, lightweight internet browser that can do optimized javascript, for what, checking your gmail, and connect to the facebook server farm? Why not using a mail application instead ? And then come up with SPDY, which is not so fast anyway.

Not everybody wants websites. People also like to share files, do video conferences, which are not possible with a centralized network, or really painful and expensive. Think about skype and bittorrent.

Those are economic alternatives because access network are not only expensive, they're not technologically viable.

Saying p2p is not viable because it is slow is a joke, because the way we use the network to make websites is already slow and require tremendous amounts of bandwidth. You're comparing airplanes and bicycles.

Also, talking about facebook, facebook is totally inefficient and not economically viable.

Seriously, a mainstream website to share photos and comments, have you seen their datacenters ? And facebook is not even a search engine. So what's google then ? Barely an assisted regex search.

You won't meet new people on facebook, you won't gather for new activities with strangers, it won't list available jobs opportunities or actually make relevant stats to tell people what jobs are needed, you won't find local businesses unless all of them subscribe to facebook, you won't post ads for something you neighbor could need.

So what is facebook for then ? Just to tell people you know, that you're making... a party. Or to advertise about your hyped bar.

Facebook is this huge device you connect to but can't use the data, because it's a trade secret. It's plain useless. That's because it use the star network pattern, instead of the mesh pattern.

I guess the main, elephant-in-the-room excuse, is because people are afraid of sex offenders.

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.

> 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.

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.

> 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.

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.

> 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.

that's the point of a network.

Wrong. The point of the network is to transmit data.

Who cares about that ?

Who cares about most efficiently getting to systems that work? Grandma, Aunt Tillie, the engineer, his boss that pays his salary, his boss's boss, his boss's boss's investor, and everyone else who matters.

Who cares about p2p? You and a couple of other people who don't matter, don't pay anyone's salary and don't get to decide anything.

You sound like a teenager. Maybe you should try to learn how technology decisions are taken in real life. You could start by getting a job in a technology company, but first here is a reading list for you:

http://soloway.pbworks.com/w/file/fetch/46715502/Crossing-Th...

http://archive.org/details/TheInnovatorDilemma

http://it-ebooks.info/book/166/

http://www.jwz.org/doc/worse-is-better.html

http://www.dreamsongs.com/Files/PatternsOfSoftware.pdf

> Wrong. The point of the network is to transmit data.

Data is mostly transmitted one way with those access network you're talking about. It makes me think of a tv or radio or cable network. That's the reason p2p does not attract anyone, because the current network is designed for people to focus on download instead, because that's the simplest software model. It only consists of supplying data to the consumer, like watching a tv channel. Internet is a mesh network first. That's how data flows.

> Who cares about most efficiently getting to systems that work? Grandma, Aunt Tillie, the engineer, his boss that pays his salary, his boss's boss, his boss's boss's investor, and everyone else who matters.

I was talking about product research. Technology does not invent itself. The boss you're talking about had no clue he could sell those things, and he did. I'm talking about lab rats, the people you fuel and inspire to make new stuff, not just use technologies in a way that has been done and done again. Facebook is already using the mesh paradigm for friend, but it requires a very high amount of computing power, and only facebook can access the data the way they want. That's a waste of good data. Other than that, facebook has brought zero new things to the internet. And it should not stop there.

> Who cares about p2p? You and a couple of other people who don't matter, don't pay anyone's salary and don't get to decide anything.

>You sound like a teenager. Maybe you should try to learn how technology decisions are taken in real life. You could start by getting a job in a technology company, but first here is a reading list for you:

Yes, I don't have a phd, I don't have a well paid job at a silicon valley company, so that's the #1 reason I'm wrong. I mean everything's its right place, why are there websites to talk about stuff, right. Let's just watch what nerds are cooking, and when it's sellable and usable, buy it and never modify it again, and lock it.

No surprise there was a net bubble.

I was talking about product research.

No you weren't! You were talking about the later stage: "going mainstream", which means efficiency, testing and deployment. But when I mentioned testing and deployment, you said Who cares about that?

that's the #1 reason I'm wrong

The reasons why you are wrong are contained in the damn links I went to the trouble of assembling for you. Why don't you go read them and learn something, instead of hitting me with a comeback? Sheesh.

> You'll have a devil of a time just deploying and testing the fucking thing, forget about development.

That's what you said.

When I said "Who cares about that ? Engineering is about results, not ease of work.", I was answering about "devil of a time".

> The reasons why you are wrong are contained in the damn links I went to the trouble of assembling for you.

The links you talk about might have to do with the IT business and how things go down between engineers, investors and managers. This is such an ugly topic, I don't even want to even try to talk about it. It's just not interesting. You don't go forward innovation by asking marketers and consumers. Marketing is about how you sell stuff, not what you actually sell.

That's what I hate about business in general. Business wants to do stuff, but it has no idea what to do. That's the reason the soviets were the first to send sputnik and a man on the moon. The US just followed. That's how innovation is broken. But that's just another topic.

I could not let you tell me that "things are good the way they are, because if they're like that, it's because we explored the business of it, and that's what we found out."

I didn't tell you things are good the way they are, I told you that the internet will always be centralized and p2p will never go mainstream.

You don't go forward innovation

Innovation is irrelevant to p2p because p2p has already been invented. Nobody wants it, and nobody has wanted it for 20 years, that's its problem.

This is such an ugly topic, I don't even want to even try to talk about it. It's just not interesting.

That uninteresting ugliness decides which technologies go mainstream and which ones fail. If you want to stick your head in the sand, this conversation is over.

> I told you that the internet will always be centralized and p2p will never go mainstream.

You seem confident about the future. I already read stats that p2p is actually quite a big chunk of the internet traffic. I never said "bring the internet full p2p", I just meant that p2p can play a role in what facebook, forums, and mail techs are already doing. p2p is not just for file sharing, file sharing is the easiest part of p2p.

> Innovation is irrelevant to p2p because p2p has already been invented.

p2p only work with protocols, and there aren't many protocols around people can actually use. There's a difference between invention and having a working standard.

And what does 'invented' really means here ? You don't just invent technologies, p2p is a very broad concept, you need to precisely define how users can make use of a protocol, what is the purpose of a particular protocol, what are the security and abuse issues, what sort of resource can be saved by using it, etc.

> if you want to stick your head in the sand, this conversation is over.

I don't want to stick my head in the sand, I'm just quite disappointed about how investors play ball with lab rats. That's how I feel IT and research business is done. Look at tesla motors, I'm sure Musk is one of the few investors who is not saying "fuck you with your nerd fantasies", and who is actually trying to change the industry.

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.

Reply on news.ycombinator.com