3 ms·
Sorry, what? I think we are talking about very different concepts, Looks like you are talking about a "distributed data system", where the OP (and me) were head
by rglullis 12y ago
Sorry, what? I think we are talking about very different concepts, Looks like you are talking about a "distributed data system", where the OP (and me) were heading to the idea of a "personal website with social functions".
For me at least, I am thinking of a system that would be some kind of "personal cloud + feed syndication + acls based on friendships". I want the opposite of a business paying to know who you are. Diaspora got the initial momentum precisely because it was pushing for a privacy-aware version of Facebook. The way I read OP, the one thing that Diaspora got wrong was by putting the "social network" as the focus of their development, while it should've been the person.
I am not even going to argue about your idea, because it seems completely far off from what OP mentioned. Yeah, of course we could a have a system that could provide Facebook features in a distributed manner, and still have business paying for it. Turns out that "have someone paying for it" violates the requirement zero (user owned, personal cloud) of the product!
Lastly, I am sorry, but there is no way I can get my fridge to run a web server, connect to the home router and store gigabytes of data, _all the while keeping your personal network secure*. Saying that the cost of devices is trending to zero is just as magic-thinking of those in the 2000s who were selling that companies could save by switching immediately just by dumping their office licenses and switching to free software. This is flat out wrong[1]
[1]: http://www.joelonsoftware.com/articles/StrategyLetterV.html http://www.joelonsoftware.com/articles/StrategyLetterV.html
- dredmorbius 12y agoLooks like you are talking about a "distributed data system", where the OP (and me) were heading to the idea of a "personal website with social functions". Those are very nearly the same thing. Most websites and online services (and I've engineered several of them) are basically: 1. An application engine, which is a fancy term for "database wrapper script". Much of this can be dispensed with if you build a largely static site, though access control and some scripted / scheduled events are likely still necessary. This talks to ... 2. A data tier. Often multiple levels of caching and / or distributed data and or map reduce and / or nosql and / or database. Again, largely dispensible in a static site. 3. A lot of front-end caching and load balancing. The less content you actually have to serve, the better. It's better to think of a personal cloud + feed syndication (pretty much my formulation as well, BTW) as a cache-tier seed. That is, whatever kit you're running pumps data out to a distributed caching framework (think DNS or bittorrent). If a request cannot be met from cache, or it's expired, then your origin gets a request. But if it can be, you're serving content without actually taking the hit. And if the caching is demand-responsive, then you actually have the situation where you get better at serving data the more requests there are. A DDoS would simply result in more caching peers feeding your data. You also get the benefit that should your residential (or colo) link go down your cached data remain accessible online. Law of large numbers says that any given user is going to have relatively little traffic, but when the vast searchlight of the Internet Hive Mind takes an interest in you, you're swamped. So the sensible thing to do is to set up nodes such that the serve both as origin for their own data and as chache peers for other nodes. And I've got some sense of how activity levels distribute across users in social networks: https://ello.co/dredmorbius/post/nAya9WqdemIoVuVWVOYQUQ https://ello.co/dredmorbius/post/nAya9WqdemIoVuVWVOYQUQ It's also possible that there might be a value in providing more robust caching services for others on a wider scale. The remaining question is: what are the enabling technologies to make all of this happen? ∙ Reasonably inexpensive hardware. ∙ Reasonably inexpensive and reliable Internet connections. ∙ Zero-configuration software. ∙ Sufficiently compelling value proposition. ∙ As for your fridge running a Web server, you'd be amazed at what does. I've found all sorts of nutty stuff scanning networks, and a webserver is a trivially simple piece of software (there are kernel-space webservers and have been for over a decade). The truth is that hardware costs are falling massively. The $35 Raspberry Pi has served 1,000 Web requests per second. A terrabyte MicroSD card will fit on your thumbnail. Power draw is around 5W. https://www.jeremymorgan.com/blog/programming/raspberry-pi-web-server-comparison/ https://www.jeremymorgan.com/blog/programming/raspberry-pi-w... https://plus.google.com/+GuidoStepken/posts/9LGGyKASb8q https://plus.google.com/+GuidoStepken/posts/9LGGyKASb8q What you permit past your network firewall is your call, but my point is that we already are polluted with server-capable computers. The question is how to harness them. As to the "providing this as a service" aspect: not everyone will, can, or wants to run their own server, no matter how zero-maintenance they are. So, if your neighborhood school, or church, or community organization, or retail center, or local ISP, or other entity, wants to take care of doing this, and can handle user request by literally having them show up at the door, well, then all the better. It's not that you have to use a service such as this. But if that works for you, so much the better and power to you. My point as that this is an option, and that there might be ways to tie this to other reasonable but revenue-positive activities.
- rglullis 12y agoLook, we are going around in circles now. I understand your point about how in principle we could have all of the functionality of a social network in a distributed manner. What seems to be point of disagreement is that you are equating "reasonably inexpensive hardware and reasonably inexpensive Internet access" to "near zero cost of operating a node". If a distributed system really was the most economical alternative, then Facebook would be giving away free raspberry pi's with a collection of sd cards to run their services. They are not building huge datacenters all around the world because they are stupid. It's just that it is the least expensive way per user to run their services. By centralizing they get economies of scale that no distributed peer can ever achieve. To me, it's plain simple: 1. A raspberry pi ($35) + 128GB microsd (~$80) card puts you already at a cost of almost $10/month for the first year of operation. 128GB, quite frankly, is an amount of data that any reasonably active user of Facebook produces in a summer vacation. So, let me be easy on you and turn that $80 micro sd into a $60 1TB external HD, and we are still talking about $95 dollars just to start playing the game. 2. Still, you are smart enough to know "distributed caching" is not a magic word. You say "the less you store yourself, the better." Thing is, if you are not storing your own data, _who is_? In fact, for a distributed storage mechanism to work well, it needs to have multiple replicas of the data, so any node will actually store their own _plus_ other people's data - I remember a while ago reading that the minimum number of seeders on Bittorrent to work well is 6. So that 1TB for your node actually becomes less than 160GB. 3. Granted, no need for you to keep a backup, but this is assuming that every user is cooperating to the network with storage capacity, which is ever growing. So every year you should be buying another disk, meaning another $60/year. Even if we argue that the cost per gigabyte is going down, we could also argue that people will produce more data at an increasing rate so it evens out.u So to me just storage alone hits easily the $5/month/user mark. Even if you get all your super smart devices to be running in a perfectly cooperative manner around the house, you still need to bank storage. I really don't see a way now or in the near future to get the operating costs of this to zero. The only distributed systems that you see working well are the ones where there is either a financial incentive to keep a node (bitcoin) or political (tor) or where the costs of distribution and storage are negligible compared to the cost of production (Bittorrent). A personal cloud offers neither of those. TINSTAAFL.