Distributed Social Networking Session Minutes
DKA: Question number one: Can a decentralized architecture be sustainable, profitable and usable?
That's an interesting important
PH: What is decentralized?
DKA asking where people are from. Handset vendors, operators, researchers, hardware vendors
DKA: I think the sustainable bit is interesting. Is a distributed architecture really what users want?
Blain: Email used to be centralized. And failed this way. Now open, distributed.
DKA: ... and triggered a lot of spam. It was so distributed that there was no authentication barriers, no way to prevent spam.
Sam: Economics really need to be considered. Key enabler for things to happen.
SP: People want to share pictures, music, voice. That can't happen for technical reasons.
Henry: Every entreprise has a social network, but people don't use it because it's too small to have a network effect. They rather use Facebook. But you cannot take Facebook to inside the company. There's a whole new market over there.
Kemp: We already have a decentralized architecture in a way: the user manages tens of identities himself, move stuff from one network to the other. Can we make a better place?
Blain: We have a fragment of the architecture. But I'm not sure we have a decentralized one.
Christine: We have silos today.
Henry: Rather than decentralization: distribution. You want every little piece to be linked together.
PH: Coming here as a newbie. What does "decentralization" mean? I put my FOAF profile somewhere and users can access it from other social networks?
Ori: On my own server, I could use services and have the same functionality as Facebook. No need to have one provider that handle millions of users at the same place.
Sam: We still rely on DNS to deliver content. URIs are not quite decentralized.
Miguel: If you distribute stuff, there is no real way to make money out of that.
Ori: Unless you have micro-payments. For the time being, there is no real business model for these big social networks.
Henry: If Facebook wants to sell micro-systems to e.g. companies, they need a decentralized architecture anyway.
Miguel: Anyone can code Facebook in a way.
Henry: talking about CIA, scribe missed that part/
Blain: Wikipedia has a list of social networkd. There are around 60 regional social networks, with no real overlap because of that. Facebook is the global face that makes it possible for a user of one social network to reach another one in another one.
Kemp: Another analogy. scribe missed it
DKA: there are lots of corporate interest to make money flow. One valuable information is to be able to profile users, which could be used. Dan mentioning he's about to have a breast augmentation operation.
Kunshik: The number of users who join the networks used to be important. It's interesting to see that all the existing mechanisms are converging, with a bit of boot-strapping, to make them inter-work.
Claudio: the question is when and who does users pay? Do you pay to use the network, send a message to another network? Roaming costs?
Kemp: or maybe we may actually end up paying users because that's valuable information. That's a possibility.
DKA: Yahoo! created a service that lets you specify which site may access your location, and specify the precision you are willing to give them (country, city, street, exact position). Links to question 5: How can we allow users who may want to deliberately fragment their online identity to do so?
Kunshik: Two perspectives for identity: from the user who logs in and from the application that uses the information. It's an individual prerogative. I don't think there's a lot we can do here.
Sam: It's important to control the amount of information application have access to, and it's also important to control how they can associate information.
Blain: For me, Facebook is equivalent to a chat room. Fragmented identity is even more important than privacy. The problem with fragmentation of course is when you want to create a bridge between two identities in some context. Fireeagle is federation stuff minus identity stuff. Fireeagle doesn't store identity.
Ori: When I posted on usenet 15 years ago, I had never imagined that information would still be there today. We have to think about 10 years in the future. If we fragment today, but if there's a way in the future to merge fragmented identities, then you can be sure that it will be done! If you have a GUID and give it to 20 different providers, you can be sure that, at a point in time, these providers will merge and exchange data.
Henry: This is just law of nature. In the end, truth comes out. It's a difficult job to have a mistress and not have your wife know anything about it.
Ori: Are you saying privacy is dead?
Henry: It's not absolute.
DKA: I wonder whether the Fireeagle architecture (no identity stored) is a good decentralization architecture? If I define policies on Facebook, I can then hold them responsible in case there's a problem.
Blain: The Fireeagle architecture was designed so that, even if Yahoo! was being asked to provide information there would be no way to disclose the identity of the user.
SP: If we define profiles in different places, we would quickly run into conflicting ones when trying to interoperate between different networks.
Miguel: Idea: You would have a number of information sources (position, friends). On another level, you would have social networks that would link these information sources in a way. Linked data.
Henry: Linked data. A way to link data and control who sees the links.
Sam: are we talking about a real infrastructure or the lack of architecture?
DKA: We're talking about extending the concept in Fireeagle.
Kunshik: The key words here are policy, and selecting link. Users in control. Policy-based service on selective information.
Kemp: I have Apache on my mobile phone.
DKA: Why would you need that?
Ori: This way you can carry on you, physically, your elements, your policies, your GUIDs. Other people cannot prove who you are in that case.
DKA: OK, so you run your own infrastructure then.
Blain: talking about centralization or not of privacy settings
Henry: User linked data. Access control. And you can do it.
DKA: Any social network, in order to survive, need to have APIs to survive.
Sam: Yes, but they want stuff in, but not let stuff go out.
Ori: What you can not do is link data.
Henry: If you take a look at LinkedIn. They are giving users a public URI.
Sam: talking about exporting contacts in Facebook and Windows Live Messenger
Ori: Do we need another protocol? I don't think so, we already have a bunch of them.
DKA: not so much about protocols, rather best practices.
Christine: Can we list the protocols that we want to use in this architecture?
Ori: OpenID, OAuth, RDF.
DKA: I worry about complexity. Going back to the Fireeagle example, I can figure out how to define my settings, but that may not be easy for everyone.
Kemp: Back to protocols. OpenID, OAuth, RDF are fine. But users talk in terms of social networks, formats. And they see a whole bunch of them.
Christine: LinkedData was one of them.
Blain: In terms of protocols, XMPP is definitely worth considering. I agree with John that it's very complicated. It's difficult to educate users, and even difficult to have developers understand and implement things correctly. My grand-ma has a list of emails written on a sheet of paper. That's her address book. Some people have thousands of contacts, but that's just a tiny percentage of users. Addressability. Latency. Data agnosticity.
DKA: I can already see themes emerging for a W3C's Incubator Group.
Christine: something to add, Dom?
Dom: Something I'd like to hear: what do you think W3C should do in the area? Let us know if you would get involved, and so on.
DKA: asking who would be interested
Dom: Do send me an email: firstname.lastname@example.org.
Henry: I'd like to create a FOAF file.
Dom: I think that's a great idea, indeed.
Henry: I read the 70 position papers for this workshop. And I had to search information on authors all over the Web. I managed to locate around 60 people. I've got some of the information, but if people have FOAF files that we could link to, send me an email.
Dom: I'll give you the list of email addresses of the authors, and if people are interested on doing FOAF files, then let's do it.