Loading ...
Sorry, an error occurred while loading the content.

7048Re: The decline of P2P and Decentralisation

Expand Messages
  • Dan Brickley
    Jan 25, 2010
    • 0 Attachment
      --- In decentralization@yahoogroups.com, Julian Bond <julian_bond@...> wrote:
      > Whatever happened to P2P and Decentralisation as a design pattern?

      Good question and an interesting thread. I got into this stuff via the digital library scene, where we prematurely ran around saying things like 'these centralised altavista-style search engines will never scale', and looked into distributed search along the Harvest (http://harvest.sourceforge.net/) and WHOIS++ (http://www.dlib.org/dlib/january98/01kirriemuir.html) lines.

      So when massmarket user-facing P2P came along, I felt like the digital library community had a big red face, and that the fact that these systems came from hackers rather than big research projects was a lesson the public-funded research scene dearly needed to learn from.

      Looking back now, I think the lessons are a little different.

      When we talk about 'decentralization' we use a fundamentally physical metaphor, to talk about some very complicated online systems and some fairly subtle multi-year situations. And by slipping too easily into thinking of sites as 'places' we frame the 'is this decentralised?' question in misleading ways. We think of Twitter or Facebook or LiveJournal as a place, and worry that too many things are centralised in that one place. A lot of the early enthusiasm for things like RSS and FOAF came from a sense that the Web is healthier when stuff we care about lives in lots of different places.

      I think we were right, but we mixed up hosting, naming and data formats, and didn't think about time carefully enough. It is very very hard for data formats to compete with hosted services. Files alone don't do much, they just sit there. Data formats plus some standard service APIs gets you closer (eg. opensocial), ... but it's a false opposition. We over-emphasised run-time decentralization, and under-emphasised decentralization in the sense of having the ability, the right and the tooling to be able to take your data off to some other hosting environment at a later date. The key thing here isn't data formats (although they are needed), but naming. DNS not XML/JSON.

      I am quite happy having had my photos all hosted by Flickr (= Yahoo now), but increasingly frustrated that the canonical URLs for everything I've published (and therefore everything that has been linked to) are Yahoo-owned properties.

      Somewhere along the way, we lost the expectation of user-owned domains being important. Never mind the hosting, it's always smarter for someone else to do that, or to share the burden with others who're paying the bills or doing some of the sysadmin. The problem is around domain names. Most Web sites are on user-owned domain names. Many many blogs are on user-owned domain names. But as we move into the microblogging and 'social network' scene, this expectation seems to drop away, and we're all happy being http://twitter.com/{something} instead of http://chatter.{something}.name/

      The difference isn't apparent to most users, and won't be until they get fed up and want to move their stuff. And all the APIs and open REST data feeds in the world won't save them from the fact that they're locked in by hyperlinkage rather than closed data.

      I'd love to be able to 'dock' my photos with Flickr/Yahoo, have those guys professionally serve my stuff and integrate it with groups and friends and so on in exchange for $ or ad-hosting, ... but be able ultimately to disentangle it and move it elsewhere with minimal link breakage. Same with blogs, same with microblogs.

      So I've no problem with centralisation, so long as we have medium-long term option to move stuff around later. Previously . Which means I think looking more carefully at making the DNS less scary for end users, and at the idea of these mega-sites spanning multiple user-owned domain names. And since this can conflict with anti-phishing user training if not done carefully, is not a trivial matter. When you click on my name in Facebook, could you be taken to a Facebook-generated or MySpace-generated page that was in a domain owned by me and 'docked' at one of those sites? Could Facebook retain a consistent user experience in the face of such Webby chaos?


      ps. related, proposal for 'dockable' accounts in identi.ca --- http://lists.status.net/pipermail/statusnet-dev/2009-August/002117.html
    • Show all 43 messages in this topic