[FoRK] p2p / decentralization comments

Jeff Bone jbone at place.org
Mon Jan 25 12:22:49 PST 2010


Jim:  yes, yes, and yes.  ;-)

JM:  last paragraph, privacy concerns:  yes.

JAR:  exactly.

Yet despite universal agreement with these comments, I do think we're  
going to see some resurgence of decentralization.  The latency- 
gradient argument (Jim makes it, for example, with respect to git) and  
the privacy argument are, I think, going to become pronounced.   
Markets, eventually, do their best to "route around" damage --- and  
monopoly and its attendant centralization, resilience (lack of),  
privacy characteristics, etc. are going to be, like censorship,  
perceived as damage eventually.

E.g., Linux as reaction to Microsoft almost-complete-monopoly circa  
early 90s.  It *had* to happen or a lot of back room sysadmins were  
going to be very, very unhappy.  Microsoft could win the desktop, but  
the market *could not* let it win the server room.  So great was the  
pressure to react that it did something completely unexpected:  it  
took "value for server operating systems" off the table entirely.

Similar things could occur to would-be monopolies on the network,  
particularly "feature" monopolies vs. data monopolies.  But even data  
aggregators and distillers are susceptible.

A few really serious privacy leaks that cause real, widely-distributed  
economic data to individuals, for example.  A few major outages as the  
big guys --- who are already in an arms race to scale up --- begin to  
break (again).

Do we really think *anyone* on the planet has the technology *today*  
to smoothly scale from the maximum-traffic, data volume, etc.  
scenarios of today --- Google, for example --- to 10x that number?   
And those order-of-magnitude compound as more motes, devices,  
localizers, sensors, etc. what have you come on line...  You can't  
very well have a single, centralized provider of any given feature  
servicing all that sort of thing, unless it's *absolutely unavoidable*  
due to data aggregation requirements ala JAR's suggestion.  Much more  
can be accomplished end-to-end;  and as an existence proof, think  
about what the "RT search" folks that are looking at wrt Twitter, etc.

It may take a long time.  I may be overly optimistic to think this  
will occur in the 10s.  I doubt I'm off by more than a decade,  
though.  And I think that what we're likely to see is more functional  
decentralization, a kind of acceleration of existing trends. Despite  
the dislike of the usual VC for "it's a feature, not a product"  
companies --- you sure do see a lot of "features" out there in the  
landscape.  (In fact, you could almost *define* "Web 2.0" that way:   
features, not products.  Ironically, this is in part a reaction to the  
other side of the same bias on the part of capital;  when the most- 
likely liquidity event for a tech entrepreneur is not IPO but  
acquisition by e.g. Google, you're going to get a lot more "features"  
vs. "products" being built.)  But that collection of stuff comprises a  
kind of decentralized operating system;  unfortunately it's a bit  
fractured and tedious to use at some point, which is why you see end- 
users flocking to all-inclusive low-end resorts like Facebook, geeks  
without too much paranoia flocking to Google, etc.  "Destinations."   
But we had that before, and it fell apart when folks started offering  
up better implementations of ala carte features at lower cost (free.)   
No reason to think it won't happen again;  and at scale.  Lots of  
little gardens.  Fewer planetary ones...  the present arrangements  
aren't going to support 7 billion people.

Cf. previously-posted link to that Blade Runner / Autonomic  
Computing / Ambient Cloud thing for a bit more similar thought.


   http://highscalability.com/blog/2009/12/16/building-super-scalable-systems-blade-runner-meets-autonomic.html


$0.02,


jb



More information about the FoRK mailing list