Munchkins are a killerApp of active messages.
See, today, on the wired Internet, only a small percentage of
intermediate machines carry the load of internetwork routing. A
ubiquitous wireless networking infrastructure, though, suggests
devolving routing responsibility to autonomous end-systems.
Munchkins represent a gedankenexperiment analyzing the possibility (and
desirability) of decentralizing routing, naming, and accounting services
to an unprecedented degree to enable untethered computing. A "munchkin"
is an extremely cheap (on the order of cents), extremely low power
(read: asynchronous chip technology), wireless (or at least untethered),
multiway, communication device about the size of a golf ball that will
be available in such quantities around the world (say, trillions,
possibly quadrillions) as to achieve "true ubiquity".
Munchkins offer the potential to surpass every existing communication
technology model, from radio to telecom to cable TV to satellites to
networked computers, and to transcend them.
When you think about it, the fundamental changes afforded by munchkins
will affect the underlying 'service model', moving away from the
'best-effort' packet-level delivery of today's Internet. New economic
and logistical constraints (i.e. ad hoc individual adminstration,
constant mobility) have implications for both the network layer and for
application protocol and middleware designers.
Imagine now that in addition to the incarnation aforementioned, the
intelligence in munchkins can be embedded in any device with a
connective cable, power cable, or wireless facility -- so we can
communicate with any such device either through electric current,
through networking, or through infrared. Basically, anything with a
(wireless or wirable) port can embed a munchkin communication chip.
What's on said chip? Why, something that speaks a universal transport
protocol, can mediate, can negotiate, and can automate tasks. And,
when a munchkin isn't doing work, it sits idle, consuming no power.
Plus, munchkins have the ability to route messages on-the-fly.
Now, as to why Web servers are important...
Imagine having an HTTPserver/ORB/InfosphereDjinnMaster as the program
that does the work of routing, transporting, exchanging, mediating,
negotiating, and automating. Imagine the use of XML for on-the-fly
typed authenticated secure document creation (complete with embeddable
components), WIDL for typed authenticated secure interface
specification, and naming/ownership/trust algorithms for adaptive
location and use of services...
How are things paid for? With kudos. I get some kudos every time I
agree to spend power routing someone else's messages for them, and I
spend some kudos every time I request information or send messages of my
own. Think of kudos as roughly corresponding to bits; an analogy is
FoRK, where the name of the game is to contribute (when you post useful
stuff as per the 10 Commandments of FoRKposting) as many bits as you
slurp (when you a. read the bits others have taken the time to harvest,
and b. waste the time of 24 people posting something they'll spend
time reading and/or deleting). Hence, if I am an infosponge, sucking up
bits like there's no tomorrow, then to hold my own in the kudo economy I
will have to contribute bits and/or bandwidth like there's no tomorrow.
But kudos are so much more. Kudos become some sort of ad hoc rating
system, because you can tell the quality of someone's information by how
much other clueful people are willing to pay for that information.
Additionally, an entire economy emerges: an advertiser can pay me kudos
in exchange for my reading the email he sends me, web surfers can pay me
kudos for the privilege to read the information-rich Web pages I make
available, and people running bit surpluses can give me kudos based on
my potential to put them to good use. In return, I can use kudos to buy
more bits and bandwidth. It's a radical bootstrapping of an economy not
possible with previous architectures; this economy dovetails the nature
of munchkins being endpoints AND routers in a trillion-node network.
One day, these may be more than just a dream.
Rohit, care to post your 20/02 vision?
:) Adam
----
adam@cs.caltech.edu
Are you beginning to see the possibilities here?
-- Strange Days