[FoRK] Stand up and face the future
J. Andrew Rogers
andrew at ceruleansystems.com
Mon Dec 21 11:31:16 PST 2009
On Dec 21, 2009, at 10:57 AM, Aaron Burt wrote:
> 2. Good partitioning and effective use of distributed hashes will get you most
> of the way to globe-scale (sorta-) shared state. The hard bit is getting devs
> to operate within the Heisenbergian constraints of a system that can tell you
> either what the state *was*, or give you a probablility distribution of what
> the state is *now*.
The serious limitations are not transactional, it really is that distributed hashes as currently imagined are deeply inadequate for global-scale shared state. Sure, it scales some concept of "state" but many (most?) interesting codes don't scale worth a damn if all you have to work with are cardinality relationships.
More information about the FoRK