[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[ih] protocol structure -> economic structure

On Mar 6, 2012, at 10:22 AM, Scott Brim wrote:

> On Tue, Mar 6, 2012 at 08:19, Craig Partridge <craig at aland.bbn.com> wrote:
>> The problem was, at the time, we didn't know how to build scalable, root-free
>> namespaces.  Grapevine showed that root-free namespaces were possible but
>> it had tremendous update costs.  So DNS had to have a root -- but the
>> expectation was that there would not be a lot of TLDS and by giving each
>> entity its own chunk of domain space, we'd achieved decentralization.
> Craig, would that not also reflect the operational orientation of the
> time?  It was important to have something that worked, relatively
> robustly?
> swb

During this same time, the routing infrastructure was also evolving from the
simpler tools we had (RIP, EGP) to BGP with explicit policy machinery. Like
the DNS, there was now explicit structure and policy boundaries which
enabled building larger networks and gave us the operational tools to
manage within subsets of the global whole.  This was "hard" before; I 
remember trying to interconnect disparate networks with RIP and trying
to puzzle over controlling route propagation in the presence of a bunch of
ad hoc interconnects.  It required very careful coordination between all
the parties.  

The introduction of these inter-domain (not DNS) partitioning tools made
all sorts of things possible operationally.  This became very important
when the NSFNET backbone appeared as an alternative "root" transit network
alternative to the ARPANET.