On Sun, Jul 25, 2010 at 5:02 AM, Toni Stoev <[email protected]> wrote:
> 1. Hierarchical or flat uniqueness system?

If you pick global for #2 then the raging success of the DNS at
extraordinarily large scales makes the answer obvious.

If you pick local then the extra management overhead of a hierarchical
system makes the answer equally obvious. There's a -reason- enterprise
users often work with bare IP addresses instead of DNS names inside
their local systems and it's not because they're dumb or lazy.


> 2. Global/universal or just local uniqueness?

For which component of the system? Either answer is obviously correct
depending on which component you're talking about.

For the communication's initiator finding the acceptor? How would we
structure a non-global uniqueness system?

For associating packets with ongoing traffic between a particular pair
of hosts where one of them is an anonymous client? Most global systems
you could construct suffer byte overhead, management overhead, a loss
of desirable anonymity for the initiator, or all three.


> 3. Shall we make identity/location split?

Do you accept that overloading network topology, packet association
and finding the remote service all onto the same data element in the
packet is the root cause of our routing problem? If so then "don't do
that" would seem to be an obvious part of the solution.

Then again, that's also the punchline to, "Doctor, my arm hurts when I
move it this way." We should understand and perhaps even enumerate
what value lies in overloading before deciding to do away with it.

Regards,
Bill Herrin

-- 
William D. Herrin ................ [email protected]  [email protected]
3005 Crane Dr. ...................... Web: <http://bill.herrin.us/>
Falls Church, VA 22042-3004
_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg

Reply via email to