Seemingly of interest specifically to:
vinton(_dot_)g(_dot_)cerf(_at_)wcom(_dot_)com
lixia(_at_)cs(_dot_)ucla(_dot_)edu
BADeemer10(_at_)aol(_dot_)com
At 08:57 PM 1/21/2002 -0800, Lixia Zhang wrote:
Note I am not saying MPLS is the right solution for the problem.
To me the right solution to the above mentioned problem should be a
multi-path routing protocol.
Dan K says:
Whether its MPLS and/or QoS or something else it occurs to me intrinsically
every extra byte that's not payload also is overhead. That's not saying its
not worth doing, just that it has a cost.
Also, Cisco et al are going to do some forklift upgrades here and there, and
there are admin costs (eg. testing) to making something new work on a
planetary scale.
(1) There should be a model thats like a spanning tree model, weighted for
any axis of freedom to make a completely deterministic solution to network
routing given different flavors or quality.
I worry though it might be a parellel to the bridge of Konigsburg problem.
Which either has no deterministic solution; (I can't understand why), or is
NP complete. Either is equally bad for a network with a billion people on it!
http://thesaurus.maths.org/dictionary/map/word/835
Its right to the philosophy of packet switching versus circuit switching. If
you book every hop and its mostly invariant, you just built a circuit
switched system from a packet one.
No body is going to *stop* anyone from doing that, but its driven by
different goals.
For instance, if in 2015 bandwidth went up 100 fold per user, would all the
QoS/RSVP etc just be overhead and get turned off? This wouldn't be so bad of
a thing.
It occurs to me its pretty complex to say the least. for instance is it
ethical to test a path from an application, yet ask for a RSVP circuit as a
backup in case congestion kills the higher bandwidth, less certain path? If
so its important the reservation processes have virtually no overhead if not
used. Seems like a hard goal.
Some infinite spanning process crawling Internet to discover paths and sort
of allocate them, that a piggy thing for sure!
And regarding tarriffed value added services, very un-internet like indeed.
Regards to all!
Regarding (1) seems like a component of that potential NSF process?