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

Re: [RRG] Geographic aggregation-based routing



I still do not understand what you try to teach me.
At first, this has nothing to do with geographical data. It would be the same task inside some OSPF-network.
 
The way I described how to use Dijkstra enables a Dijkstra shortest path tree which respects constraints like uni-directional forwarding. E.g. from node N1 to N2 but not from N2 to N1. And even conditioned: .. but not from N2 to N1 unless N1 is the destination node.
 
 
I didn't ask for adding red arrows but for expressing the task differently, by telling me which black line may be passed in which direction(s). If every node not only sees the topology of black lines but also the attributes per line (allowed forwarding direction(s)  ), then I don't see why my Dijkstra-based computation should produce incorrect paths.
 
Heiner
 
In einer eMail vom 13.07.2008 22:53:42 Westeuropäische Normalzeit schreibt bill@herrin.us:
On Sat, Jul 12, 2008 at 3:40 PM,  <HeinerHummel@aol.com> wrote:
>>> http://bill.herrin.us/network/geoag-h1.gif.
>
> Well, it requires a red arrow not only from B to C but also from C to B.
> But then all black lines are replaced by pairs of oppositely directed
> arrows.
> So we have no restricted situation anymore.
>
> Heiner, shruging my shoulders

Precisely. There is no configuration of red arrows you can draw that
correctly respects the permission constraint described by the green
arrows. Nor can we simply do away with the green arrows; they describe
the technical portion of the Internet's core economic model.

Hence we have found a contradiction: a valid network configuration for
which your algorithm can't meet the constraints. If this was math,
we'd say it was disproven. Since it's engineering, we say something
along the lines of: disqualified for technical error.

It turns out that -all- geographic aggregation strategies suffer from
a variant of this problem. That's why none of the rest of us are
interested in talking about geographic aggregation.

Your insights into other approaches, however, continue to be welcome.


> [the described approach] has always been useful for QoS/SLA/available bandwidth
> sensitive/etc  routing inside any OSPF network.
> Can either OSPF-experts or edu-folks comment on my question?

I used OSPF for my interior routing back in my ISP days. I have the
following two answers to that question:

1. OSPF is intradomain. Because all links are owned by the same
organization, permission is not a hard constraint the way it is in
interdomain routing.

2. I never got more than a trivial amount of aggregation with OSPF.
I'm not sure if the fault was OSPF or my configuration, but it didn't
offer up aggregation easily.

Regards,
Bill Herrin


--
William D. Herrin ................ herrin@dirtside.com bill@herrin.us
3005 Crane Dr. ...................... Web: <http://bill.herrin.us/>
Falls Church, VA 22042-3004