[Interop-dev] Topology JSON

[Gabriel]

On 15/02/2015 17:55, Henning Rogge wrote:
> On Sun, Feb 15, 2015 at 5:16 PM, Nemesis <(spam-protected)> wrote:
>>> I think NetworkRoutes is the representation of the routing table
>>> created by a routing agent. Doesn't matter how it was calculated, just
>>> a bunch of destinations, next-hops, interfaces and "costs" to give you
>>> an idea how far the destination is away compared to other routes.
>>>
>>> NetworkTopology is the knowledge of the links in the whole mesh...
>>> maybe we should split this into two "subgroups".
>>>
>>> The "neighborhood" should contain knowledge about other routers you
>>> can directly communicate to (and maybe data you directly got from them
>>> for two-hop knowledge?). This is something the local router collected
>>> itself and has a very accurate knowledge about. I think both
>>> distance-vector and linkstate protocols collect this information.
>>>
>>> The "topology" is knowledge about the whole network graph, most likely
>>> based on information you got from other routers. Linkstate protocols
>>> need this data to do their job, distance-vector protocols don't (but
>>> might deliver it through some optional query mechanism).
>>
>> Do you think something like this would suffice?
>>
>> http://pad.eigenlab.org/p/netjson-topology
>>
>> Or do you think something is missing?
> 
> The metric should be neighbor dependent... with a topology, you are
> representing (at least for OLSR) data between nodes.
> 

with olsr2 can you have different metrics in the same topology?

we should find a normalized way to memorize the weight on each edge, so
we can have drawings app that don't have to care about the metric used
by each routing protocol.

Gabriel