[Olsr-dev] Fwd: Re: netjson plugin

[nemesis]

 Hi everyone,

 thank you all for bringing this up, by reading your discussion I 
 understood that the RFC is not crystal clear and needs to be improved.

 the goal of NetworkGraph is to represent a Network Topology for 
 visualization and monitoring purposes but the initial definition did not 
 allow distance vector protocols to return their partial graph using 
 NetworkGraph, so we changed the definition to allow this (so we could 
 write a generic collector, which we did). Too bad now the definition is 
 vague and can lead to misunderstandings.

 I believe in this case it's appropiate (and surely more useful) for the 
 olsrd1 implementation to return the graph of entire network.

 The discussion about local vs full topology started already a few 
 months ago and I opened an issue on github in order to avoid forgetting 
 the various proposals, see here:
 https://github.com/interop-dev/netjson/issues/25

 I did not have the time to work on that specific issue because I was 
 more focused on writing the RFC draft and write some implementations 
 (rough consensus and running code), netdiff was one of the first 
 implementations of which Gabriel is one of the core contributors. 
 Gabriel also proposed the creation of NetworkGraph and the addition of 
 the "local_addresses" attribute, he knows quite well the matter.

 In the following example, you can see an API that implements NetJSON 
 NetworkGraph, returing the output of the OLSRd1 ninux network in 
 Florence:
 http://ninux-graph.netjson.org/api/topology/643c4577-cef2-4b5e-b8a4-c29756b10748/

 The generated graph can be viewed at this URL:
 http://ninux-graph.netjson.org/topology/643c4577-cef2-4b5e-b8a4-c29756b10748/

 Ferry did send me a sample of the new plugin output, but I did not 
 understand it was intentionally limited to local links, my bad.

 I hope this clears the misunderstanding and I promise to improve the 
 spec and work on issue #25.
 If you have suggestions and alternative proposals please let me know.

 I will also meet Ferry, Henning and everyone interested in improving 
 NetJSON at the battlmesh v9 in Porto (1-7 May 2016).

 Best
 Federico Capoano

 PS: Lara is spot on :-)


 On Thu, 21 Apr 2016 22:47:17 +0200, Gabriel <(spam-protected)> 
 wrote:
> Could you give us your opinion on this?
>
>
> Thanks,  Gabriel
> ---------- Messaggio inoltrato ----------
> Da: Conrad Lara <(spam-protected)>
> Data: 21 apr 2016 22:23
> Oggetto: Re: [Olsr-dev] netjson plugin
> A: Ferry Huberts <(spam-protected)>
> Cc: Gabriel <(spam-protected)>(spam-protected)
>
>> I'm not an expert on netjson,   did I miss a line that says "known" 
>> links are local only on my quick read? If not doesn't OLSR "know" 
>> about all the links because of the OLSR packet forwarding?(dot_draw 
>> determines this already somehow to draw it's graphs)
>>
>> Just because one isn't directly connected to the link doesn't mean 
>> (unless RFC says so) that you don't know about it.  The RFC calls for 
>> a source and  destination id for the link mapping. That could easily 
>> handle showing the remote node. If the source was just the same node 
>> why would the protocol call for a source node ID definition when's 
>> just a destination would tell you everything you need because the link 
>> could only be local? It also doesn't state that the source id must be 
>> the presenting node.
>>
>> A quick google already showers at least one test sample file that 
>> shows this in action as part of the system test for netdiff 
>> https://github.com/ninuxorg/netdiff/blob/master/tests/static/netjson-3-links.json 
>> that it shows more then one source for the links to display (this is 
>> by no means a RFC proof of course as its not from the official source 
>> but makes me think this deserves more in depth discussion)
>>
>>
>> Sent from my iPhone
>>
>> > On Apr 21, 2016, at 12:48 PM, Ferry Huberts <(spam-protected)> 
>> wrote:
>> >
>> >
>> >
>> >> On 21/04/16 20:47, Gabriel wrote:
>> >> Hi, I've finally managed to get the netjson plugin working.
>> >>
>> >> I noticed that the "links" section contains just the 1-hop links.
>> >> The NetworkGraph should be a representation of the whole network 
>> as
>> >> edges and nodes.
>> >
>> > Incorrect, the RFC says:
>> >
>> >> Definition: a list of nodes and links known by a node.
>> >
>> > See http://netjson.org/rfc.html#rfc.section.4
>> >
>> >
>> > If you want the whole network, you must query all nodes and 
>> combine the info.
>> >
>> >
>> >>
>> >> Is it just a bug or it has been implemented to contain just the 
>> 1-hop links?
>> >>
>> >>
>> >> Gabriel
>> >
>> > --
>> > Ferry Huberts
>> >
>> > --
>> > Olsr-dev mailing list
>> > (spam-protected)
>> > https://lists.olsr.org/mailman/listinfo/olsr-dev
>>