[Olsr-dev] UUID plugin

[Ferry Huberts]

On 05-06-12 15:14, Henning Rogge wrote:
> On 06/05/2012 03:03 PM, Hans-Christoph Steiner wrote:
>> On Jun 5, 2012, at 2:33 AM, Henning Rogge wrote:
>>> Dependencies between plugins are a real pain in OLSRd at the
>>> moment. And constantly changing the MAC sounds like a good way to
>>> shoot yourself in the foot. ;)
>>
>> Depends on what you mean by "constantly". I was thinking that before
>> associating with a network, the device can randomize the MAC. That
>> way it should make it much harder to track a given device has it hops
>> around from network to network and back. I can't see anything there
>> that would shoot my feet. This would be primarily for mobile
>> devices.
>>
>>> There is also the problem that generating an UUID is not trivial on
>>> simple embedded hardware. If you cannot trust the MAC and the IP,
>>> there is not much you can use to generate a good UUID.
>>
>> If it uses the MAC and IP on the first time it generates it, then the
>> UUID is saved to a file, that seems like a workable method.
>
> Generating one during boot sounds like a sane solution. Still, you need
> to be prepared to have a MAC collision.
>
>> As for cross-plugin communications, now that I think about this, it
>> doesn't really need to be a plugin at all, but perhaps just a feature
>> of jsoninfo. If the file exists, jsoninfo will use its info. And
>> this file would be generated outside of olsrd.
>
> Which would mean you cannot access the UUID through the txtinfo.
>
> Maybe in case of the UUID it might make sense to store it in the core,
> because it could be also used as a seed (in addition to time) for all
> kind of randomness some plugins need.
>


it would be awesome if we could integrate UUIDs into the core of olsr v2 
as _the_ identity of the stack.

I've worked with jgroups for a long time and they discovered that using 
IPs/MACs presents problems and then switched to UUIDs. IIRC they 
generate a new UUID every time the stack is started, but I can check if 
anybody really wants to know.

currently we have the mainIP as identity of the stack and this is 
presenting problems as multiple people have already found out.

> Henning Rogge
>
>
>

-- 
Ferry Huberts