[Dundi] [RFC] Reliability of contact information

[Florian Overkamp]

Hi, 

> -----Original Message-----
> Right, but if you only "peered" (shared keys and configuration
> information) notsoevilbox with your GPA counterpart, then 
> evilbox is a "secondary" node within your network. My point 
> is that it's possible to interpret the GPA language in such a 
> way that since evilbox is not _registered_ with any GPA peers 
> outside of your organization, it's not allowed to directly 
> place calls using GPA-controlled contexts (it would have to 
> place its calls through notsoevilbox, which is "registered"). 
> This may be completely bogus; I'm just suggesting that it's 
> one possible interpretation, and it may be a useful 
> interpretation if we want to actually be able to trace/track 
> calls back to their originators.

I agree that might be a way. However:

- article 2e defines that all initial control signalling must be sent from a
member of the Peering System (which would be either machine in the gpa
context). Technically this makes it very well possible that the initial
control signalling be sent from notsoevilbox. The example behind that
statement however makes clear that that is not how the article was intended:
'...commands must be sent from the requesting DUNDi node'.

A loophole in this would probably be if notsoevilbox would peer in GPA
context with the world and another context with evilbox. In this case, all
initial signalling will seem to come from notsoevilbox, but the (network of)
evilbox(es) would still be pretty much untraceable. However, the entire
network of evilboxes must be under your control or a GPA needs to be
executed with all parties in that network. (Remember the GPA is not strictly
tied to the e164 context in this aspect, since the term 'Propagate' is not
used. The network of evilboxes may ofcourse NOT publish any routes - that
would be a GPA violation if outside of the e164 context. Workaround: you
could create your own e164 context which is never connected to the 'public'
e164 context)

Florian