Re: [Anima-signaling] GRASP issue 49: More text about inter-domain GRASP

[Brian E Carpenter <[e-mail]>]

On 02/08/2016 23:50, Toerless Eckert wrote:
> Its really hard to define requirements when you don't know what the
> communication should ultimately do.

Absolutely. And we shouldn't stretch the protocol spec to cover future
application scenarios that we don't understand. I'd rather limit what
we say to guaranteeing security properties.

> So, for example:
> 
> In our Cisco implementation, we allow services to run on systems that
> do not have ACP - just because its evolving work to get the ACP onto
> such systems. Eg: TFTP servers, Radius Servers - or the like. SO there's
> a leg of the ACP that's unencrypted. And the deployment requirement is
> to physcially protect this segment - aka: router with ACP co-located with
> those servers n a secure NOC room. And we use mDNS there.
> 
> The same setup would make sense to me to accelerate deployment also of
> course with GRAP - instead of mDNS.
> 
> So, this is an example of expanding the secured insance of GRASP across
> a segment thats not cryptographically (IPsec) but physically protected.
> 
> Now i could easily imagine that the next step would be to have multiple
> disjoined autonomic networks, but a shared NOC. In that case, the
> reason that we don't use encryption is not only "server systems suck, have no ACP",
> but also: If a server should provide objectives (services) to multiple
> autonomic networks, then we would need to solve how it could be cryptographically
> part of multiple ACPs. Thats even more work.
> 
> If i take your text and would want to build a solution around it, then
> i could think of some gateway-device thats part of two ACPs, and
> has an ASA participating in each ACPs GRASP instance and is filtering/forwarding
> objectives that should be allowed to be used across ACPs. Makes sense too...

Yes. I keep coming back to my old love: diffserv and how to deploy it between
domains. I can see ASAs being deployed as inter-domain bandwidth brokers,
but interworking with intra-domain bandwidth brokers, exactly as you say.

Rgds,
   Brian

> Cheers
>     Toerless
> 
> On Tue, Aug 02, 2016 at 01:39:35PM +1200, Brian E Carpenter wrote:
>> Hi,
>>
>> I came up with an idea of how to tackle this: state that any such inter-domain
>> operation MUST be a separate (D)TLS-based instance of GRASP that does not
>> share common data structures with the normal (ACP-based) instance.
>>
>>     Brian
>> On 01/08/2016 09:11, Brian E Carpenter wrote:
>>> Hi,
>>>
>>> Here's another issue for design team comments.
>>>
>>>    o  49.  Section 3.3.1 should say more about signaling between two
>>>       autonomic networks/domains.
>>>
>>> I need ideas about what to say.
>>>
>>> Regards
>>>    Brian
>>>
>>
>> _______________________________________________
>> Anima-signaling mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/anima-signaling
>