Re: [manet] Fwd: I-D Action: draft-ietf-manet-olsrv2-management-snapshot-01.txt

[Ulrich Herberg <[e-mail]>]

Hi Henning,

thanks for your review! See inline (only my comments, not yet
discussed with my co-author)

On Tue, Jul 8, 2014 at 12:36 AM, Henning Rogge <[e-mail]> wrote:
> [...]
>
> I think section 3.1 is worded a bit strong... its true that RFC5444
> messages can only contain a single address type as addresses, but
> RFC5444 packets can contain messages with different address lengths.
>
> The section reads (for me) that you cannot support both IPv4 and IPv6
> within the same routing daemon, which is not true. The olsr.org
> implementation does exactly this (even without my optimization
> tricks), it even runs a single dijkstra over a set of both IPv4 and
> IPv6 OLSRv2 nodes and edges.


I agree that this section needs some word-smiting. Good points!


> Do we want to mention that most Freifunk/Funkfeuer networks use TCP
> based queries (either pure text of HTML) with JSON output to query the
> state of the OLSR (v1) implementation... the OLSRv2 implementation of
> olsr.org will allow to do the same.

We could indeed mention that as an example of alternative interfaces
for managing MANET routers. I think it's a good idea.


> Would it be interesting to see if we can standardize a JSON query
> interface? Or if we can map the SNMP definitions to JSON in a
> reasonable way?

That is a bigger question than can (or should IMO) be handled in this
draft, but a very interesting one. And it may be even bigger than
MANET. Are you aware of the COMAN mailing list? They have discussed
management of constrained devices for some time, although so far no
BoF has resulted from these discussions. In the 6lo WG, we have
similar discussions for how to manage constrained devices. There are
proposals to reuse Yang models via HTTP
(draft-bierman-netconf-restconf-04) or COAP
(raft-vanderstok-core-comi-04). It would definitely be good to reuse
existing management interfaces (such as MIB modules and YANG). The
IESG has released a statement to prefer YANG over MIB modules for
writable configuration:
http://www.ietf.org/iesg/statement/writable-mib-module.html
I just recently had a phone call with three "B"-ADs on this topic
(Benoit, Barry and Brian), and they promised to send out an email to
multiple WGs regarding this topic, as it may require a cross-WG effort
(the email is still pending).

Coming back to your suggestion, JSON may be a good approach to manage
devices. And it's good to see that you already have an implementation.
I think we should discuss this more in Toronto. Will you be there?

Best regards
Ulrich