On 10/22/2012 09:30 AM, Kerry Lynn wrote:
On Mon, Oct 22, 2012 at 12:15 PM, Michael Thomas <[email protected] 
<mailto:[email protected]>> wrote:

    On 10/22/2012 07:22 AM, Tim Chown wrote:


        And I don't think this possibility is excluded at all by the current 
arch text either. It talks of an authoritative name service running on the CER, 
with a dynamic registration service (e.g. dyndns).  I have added 'as far as 
possible' to the unmanaged text.


    It's been hard for me to gauge what is and isn't in scope on the naming
    front. For the most part, this wg looks like it's only interested in putting
    some band-aides on mdns/dns-sd. That's a mistake, IMO, because there's
    nothing inevitable about a zeroconf driven design. We should consider
    littleconf solutions especially for naming because there is no way that
    mikes-toothbrush.mtcc.com <http://mikes-toothbrush.mtcc.com> got named that 
way coming from the factory.


Not sure how you can draw that conclusion; there seem to be lovers and
haters of mDNS.  To the extent that it's unsuited for multi-link topologies,
a BoF has been approved to discuss problems and requirements.  I hope
that folks interested in a DNS-based solution to service discovery and
naming will join the mdnsext mailing list.

As I mentioned, what I'm thinking about doesn't need multicast at
all. So it would be pretty bizarre to discuss getting rid of the need
for multicast dns on a mailing list that is bent on extending it.


I have nothing wrong with littleconf per se, but there seems to me a chicken
and egg problem: how does one engage a device (presumably running a
web-based UI) initially in order to administer it?  Is that by address or by
name?  If by name, then how does it work?

The device just announces itself to a repository, and the repository serves
up its name. No need to multicast each time you need to resolve a name,
it just unicasts it or maybe anycasts it at boot time, perhaps on a renewal
timer. If the rfc's that Brian's note references are applicable, so much the
better.

Mike

_______________________________________________
homenet mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/homenet

Reply via email to