Re: I-D Action: draft-carpenter-6man-rfc6874bis-00.txt

[Nico Schottelius <[e-mail]>]

Jared Mauch <[e-mail]> writes:
> I hate to say it but mDNS is likely the way to go for many things, but
> you may not be able to tell the difference between raspberrypi.local
> vs raspberrypi-2.local which can often be a challeng

I don't think mDNS is feasible, as you often don't know the device
name. Let me illustrate a use case that is already there and is indeed
tricky for us at the moment:

We have a "bootstrap network" (layer 2) that contains dozens
or even low hundreds of unconfigured devices. All of these devices can
easily be *found* by using IPv6 multicast ff02::1.

However *some* of them might be broken, have configurations that need to
be repaired via HTTP. The device configuring these devices is usually
some kind of "admin notebook (or desktop)", which we try to configure
"as robust as possible", because there are tons of devices providing
IPv4, IPv6, etc. - it's basically the "device zoo".

Overall, most of these devices can be identified (EUI-64 embedding) and
configured (ssh, dhcp, etc.) automatically. However *some* need manual
care and the trend goes towards requiring an HTTP(s) connection.

What I want to show with this example is that you cannot easily rely on
more complex/upper layer protocols - the only identifier you have is the
link local address, the MAC and the open port list.

Cheers,

Nico

--
Sustainable and modern Infrastructures by ungleich.ch