Re: [dnssd] Setting device friendly names

Right, the security model for having "the network" change the device's name
is wrong.   That doesn't mean it's not a problem that needs to be solved.
It may indeed be that there is no point in trying to address it in homenet;
I would certainly agree that it's not a high priority.   The main reason
it's of interest at all is for resolving name collisions.

In that sense, we _do_ have a way of making the device change its name--we
just don't get the opportunity to suggest a new name for it.

On Thu, Jul 21, 2016 at 3:48 PM, Alf Watt <alf@istumbler.net> wrote:

> Not have I, which seems suggestive: the feature has at least two proposed
> methods (UPnP and DHCPv4) and neither has been widely implemented in
> clients.
>
> It appears that clients which the IT staff doesn't already control do not
> want the network to be able to choose their name. I expect for the
> usability reasons already discussed: if the name is changed the user will
> have to learn the new name somehow.
>
> One other common method of host name configuration which hasn't come up is
> configuration profiles, which can automate away having (or allowing) the
> user to set the host name (or have one generated for them in setup).
>
> Best,
> Alf
>
> On Jul 21, 2016, at 6:13 AM, Ted Lemon <mellon@fugue.com> wrote:
>
> I have never seen a host that would accept an unauthenticated hostname
> update from a DHCP server.
>
> On Thu, Jul 21, 2016 at 3:10 PM, Alf Watt <alf@istumbler.net> wrote:
>
>> Don't forget, good ol DHCPv4 provides option 12 for setting host names by
>> the server:
>>
>> https://tools.ietf.org/html/rfc2132
>>
>> Support will depend on the client respecting the option, which at least
>> some Linux distributions don't seem to out of the box.
>>
>> The option wasn't included (even for the more common case of the client
>> requesting a name from the server) in DHCPv6, which suggests using DDNS for
>> allowing the host to update the name after auto-configuration.
>>
>> http://www.ietf.org/rfc/rfc3315.txt
>>
>> Best,
>> Alf
>>
>> On Jul 21, 2016, at 3:51 AM, Ted Lemon <mellon@fugue.com> wrote:
>>
>> I think the point Barbara is making about consistency is right, and that
>> we may be in violent agreement. :)
>>
>> In general there are devices that you'd like to be able to configure
>> using the network, and devices that you definitely do not want configured
>> by the network.
>>
>> On Jul 21, 2016 12:31, "Stuart Cheshire" <cheshire@apple.com> wrote:
>>
>>> On 21 Jul 2016, at 02:54, STARK, BARBARA H <bs7652@att.com> wrote:
>>>
>>> > UPnP SSDP announcements include a friendly name in the SSDP header. To
>>> me, this is comparable to the instance name in DNS-SD / mDNS. And
>>> comparable to a NetBIOS name.
>>>
>>> Yes. I don’t know any service discovery protocol that *doesn’t* have
>>> “friendly names”.
>>>
>>> The question was about a ubiquitous protocol for *setting* the “friendly
>>> name” that’s universally supported on (virtually) all devices.
>>>
>>> > When I ask my router for a list of hosts on the network, and when I
>>> ask my BluRay player to show me a list of content sources (UPnP), the lists
>>> use the same names for devices (because these devices use the same name
>>> across protocols).
>>>
>>> If one computer shares two USB printers on the network, do both shared
>>> printers have to appear with the same name (the same as the name of the
>>> computer that’s sharing them)? Restricting services to
>>> one-instance-per-host is quite limiting.
>>>
>>> <https://tools.ietf.org/html/rfc6760#section-3.3> Address Services, Not
>>> Hardware
>>>
>>> > Yes, there is also a UPnP Service defined that allows for remotely
>>> changing the friendly name.
>>> https://openconnectivity.org/upnp/specifications/friendlyinfoupdate1.
>>> But this service was only published in 2014 (long, long after the UPnP
>>> Device Architecture required "friendly name" in SSDP advertisements), it is
>>> not widely implemented (not mandatory to implement)
>>>
>>> Do you personally own *any* device that supports this? I know the
>>> “standard” exists. My question was how many products actually implement it.
>>> The world is full of so-called “official standards” that no real-world
>>> products implement.
>>>
>>> > FWIW, UPnP is an ISO/IEC standard.
>>>
>>>
>>> Yes, a great example: OSI networking.
>>>
>>> Stuart Cheshire
>>>
>>> _______________________________________________
>> dnssd mailing list
>> dnssd@ietf.org
>> https://www.ietf.org/mailman/listinfo/dnssd
>>
>>
>