Re: [core] [SenML] Multiple values in a measurement

[Ari Keränen <[e-mail]>]

On 09/11/2018 1.37, Christian Amsüss wrote:
> Hi Michael,
> 
> On Thu, Nov 08, 2018 at 10:17:41AM -0800, Michael Koster wrote:
>> I don't believe RFC8428 places any normative constraints on the scope
>> of the identifiers that result from bn+n, either in section 4.5.1
>> names or 4.6 resolved records.
> 
> Quote:
> 
> | Name:  Name of the sensor or parameter.  When appended to the Base
> |      Name field, this must result in a globally unique identifier for
> |      the resource.
> 
> I'm not particularly fond of of that, but remember very clear reasoning
> from the authors that having unique IDs in there is a stronger
> requirement than playing nice with URI references.
> 
>> Specifically, I don't imagine there being a senML validator that would
>> check for the uniqueness of bn+n resolved names. I don't even know how
>> right offhand.
> 
> I figure "We funnel data from all our devices into our Enterprise
> server whose database behaves like a concatenation of all SenML files"
> was one application. That'd get confused quite fast if different devices
> all reported their first accelerometer as "3313/0/".
> 
>> What do the authors say?
> 
> I'd be very happy to be told I'm wrong on this (it would allow more
> elegant core-interfaces), and look forward to a response from them.

Yes, one of the design principles was that SenML names need to be 
globally unique so that they can be used e.g., as data base keys without 
conflicts.


Cheers,
Ari