Re: [netmod] Benjamin Kaduk's Discuss on draft-ietf-netmod-geo-location-08: (with DISCUSS and COMMENT)

[Christian Hopps <[e-mail]>]

> On Jul 17, 2021, at 6:14 PM, Benjamin Kaduk <[e-mail]> wrote:
> 
> On Sat, Jul 17, 2021 at 02:38:55PM -0400, Christian Hopps wrote:
>> 
>> Benjamin Kaduk <[e-mail]> writes:
>> 
>>> Hi Christian,
>>> 
>>> Sorry for the very delayed reply (and thanks to Rob for the nudge).
>>> 
>>> On Wed, May 26, 2021 at 06:04:58PM -0400, Christian Hopps wrote:
>>>> 
>>>> Benjamin Kaduk via Datatracker <[e-mail]> writes:
>>>> 
>>>>> Benjamin Kaduk has entered the following ballot position for
>>>>> draft-ietf-netmod-geo-location-08: Discuss
>>>>> 
>>>>> When responding, please keep the subject line intact and reply to all
>>>>> email addresses included in the To and CC lines. (Feel free to cut this
>>>>> introductory paragraph, however.)
>>>>> 
>>>>> 
>>>>> Please refer to https://www.ietf.org/iesg/statement/discuss-criteria.html
>>>>> for more information about DISCUSS and COMMENT positions.
>>>>> 
>>>>> 
>>>>> The document, along with other ballot positions, can be found here:
>>>>> https://datatracker.ietf.org/doc/draft-ietf-netmod-geo-location/
>>>>> 
>>>>> 
>>>>> 
>>>>> ----------------------------------------------------------------------
>>>>> DISCUSS:
>>>>> ----------------------------------------------------------------------
>>>>> 
>>>>> I think we lack sufficient precision (forgive the pun) in how we talk
>>>>> about "accuracy" and "precision".  Are the leafs that claim to specify
>>>>> "accuracy" specifying a precision?  If so, the precision of a specific
>>>>> measurement, the precision of the measurements that led to the creation
>>>>> of the coordinate frame, or something else?  Are they doing so in
>>>>> relative terms (e.g., percentage) or absolute terms (e.g., degrees and
>>>>> meters)?  (There are "units" directives only for "height-accuracy" and
>>>>> not the others.)  How can we we say that we'll have 16 fraction-digits of
>>>>> precision for lat/long when the maximum accuracy we can say that a
>>>>> geodetic-system has only gives us 6 fraction-digits for coord-accuracy?
>>>>> When we say that the "precision of this measurement is indicated by the
>>>>> reference-frame" is that the same thing as the relevant "-accuracy"
>>>>> nodes, or something else?
>>>> 
>>>> Yes, the geodesic-datum is what defines the values and their accuracy. For the
>>>> precision in the value we choose the fractional digits based on what might be
>>>> needed, but not to prescribe anything. For decimal degrees e.g., we only need
>>>> 100s values the rest can be left to the fractional portion.
>>> 
>>> Unfortunately, even your description here still doesn't help me understand
>>> what the intended semantics of these values are.
>>> 
>>> To help illustrate my confusion, here are a few possible things that could
>>> be what is intended to be conveyed:
>>> 
>>> - the geodetic-datum description of the object has been measured to be
>>>  within a known delta of the actual object being described, at all points
>>>  on the object that the coordinate system can describe
>>> 
>>> - the geodetic-datum description of the object is capable of determining
>>>  relative differences between points on the object to within a particular
>>>  delta of precision, but those individual coordinate values may be farther
>>>  than that delta from the actual point on the object that was referred to
>>> 
>>> - the values that are reported in this YANG module reflect measurements
>>>  that were made and are known to be within some delta of the coordinate
>>>  system's value that they are reported as
>>> 
>>> - the values that are reported in this YANG module reflect measurements
>>>  thare are known to be distinguishable from other measurements to within
>>>  some delta of other measurements relative to that coordinate system, even
>>>  though the actual position being indicated may diverge from the reported
>>>  value by more than that delta
>>> 
>>> - the values that are reported in this YANG module reflect measurements
>>>  that were made and are known to be within some delta of the actual point
>>>  on the object that the coordinates refer to
>>> 
>>> - the values that are reported in this YANG module reflect measurements
>>>  that were and are known to be distinguishable from other measurements of
>>>  points on that object within some delta, but the actual distance from the
>>>  measured point to the point on the object indicated by the reported
>>>  coordinates may be larger than that delta
>>> 
>>> In short, there are at least three classes of things at play here: the
>>> actual object itself, the coordinate system used to model the object, and
>>> values reported in the YANG module (which are assumed to ultimately derive
>>> from some form of measurement).  To talk about accuracy or precision
>>> implies a relationship between elements of two of those classes, and I
>>> don't even know which of those classes you're trying to talk about.
>> 
>> Let's start with a simple baseline, if you want to dig any deeper than the well understood Lat+Long; do you know what a geodetic datum is? This is required knowledge if you want to get into anything more than the obvious Lat+Long use of this grouping. It defines the coordinates and also the accuracy of measurements.
> 
> Well, I thought I did, but the fact that you are asking me makes me less
> sure that I actually do.
> 
> Limiting just to the Earth for simplicity of discussion, it is "well known"
> that the earth is not a perfect sphere; it's not even a regular ellipsoid.
> Even discounting local topography on the surface, "mean sea level" varies
> due to the differing internal density, angular momentum, and myriad other
> factors.  So if we want to talk about coordinates of a point on the earth,
> we have to build a model of the earth in which we define what our
> coordinates mean.  We'll have to anchor our coordinate system to actual
> points on the earth in some way, whether by defining an arbitrary origin at
> a physical object, using the center of mass (which can at least in theory
> be measured to very high precision), or some combination thereof.  But the
> coordinate system remains a model of the actual earth, and there will be
> some skew between them for points that in an ideal coordinate system would
> match up exactly with the physical object.
> 
> The coordinate system will have inherent accuracy limits based on how much
> skew there is between the idealized points in the coordinate system and the
> actual points on earth they're supposed to represent.  When one makes a
> measurement with respect to a given coordinate system there may also be
> inherent limits to the precision of measurement that can be made with
> respect to the coordinate system, e.g., if the coordinate system is defined
> with respect to GPS points, the limits of GPS resolution are a bound on how
> precisely one can make a measurement.
> 
> That's all well and good, but what I describe above is a property of the
> coordinate system itself, not a property of individual measurements made
> with reference to that coordinate system.  The YANG grouping we define here
> allows overriding the coord-accuracy and height-accuracy on a
> per-grouping-instantiation basis, i.e., for a single list of coordinates.
> But those coordinates in the instantiation are certainly sometimes going to
> be derived from measurements, and I expect that measurements will be the
> overwhelming majority of usage.  Measurements, however, *also* have
> accuracy and precision, but this time with respect to the coordinate system
> they are being measured in.  Instrumental error and other factors can
> introduce a systemtic bias in the measured values, leading to bad accuracy,
> even if the precision of the group of measurements remains quite good, so
> that relative comparisons within the dataset are reliable even if the
> absolute numers are not reliable with respect to the coordinate system.
> 
> So, when we refine the coord-accuracy and height-accuracy for an
> instantiation of the grouping, what does that mean?

It’s supposed to mean the accuracy of the measurement that is recorded in the grouping. So if the coord-accuracy is .1 and the measurement is lat/long then the accuracy is within 1/10 of a decimal degree. if the measurement is in cart coordinates the accuracy would be 100cm. I don’t think we need to make this anymore complex than that. Is there some text you would like to see to make that clearer?

Thanks,
Chris.

> 
> 
>> It is out way out of scope for this YANG grouping to try and explain the huge field of geographic locations and geodetic datum and systems.
> 
> Of course.  But we should have enough of a reference so that people can
> have a way to read up and understand what the fields we are defining
> actually mean.
> 
> -Ben