CTI STIX Subcommittee

 View Only
  • 1.  Location, latitude/longitude, and precision

    Posted 07-18-2017 21:33




    Hey everyone,
     
    We’ve been having a conversation about the best way to represent the precision of a given latitude and longitude pair. What I mean by that is, if I give you 45.234234 latitude by 45.2342 longitude (literally
    just typed random numbers there) is that describing an exact point, just pointing to a city center, or even just pointing to a country?
     
    We have a couple options here:
     

    Just define in normative text that, for STIX, latitude and longitude is approximate. This would mean that you couldn’t really rely on it to be accurate
    to an address, even in cases where it was. Have an optional property to represent the precision. If the property is present, it describes the precision in terms of meters. If absent, we have two
    options:

    Precision is unspecified, and consumers can take that however they want Default precision to perhaps 10km (roughly a zip code), so consumers can pretend the value was that
    Have a required precision property
     
    We also talked about significant digits but due to the ways JSON parsers read data it would mean using a string and likely that data would also be lost in those cases too (and even if there it’s harder to
    get to). It didn’t seem like a great option compared to any of the above, but maybe we’re missing something.
     
    So, what do people think? Any other arguments for one or the other?
     
    I don’t have any strong opinions really other than that #3 is probably not workable (producers would just make up a value in a lot of cases).
     
    John
     






  • 2.  Re: [EXT] [cti-stix] Location, latitude/longitude, and precision

    Posted 07-19-2017 04:51



    Since we do not yet really know what we need here, I would propose that we add this field at a later time. I fully understand the problem we think we are trying to solve, but I am not sure we have identified the best solution. It would be nice if we had
    a bunch of real-world problems in products that were driving this.


    Further, if you are uncertain about a location you should NOT be using lat/long to represent the location. You should be using one of the other properties like region, country, city, administrative_area, etc. Those are the named
    non-precise ways of doing location.


    Bret 

    Sent from my iPhone

    On Jul 18, 2017, at 11:33 PM, Wunder, John A. < jwunder@mitre.org > wrote:









    Hey everyone,
     
    We’ve been having a conversation about the best way to represent the precision of a given latitude and longitude pair. What I mean by that is, if I give you 45.234234 latitude by 45.2342 longitude (literally
    just typed random numbers there) is that describing an exact point, just pointing to a city center, or even just pointing to a country?
     
    We have a couple options here:
     

    Just define in normative text that, for STIX, latitude and longitude is approximate. This would mean that you couldn’t really rely on it to be accurate
    to an address, even in cases where it was. Have an optional property to represent the precision. If the property is present, it describes the precision in terms of meters. If absent, we have two
    options:

    Precision is unspecified, and consumers can take that however they want Default precision to perhaps 10km (roughly a zip code), so consumers can pretend the value was that
    Have a required precision property
     
    We also talked about significant digits but due to the ways JSON parsers read data it would mean using a string and likely that data would also be lost in those cases too (and even if there it’s harder to
    get to). It didn’t seem like a great option compared to any of the above, but maybe we’re missing something.
     
    So, what do people think? Any other arguments for one or the other?
     
    I don’t have any strong opinions really other than that #3 is probably not workable (producers would just make up a value in a lot of cases).
     
    John