PIDF-LO Profile

1
PIDF-LO Profile

• James Winterbottom,
• Martin Thomson
• IETF-65

2
Whats Changed?

• Updated the rules based on comments to make them
  less ambiguous.
• Removed much of the geospatial section and made
  extensive references to the GeoShape draft.
• Modified examples to be in line with Revised
  Civic draft and GeoShape draft.

3
Issues

• Recommend a restriction of polygon to 16 points
  when transporting own location.
• Do we want to recommend shape types for emergency
  calling?

4
LCI and Uncertainty
5
LCI Uncertainty

• In LCI (RFC 3825) defines a resolution that
  describes a region of uncertainty.
• Appendix A of PIDF-LO Profile describes how to
  get a PIDF-LO from LCI.
• Resolution indicates a number of bits that can be
  considered valid.

6
Converting from LCI

• (From RFC 3825) 38.89868 is
• 000100110.1110011000001111111001000
• A resolution of 18 indicates a range using the
  first 18 bits
• 000100110.111001100xxxxxxxxxxxxxxxx
• x bits can have any value from all 0s to all 1s
• 000100110.1110011000000000000000000 to
  000100110.1110011001111111111111111
• 38.8984375 to 38.9003905951976776123046875

7
Converting to LCI

• (RFC 3825 doesnt include any example of this)
• When converting a value with uncertainty to LCI,
  start with the maximum and minimum values
• 32.98004 to 32.98054397 (56 metre range)
• 000100000.1111101011100011111001110 to
  000100000.1111101100000100111011100
• Resolution is the number of identical bits
• 000100000.1111101, which is 16 bits
• Converting back to check accuracy
• 32.9765625 to 32.9843745 (870 metre range)

8
That wasnt a contrived example

• That was a randomly selected point.
• This gets much worse for certain border cases
• 31.9999985 to 32.00000274 (0.5 metres)
• 000011111.1111111111111111111001110 to
  000100000.0000000000000000001011100
• The resulting resolution is only 3 bits!
• Conversion gives a range from 0 to
  31.9999999701976776123046875
• The error has increased to 3,500,000 metres!

9
Conclusion

• LCI encoding has a major flaw.
• This flaw causes it to break down near certain
  boundaries.
• This problem can occur all over the world
  wherever the minimum and maximum values for a
  location differ in too many bits.
• It doesnt matter what the original precision
  was!
• A location with uncertainty that spans the
  Greenwich Meridian or the Equator cannot be
  represented
• Resolution 0 bits!
• Resolution is effectively useless.

10
For my next trick

• I will demonstrate how important uncertainty is.
• and why it should not be ignored.

11
Where is the Target Zone A or B?

• This is easy its clearly in Zone A, right?

12
Where is the Target Zone A or B?

• If you include uncertainty, the answer is less
  clear.
• To be certain, the overlap with Zone A must be
  compared with the overlap with Zone B.

13
Where is the Target Zone A or B?

• The amount of uncertainty determines which zone
  the Target is most likely to be in.
• The bigger the uncertainty, the more likely the
  Target is in Zone B.