This is a (probably pointless) discussion on the 100 meter uncertainty number that is found in -framework. It is a realistic number today, and a much smaller number is not, in my opinion, realistic.
For emergency calls, you care about getting an accurate routing in around a minute (fastest response time practical), and 2 minutes is probably okay. The U.S. Regulation is based on average results for emergency calls within a county (changed from average within a PSAP service boundary). The requirement is that EVERY county get 100 meter uncertainty, 95% confidence. As a practical matter, they drive counties to determine the results, rather than attempt to actually locate callers. They use real cell phones. I'm a bit fuzzy on when they get the accurate location, but the cell system determines this (a poll from the PSAP doesn't get a better result until the network is ready to give good results). I do think it's a minute or two. Brian On 3/10/10 9:28 AM, "Henning Schulzrinne" <[email protected]> wrote: > > On Mar 10, 2010, at 8:36 AM, Brian Rosen wrote: > >> I suppose that it's conceivable that every carrier, the FCC, NENA and APCO >> are all stupid. > > The 100 m/95% is a requirement, not an attempt to describe current GPS > performance. As far as I know, the requirement was stated when GPS SA > (selective availability) was still enabled. > > >> >> On the other hand, the carriers spend millions of dollars testing every >> year. So, perhaps they aren't quite so stupid. > > I'm afraid I'm missing how this relates to my note. Did I accuse anybody of > being stupid? > >> >> I don't have access to the full paper to understand what a "20 minute field >> test" is, but the largest problem for 100 meter 95% confidence is the fact >> that the GPS is not always enabled (for battery drain reasons) and unless >> it's on and seeing satellites for many minutes, you don't get really good >> accuracy. Also, field tests are usually conducted in open sky situations, >> and not urban canyons, or even tree leaf cover situations, which degrade >> performance considerably. > > > As an aside, this illustrates that the 95% number is pretty meaningless. Does > it mean 95% of all calls (lots more emergency calls are made from indoors and > the urban canyons of Manhattan than from open fields) or 95% of all sample > points? Is it measured immediately after turning on the device or after > waiting N minutes? > > Without such detail, making precise statements is just pretend-precision. > > Henning _______________________________________________ Gen-art mailing list [email protected] https://www.ietf.org/mailman/listinfo/gen-art
