[email protected]
[Top] [All Lists]

Re: [Ecrit] LoST

Subject: Re: [Ecrit] LoST
From: Henning Schulzrinne
Date: Wed, 11 Jul 2007 08:53:49 -0400
Three quick comments:

As far as I know, GPS devices produce a point plus (small) uncertainty, e.g., in the common NMEA output format (http://vancouver- webpages.com/peter/nmeafaq.txt). The uncertainty is so small and tends to fluctuate so much that including it in location mapping is pretty pointless, and it's probably of dubious value even for dispatch. (When I went geocaching recently with my kids, uncertainty on a modern SiRF-based device was around 18 - 25', under dense tree cover.)

That said, I think the complexity of including circles, arc band and ellipse is minimal. In all cases, the description includes a center point, which the LoST machinery could easily use to do the actual lookup. Thus, the additional complexity is mainly parsing these XML structures, which strikes me as not a big deal.

What gets complex is if a LoST resolver is forced to map shapes to multiple branches of the tree, what we have discussed in the past as the "forking" problem. For example, if a geo shape straddles the boundary between the US and Canada, you really don't want the LoST resolver gather results from both trees, decide how long it needs to wait for one or the other, and other issues. This requires significantly more protocol work, creates additional error conditions and increases implementation complexity.

A third point relates more to HELD than LoST: It's probably a good idea to be able to request a specific location profile, in the same spirit as in LoST, so that if newer, complex shapes can be accommodated easily in the future. We can then decide whether to define a LoST-like profile for HELD that only includes the current geo shapes, without endangering interoperability.

In summary: I'm not opposed to including the three additional shapes, as long as it is clear that the LoST server MAY use the center point for mapping.


On Jul 11, 2007, at 6:58 AM, Hannes Tschofenig wrote:

For GPS usage at the end host only you are certainly right that circles & ellipse is a good choice. The big question is here:
* is it a big issue to translate them to a point
* does it introduce big error rates
Please also note that we are talking about emergency call routing in the context of LoST and not about dispatch. To approximate a circle with a radius of 3 meters with a point might not be a big problem.

Ecrit mailing list
[email protected]

<Prev in Thread] Current Thread [Next in Thread>