[email protected]
[Top] [All Lists]

Re: [Ecrit] Profiles (was: Consensus Call) - hole

Subject: Re: [Ecrit] Profiles (was: Consensus Call) - hole
From: Henning Schulzrinne
Date: Tue, 24 Apr 2007 21:30:19 -0400

Also, there's the problem I brought up yesterday. If a server sees this input, where a polygon has a whole in it:

<gml:Polygon srsName="urn:ogc:def:crs:EPSG::4326">

Now, is the server to find the centroid of the polygon or is it suppose to look for an exact match of the polygon?

That seems like a completely different question compared to the profile issue, or at least I'm missing the connection.

This is indeed a real problem, but only if the location of the querier itself has a hole. There is no problem for the case where the service boundary has one or more holes. The location-with-hole case only occurs in the location hiding case, as far as I can tell, and is relatively uncommon even there.

There are two solutions, neither of which is totally satisfying.

(1) Exact match, as you indicate. In other words, a LoST server first tries to find an exact match for the boundary. For the location- hiding case, this will always work.

Advantage: The client doesn't have to know.
Disadvantage: LoST servers need a special case.

(2) Transformation in LIS: The carrier that wants to hide a location creates a convex location region that has no holes and includes the location of the UA. A very simple way to do this is to return a triangle (square, circle, ...) that includes the caller's location (not in the center) and whose centroid falls into the PSAP service area. This has to be done only once (unless the PSAP boundary changes), so the computational burden is trivial.

Advantage: No special case in LoST servers.
Disadvantage: Requires some special geometric treatment for special cases.

I have some preference for (2), since it's a bit cleaner and it's conceptually no different from a wireless location system with a very imprecise location technology, such as cell-face only. It requires no standardization as such, as long as we standardize the centroid behavior.


Ecrit mailing list
[email protected]

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