[EAS] USVI and PR: polygons versus political boundaries

[Botterell, Arthur@CalOES]

I'm afraid you're confused, Sean. CAP isn't a transport standard, it's a data structure.  (In OSI terms it's mainly a Presentation-level standard, although it does have a few features that could be used at the Session layer if an implementer wanted to.)

CAP is deliberately agnostic as to transport.  The closest thing we have to a de facto transport standard for CAP is the RSS-like polling mechanism used by IPAWS and a number of commercial products, although that's run up against latency problems in some applications.  There's also XEP-0127 for CAP transport over XMPP ("Jabber")... and lately I've been working on CAP transport over other pub/sub transports as well as digital broadcast, but not aimed at a standard, just an application.

Note that CAP can also be encoded in ASN.1 per the spec... and it's commonly transformed to JSON inside applications, although that hasn't been recognized in a standard yet.

You're right that the recommended limit for coordinate in CAP is 5 sig-figs... which translates to precision on the order of a meter.  My example offers quite a bit of false precision, but that's not really a concern in the application.

I'm not sure when systems would need to determine if two polygons were exactly the same.  Certainly there are cases when it needs to be determined whether two polygons (e.g., an alerting target and a cell site or siren coverage contour) intersect, but the most common calculation is whether a given point location lies within or without a specified polygon.  

You're right that having both geometries and labels risks inconsistency, but that was the price we paid for SAME back-compatibility.  Personally I'd love to quit using geocodes entirely, but that's not going to happen for awhile yet.

And certainly anytime one simplifies polygons one has to consider the question of overlaps.  Again, in the PBS map that's a minimal consideration, although I might go back and re-simplify the counties as a group instead of individually some day.

Again, though, if we're looking at automated processing... which we almost always are these days... explicit geometries really are more useful than human-style names.  The folks who are actively working on such things... at NOAA and elsewhere... got over their reluctance to deal with geometries instead of names years ago.

So look, I'm glad you're so interested in this stuff, but I think we've drifted pretty far from the topic of this list.  If you want to go out and build some systems I hope you will, as that's the best way to get educated on this topic.

Art