Im GIS Monitor, April 29, 2004 from GITC America bin ich auf folgenden Beitrag zu GML gestossen, welcher Erfahrungen von Safe Software (Anbieter von FME) beschreibt:
"Don Murray, the founder and president of Safe Software, explained his experiences with Geography Markup Language, GML, an XML encoding for geospatial data. In short, he argued there are two roles assigned to GML. One is to be an "expressive" tool for complexly modeled data. That is, it can capture and distribute all the complexity of the Ordnance Survey's MasterMap in a vendor-neutral way. From his account, it works great for that type of use. The second use is to deliver a solution to the request, "Hey Fred, can you get me some data about this area of town?" For this type of "quick sharing across GIS packages" GML has not fared so well. That was his conclusion based on experience from a recent GML Relay where data ideally was to be loaded into and edited in a software package then passed to the next participant to do the same. It didn't happen as was hoped, to put it gently.
Still, the Relay experience was valuable as were some of Safe Software's implementations for customers. Murray noted that he and his colleagues have learned that GML:
-is not a database
-is slow to use as is but great to load once into a database
-is built for expressiveness
-typically needs an external file to ensure effective understanding of that expressiveness
-that the external file sometimes needs to be "handcooked"
-is effective in a controlled environment
The challenge then is to explore how GML can better serve the quick data sharing needs of geospatial (and other) users. And, the answer seems to involve some type of profiling that will simplify the process. A profile basically means "rules about what types of things go into" GML. As Murray put it, one part of a profile might be "Thou shalt not have embedded objects" (object made up of other objects). Until one or more profiles are defined and agreed upon, most users will rely on the "old faithful" ways of quick data transfer, via shape files or MIF/MID. For GML to take hold, it needs to be at least as easy and effective as these methods. The example Murray used that sticks with me: imagine if when you tried to load a Web page you had to "spend a morning" deciphering its language. The Internet would never work. That's akin to the current state of GML for data exchange."
Für diejenigen, die GIS Monitor noch nicht kennen: [
www.gismonitor.com/]