Re: [cf-pointobsconvention] Draft 2


I was lucky to hear Andrew Woolf talk about CSML a couple of weeks ago in Boulder. I really liked his talk and think I understand the benefits of keeping data in netCDF. I am thinking about the representation of the spatial features themselves in the netCDF file and thinking that good things might happen if we thought about the general class of geographic features as a single class rather than as separate points, lines, ... In that case, we already have a well established and widely used standard for storing that information.

The reason i said that the connection to spatial databases was straightforward is because the well-known binary representation of the features is actually the native binary representation of features in PostGIS, which is a database that I expect many in the netCDF crowd would be interested in using. In addition to having the same binary representation, all spatial databases understand the simple features spec which would be easy to support if we used WKT inside of the netCDF or HDF5 files.

I think that CSML would have an important role as a transport or as essentially an XML wrapper around these netCDF files, but I don't understand it completely.


On Oct 12, 2007, at 9:57 AM, Bryan Lawrence wrote:

Well, it's not that ridiculously straightforward :-(

CSML: The climate sciences *modelling* language, is the first (as far as we know) attempt to build a *model* of climate science features ("beyond" lines polygons etc, but "less than" grids) which are consistent with the geospatial
communities ideas (especially as expressed in GML, but based on the
underlying ISO and OGC standards) see ...

But, having done that, we still have to store the underlying binary data, and for that we prefer netCDF, and that does bring us back to initiatives like this. I don't think we can escape these thought processes :-) ... which will
then feedback to iniatives like CSML and on a longer time scale, the
harmonisation of the Unidata SDM, CSML and OGC fundamentals ...