[thredds] Fwd: Re: Scan Aggregation vs. Feature Collection

• To: THREDDS community <thredds@xxxxxxxxxxxxxxxx>
• Subject: [thredds] Fwd: Re: Scan Aggregation vs. Feature Collection
• From: John Caron <caron@xxxxxxxxxxxxxxxx>
• Date: Tue, 24 Jul 2012 08:10:34 -0600

Some typos corrected:

Well, its complicated.

Old style aggregations, aka "NcML aggegations" remain supported. I think
of these as low level index manipulations. These are embedded in NcML,
either on the client or in the TDS config catalog, see:

  http://www.unidata.ucar.edu/projects/THREDDS/tech/tds4.3/tutorial/NcML.htm

We can do much better when we know the "feature type", and the new way
is called "feature collections":

1) type=FMRC is for gridded data. Works mostly ok in 4.2, but refactored
in 4.3 and still testing. However, usually problematic for GRIB files.

2) type=GRIB is for GRIB data, which is a special kind of gridded data.
This supersedes FMRC when the data is in GRIB. Only in 4.3, ready for
beta test.

3) type=Point are for point feature types. Still in alpha in 4.3

NcML works fine when you know what you are doing and you have homogenous
files. feature collections are intended to be easier to use and tolerate
heterogeneous files.

John

• References:
  • Re: [thredds] Scan Aggregation vs. Feature Collection
    • From: John Caron