1. Definitions

A "bufr file" is a collection of bufr records.
A "bufr file index" is an index for the bufr file, used by the netcdf-java-2.2 
library (nj22). 

If the bufr file has a filename or URL "bufrURL", its index file has the name 
"bufrURL.bfx" . The nj22 library looks for it in the same location as the bufr 
file, and when possible, creates it in the same location if it doesn't exist. 
It may create it in other locations if it can't create it in the same location 
as the bufr file.

The bufr file index has enough information so that nj22 can create the netcdf 
objects that represent the bufr file, without having to scan the entire bufr file. 


2. Bufr file index format

name = value of the attribute
name must be netcdf name (no blanks etc)
value is any string, do not quote

------------------------------------------------------------
Section 1 Global attributes
------------------------------------------------------------
type = BUFR       // data type
bufr_edition = 3  // edition of encoding
location = data/PROFILER_3.bufr
length = 1115527  // file size
created = 2006-05-19T14:25:34Z  // date index create
index_version = 1.1  // version of this index
header = IUPT43 KBOU 240000  // product header of first record
reference_time = 2006-03-24T00:00:00Z  // reference time for daily file
center_id = 59 The NOAA Forecast Systems Laboratory
sub_center_id = 0 WMO Secretariat
category = 2 Vertical soundings (other than satellite)
table = B3M-000-009 // table used for decoding data
compressdata = false // type of raw data, determines field data arrangement

------------------------------------------------------------
Section 2 Bufr records
------------------------------------------------------------
one line for each observation in the file, blank separated fields. It is possible
that one line could represent a set of obs if the data is in compressed format.
Point/Station ISODate DataDimension Latitude Longitude Altitude DDSoffset
ObsOffset BitPos BitBuf

WSMN5 2006-03-24T00:00:00Z 37 32.4 -106.34 1224 1408 1462 0 0

-DataDimension  size of the larger fields. Not necessary regular for each ob
-DDSoffset  offset in file for Data Descriptors used to interpret the data
-ObsOffset  offset in file for the start of Observation
-BitPos since obs don't necessary start on a byte boundary, position in byte
-BitBuf since obs don't necessary start on a byte boundary, starting byte
-BitBuf is equal to the number of obs if this is a compressdata set because
they always start on a byte boundaries.
------------------------------------------------------------
Section 3 Parameters for each Observation

one line for each parameter in the observation
"Descriptor Key" "Max Dimension of Field" "type of Data" "Field Name" Units

0-1-2   1       true    WMO_Stn_number

-type of Data  equals true if numeric otherwise false equals String data.