Re: [bufrtables] More on table versions

Hi Milan:

Thanks for your thoughts on this. One of the interesting questions is whether software 
can automatically detect if the wrong table is being used to decode. When the bit widths 
are wrong, the answer is usually yes, since one can "count bits" and compare to 
the data section length.

However, if scale/offset/units is wrong, then one must actually understand the 
data in order to detect problems, and even then it way not get detected.

Something that perhaps you have thought of before is to embed something like an 
MD5 checksum into the BUFR message, to be compared against the MD5 of the 
master table stored at the WMO. Obviously a lot of trouble, and even then not 
foolproof, but perhaps worth considering, especially for archived data.

Regards,
John

Milan Dragosavac wrote:
Hi John,

I believe you might find almost any master table version number in the
data. In practice in the section 1 you will find local version numbers use as well although the local entries are not really used. In any case you can always first try making link to version 13 and find out how it goes. I am sure it will work in almost all cases. I will try to address
this problem at WMO ET/DRC Meetings.

Best regards

Milan

Milan Dragosavac

ECMWF
Shinfield Park, Reading, Berkshire, RG2 9AX, UK

Tel: (+44 118) 949 9403
Fax: (+44 118) 986 9450
Telex: 847908 ECMWF G
E-mail: milan.dragosavac@xxxxxxxxx