Re: [bufrtables] More on table versions

Hi Milan:

Some comments are below.

Milan Dragosavac wrote:
Hi John,

The check some can be done only on particular format of the tables.
Almost  every bufr software ( centre) got different format of bufr
tables although
the content shall be the same in the case of standard WMO table entries.
If local entries are added, the tables would be different anyway from
centre to centre.

yes, you are right that it would require the use of a standard format. I think 
that this could be a good thing, even if centers decide to use a special format 
internally, it would be helpful to be able to translate that to a standard 
format for comparison.

 Therefore checksum on tables would not work and really is not needed.
Ecmwf is using bufr for more than 20 years and we have not had a case that
we can not decode historical data because of tables.

i understand that ecmwf can decode the messages that ecmwf has written, since you have been careful to preserve all the tables that you need. But perhaps an outside decoder (such as the one i am trying to write) might have more trouble being sure that it has the correct tables?
If incompatible changes are introduced in new table versions in the future, and 
coders are not being careful about what version they put into the bufr record, 
then this may also cause problems. still, if they are careless about the 
version, they could be careless about the checksum also, so this solution is 
not foolproof either.

thanks for your time in thinking about these issues. BTW, I have been cc'ing 
the email group so others can think about them also, and add their thoughts if 
they like.

Ciao,
John


Best regards

Milan
Milan Dragosavac