[MITgcm-devel] this weeks share of stupid suggestions about netcdf

[Martin Losch]

Hi there,
I have a few questions/comments/suggestions about netcdf:
1.There is no reason for not having an iteration_count (really output 
count) in state , but the coordinate variable T(T) does not seem the 
right place for this. Instead T(T) should hold the model time (which is 
now stored in model_time). What do you think?

2. another nice thing to have would be the attribute missing_value for 
record variables. This attribute could be 0 for a start (although some 
"special value" such as -9999 or 1e23 makes more sense). Applications 
such as ncview or ferret recognize this attribute and automatically 
replace all "missing_value"s in the record variables with some internal 
special value for displaying ( a comfortable way to have the land masks 
included ). Is that difficult to implement? Is that something that is 
of interest? In which routine are the attributes declared and filled?

Martin