netCDF says all subscript ranges are 1:n for Fortran, 0:n-1 for C.
Let us consider a minor extension to specify the logical subscript range
in the file. I am doing this via an attribute consisting of pairs of integers,
but maybe there is some legitimacy to be recognized.
Thirty years ago, Algol's array limits had both lower and upper limits, as in
1982:1987 for yearly stock market data. Derivative languages like Pascal
and Modula also. Even fortran 77 grudgingly allows phi(0:n) for my
electric potentials defined in the physical interval (0,n*dx),
and (-m+1:m) for my Fourier coefficients --just like the sum limits in the
equations in the reports. Why not? (Not in C? Yeah, well that's C)
All I'm suggesting is that the file legitimize the "self-description" that
the natural subscript range not start necessarily at 0 (C) and 1 (Fortran).
Actually, writing this reminded me that netCDF doesn't recognize in the
self-description the existence of complex-valued quantities, another concept
popular for some time.
------
I am very encouraged by the development and availability of netCDF and HDF,
and building on both that is happening.
