Re: [ferret_users] MULTI-FILE NETCDF DATA SET

[Ryo Furue <furue@xxxxxxxxxx>]

Hi Steve,

| Thanks for your comments.  We (the Ferret developers) understand the
| situation and agree with your points.  For several years we have been 
| encouraging Ferret users of netCDF to solve the problems of file 
| aggregation through the more advanced techniques offered by NcML and the 
| Unidata TDS server.  Understandably, there have been many Ferret users 
| for whom the effort of setting up a TDS was simply more than they were 
| willing to undertake.  In order to support those users we have kept 
| alive the Namelist-based technique for file aggregation.  Despite its 
| imperfections with respect to portability, it is effective in most cases.

There seems to me to be some misunderstanding.
I don't have any objections to the namelist-based technique.
The namelist-based technique is portable, as long as the users
use the standard-conforming syntax, which is something like

  &namelistname
     itemone = blahblah,
     itemtwo = foofoo,
  /

That's all I say.

As a footnote, if you use a pure F77 compiler the above may
not work because namelist wasn't in the F77 standard.  In that
sense, namelist isn't perfectly portable.  But, it's harder
and harder to find a pure F77 compiler.  These days, most
Fortran compilers are F90-conforming, meaning in particular
that they support namelist of the syntax above.  In addition,
even so-called F77 compilers tend to support the F90 syntax
of namelist. For example, g77 does.  (Or did.  I don't find
g77 any longer for the OS I use.  It's replaced by gfortran.)

Regards,
Ryo