Re: NetCDF Perl documentation

> From: address@hidden (Steve Diggs)
> Subject: NetCDF Perl documentation
> To: address@hidden
> Date: Sun, 21 Nov 1999 19:13:13 -0800 (PST)
> Keywords: 199911220310.UAA14898

Hi Steve,

> We actually met years ago during the CEPEX experiment when I was visiting
> Dick Dirks about logistics in Fiji.  I've since moved on, more liked moved
> around SIO, but I've remained a loyal fan of NetCDF.  In the last 6 years,
> I've also become a big fan of Perl, since it has unparalleled string
> processing capabilities and lots of other nice features.

Yes, I remember meeting you.

> Recently, I've convinced the powers that be on my international data project
> (WOCE) to convert and exchange data in NetCDF.  So far, so good. 
> Since the software group that I manage uses Perl almost exclusively, we
> decided to make all of our NetCDF code use the NetCDF/Perl interface.
> 
> Well, to put it lightly, it's been a struggle.  Robb Kambic has been quite
> helpful, but I feel as though I'm taking up a lot of his time asking dumb
> questions that could have been answered much more efficiently if I only had
> access to a comprehensive NetCDF/Perl manual.  I have to keep asking, "how
> does one do blah, blah....".

I'm CC:ing Steve Emmerson on this reply, since he developed and
maintains the netCDF Perl interface.  But I have to take most of the
responsibility for not assigning more resources to continuing
development of the Perl interface, since I didn't think it was getting
used enough to justify a lot of time when we were short on programming
for other projects.

> When I was using C to do this, the support was great and the documentation
> plentiful.  Imagine trying to get users to try to use NetCDF in C or
> FORTRAN w/o extensive documentation with examples!  You may think that
> there aren't a lot of people using your NetCDF/Perl interface, but that may
> be because they tried and couldn't make heads or tails of it.  BTW, there's 
> a quote in the man page that really gives me heartburn:
> 
> "     In addition to the above functions, most C  macro  constants
>      that are defined in the netCDF header file netcdf.h are also
>      available to a perl script by dropping any  `NC_'  substring
>      and using the NetCDF:: prefix, e.g. NetCDF::LONG."
> 
> Not NetCDF 3! 

Thanks for pointing out that problem.  I think the man page should at
least be updated to be consistent with the current netCDF 3
documentation (which also needs work, by the way).

> Well, you can't just use the docs for the C interface since the function
> names are different!  I found out (on my own) that one must refer to the
> obsolete NetCDF 2 (1996) docs.  Why isn't Perl getting the attention that
> FORTRAN and C are?  Has it really been three years since the Perl interface
> has had any attention?  I would switch to C, but trust me, it would consume
> a lot more resources here and would be very inefficient for us.

It may be a chicken and egg problem; if we devoted more resources to
the Perl interface, there would be more users, which would justify the
effort.  But the bottom line is that we are being funded by the NSF to
do the work we proposed for atmospheric science departments, which
means developing a suite of Java applications for meteorology and
enhancing the Internet Data Distribution system are currently the
highest priorities.  We're also being stretched by demands for a
better Fortran 90 netCDF interface and collaborating on a merger with
HDF-5, among other things.  But now that we know the current state of
the Perl interface is a problem, I'll put looking into upgrading it on
our "to do" list.  And we have no plans to drop it, but would gladly
adopt an updated version if someone out there in the user community
developed and donated a new netCDF-3 compatible Perl interface.

> I don't mean to complain incessantly, but I'm finding the going hard and
> slow developing my software in Perl to use NetCDF.  Usually, this indicates
> that a particular interface will soon be dropped from official support due
> to lack of interest, and that would be a shame, for there are a lot of Perl
> users out there (I teach a class for UCSD extension and they're *always*
> sold out) that use the language in a lot of science-related applications.
> Many more than, say, Java.
> 
> Please reply, I'd like to know your thoughts on this subject.  I still
> remain a fan of NetCDF.

Thanks for the feedback, both positive and negative; it's useful.
You're probably right that there are currently more Perl users than
Java users for science-related applications, but I think the opposite
will be true in 2 or 3 years.  And our development priorities are set
by our User and Policy Committees, who have both recently re-endorsed
the Java direction.

--Russ

_____________________________________________________________________

Russ Rew                                         UCAR Unidata Program
address@hidden                     http://www.unidata.ucar.edu