[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: HDF data types on UNIX and windows

H C Pumphrey (hcp@newsread.ed.ac.uk) writes:

> And that therefore tallies with the message you got from IDL. Maybe
> only a small fraction of HDF files have this data type in and support for
> it is therefore unreliable in many packages.
> I have tried a few HDF viewers and readers and found that most are flaky. 
> They tend to support the subset of HDF files that the writer uses them on, 
> and fail on others. (I had assumed that IDL's HDF support was better than 
> this, maybe I was not quite right in that assumtion.)

I offer an alternative hypothesis: the HDF documentation
as it comes with the HDF libraries is atrocious.

Readers are good for data the writer uses because the
writer has spent hour upon hour figuring out what works
empirically, not with what works according to the lousy
documentation. And when you find a discrepancy, what do
you do? Implement a work-around? Assume the documentation
is wrong? Suggest it is the library writer's fault?

Packages like IDL have to be just slightly behind the
leading edge simply to have enough people testing the
implementation to make *some* kind of implementation
possible. The surprise to me is that HDF readers are
as good as they are!

> This in turn suggests to me that HDF is too big and 
> complex. I get the impression that HDF5 is intended to cut exactly
> this Gordian[1] knot.

Heaven help us. :-(


David Fanning, Ph.D.
Fanning Software Consulting
Phone: 970-221-0438 E-Mail: davidf@dfanning.com
Coyote's Guide to IDL Programming: http://www.dfanning.com/
Toll-Free IDL Book Orders: 1-888-461-0155