Mike = first conversion utility ?
Yes

HDF4 tool, fits to hdf4 exists
Fits users input 

Linear objects, no index
Header = they have API functions that deal with it


Use HDF5 table if possible

How organize the file?
Have a main group called FITS (ala netCDF)
Ask for input of several fits files in one HDF5 file

Header = should be attribute ? where ? on the group ?


Usage: more options, select one particular object, use HDF5 compression, other
HDF5 features

Astronomers asked for the tool
1KM telescope array

H5 to fits ? yes

Many fits files in one HDF5 file

Elena = headers should be attribute 
Table 

Quincey = use same type as fits uses in mapping

Raymond= Not use NATIVE types in the conversion

Pick the endianess and the size of the original type, provide options for using NATIVE
Use STD type as first option

Usage = more options

Peter= Reach other communities
Advertise it

Benchmark tests, IO performance
Attributes better for headers

Frank=Naming of objects, _N, preserving the original ordering of objects

Failure in reading ? stop it and message

Kent=should be FITS_IMAGE_N
Header= should be attribute

John= attributes, one for each name, problem of orders, what about the comments ?

Peter= header as a comment


Several options of saving the header . as a string or as different attributes ? ask users .

Quincey= h4toh5 tool. Put the metadata in the hdf5 file, point to the raw data as external dataset
Elena= pros and cons, not able to use all HDF5 features (compression, etc)

Funding ?