Re: Options for Groups in netCDF-4

Hi Russ,

> I've begun to think about whether we should make use of HDF5 Groups in
> netCDF-4, and if so how we should make use of them.  I don't have any
> recommendations yet, but here's a short discussion of some of the
> issues:
> 
>   http://my.unidata.ucar.edu/content/staff/russ/groups-rfc.html
> 
> This is a request for comments on the above draft.  I'm leaning
> towards using HDF5 Groups, but haven't decided which of the two models
> described in the above draft to recommend.
    The issues you discuss in your draft make good sense.  We've been planning
on adding an optional "creation order" index to HDF5 groups, which would make
the "Groups as netCDF files" option fairly easy to implement for you, since
most of the work to track the order that objects are created could be done at
the HDF5 level.  With the "Groups with string syntax" or "Hybrid" models it
would require either a more comprehensive indexing model at the HDF5 level, or
more work done at the netCDF level, to track the order the objects are created
through the nc4 API.

    Quincey