Re: [h5md-user] H5MD based on HDF5 1.8 file format

[Pierre de Buyl]

On Wed, Jul 17, 2013 at 11:45:11AM -0400, Peter Colberg wrote:
> On Sat, Jun 01, 2013 at 09:53:44PM +0200, Pierre de Buyl wrote:
> > On Fri, May 31, 2013 at 01:37:57PM +0200, Felix Höfling wrote:
> > > I see no problem in requiring HDF5 version 1.8 for the feature of
> > > object tracking. In the rare case that someone wants to write H5MD
> > > using HDF5 version 1.6, this feature is not available. Since all
> > > other H5MD specifications work also with 1.6, I'm a bit hesitating
> > > to make 1.8 compulsory.
> > 
> > I like the idea but honestly, we may just mention on the webpage that HDF5 
> > 1.8
> > and later is suggested and not make it mandatory. If we want to impose that
> > restriction later on, it is a very easy one to add to the specification.
> 
> Alright, then the use of the HDF5 1.8 file format shall be optional.
> 
> I would like to make the following change:
> 
> The h5md/creation_time attribute is dropped.

It seems very reasonable to me to make use of HDF5 features instead of supplying
our own. Letting HDF5 handling the date and time should be less error prone,
too.

> If a user wishes to track creation and modification times of the file
> and its objects, the H5MD file shall be written in the HDF5 1.8 file
> format. The file format corresponds to the H5F_LIBVER_18 constant of
> the HDF5 API.

This should then be mentioned in "General organization", for instance.

Regards,

Pierre