Re: [h5md-user] units module

[Pierre de Buyl]

On Sat, Oct 19, 2013 at 01:28:05PM -0400, Peter Colberg wrote:
> On Sat, Oct 19, 2013 at 09:44:50AM +0200, Konrad Hinsen wrote:
> > I am all for Unicode in human-readable text, but does it really
> > provide an advantage for a binary format like H5MD, which is always
> > handled by specialized code? Unicode handling is still a hassle in
> > many situations, so it needs to be justified by concrete advantages.
> 
> The short symbols are nice for display. The long identifiers require a
> reader to use the udunits2 library, or at least a self-written lookup
> table of long identifiers and short symbols.
> 
> Since ASCII is a subset of UTF-8, a writer can set the encoding to
> UTF-8 and write ASCII units exclusively. It is only when Å is used
> that unicode support is required.
> 
> My suggestion would be to support one <symbol> for each unit in
> 
>   udunits2-accepted.xml
>   udunits2-base.xml
>   udunits2-derived.xml
>   udunits2-prefixes.xml
> 
> Alternatively, the units module could define one <name> for each unit.

Alter-alternatively, the base "H5MD" could use ASCII only and the specific
"units" module could request UTF8. This means that the choice of UTF8 would done
knowingly by the creator of the file.

P