Re: [h5md-user] Finalization of units module

[Felix Höfling]

Am 09.01.2014, 09:47 Uhr, schrieb Konrad Hinsen
<address@hidden>:

> Felix Höfling writes:
>
>  > Thinking about the "system" attribute again during the vacation I'm a  
> bit
>  > unsure whether it is really needed. In particular given the fact that  
> it
>  > can (and must) take a single value only.
>  >
>  > Extending the set of units could be achieved by simply increasing the
>  > version number of the units module.
>
> The underlying question is: do we want a single unit system, able to grow
> as new needs are identified, or do we envisage distinct unit systems
> that are completely independent?
>
> I'd go for the second approach, and keep the "system" attribute, even
> if initially only "SI" is defined. The major alternative unit systems
> for which I see a need are domain-specific unit systems defined for
> theoretical models. A simple example is "reduced LJ units" but there
> are more elaborate ones. I see them arriving in coarse-grained models
> for macromolecules. These unit systems have no reference at all to the
> real world (and thus to SI), and may use arbitrary symbols.
>
> Finally, a minor point but I think it's worth considering. The SI
> system defines a fixed set of units and prefixes, plus some tolerated
> units that don't fit the system but are convenient, such as liters for
> volume.  A major goal of SI is simplification through standardization
> and thus the elimination of older units.  A unit system designed to
> include lots of other units as long as no name conflict arises should
> not be called "SI".
>
> Konrad.

With respect to abstract unit systems such as "LJ units" I agree, they can
and should not be combined with physical units.

Concerning the second point: I think that H5MD should just reflect the
common usage of units and not enforce the use of SI. Hence, I would
like to keep the list of unit symbols as it is. To avoid confusion, we may
choose "physical" rather than "SI" for the system attribute, in contrast
to e.g., "LJ".

Felix