VOUnits RFC

[Norman Gray]

Greetings, all.

The VOUnits recommendation process trundles on...

On 2012 Aug 17, at 09:57, Norman Gray <norman at astro.gla.ac.uk> wrote:

> This message announces the start of the VOUnits RFC, the page for which can be found at <http://wiki.ivoa.net/twiki/bin/view/IVOA/VOUnitsRFC>.

We have had a good collection of comments on this RFC page (though from only two of the 12 working groups -- thank you, Applications and Data Modellers!), and I've received some comments by email and at the Heidelberg interop.

After consultation with Sébastien (the original author of the VOUnits document) and Mireille (the Semantics vice-chair), I've released another version of this document, PR 1.0-20130724.  This should appear at <http://www.ivoa.net/documents/VOUnits/> before long, but at the time of writing, the version there is still 1.0-20130429.  For the moment, therefore, use instead the version attached to the wiki page.

I haven't taken all of everyone's advice on this, but where I haven't agreed with someone, I believe I've added rationale to the document which should explain why, at least by implication.  I'm very happy to be disagreed with.

I don't know what precise stage we're in, procedurally, since the RFC period is long over, but we've had an incomplete set of approvals, and I've just released an edited version of the document!  @Severin: do you have an opinion here?

----

I've updated the RFC page with some specific responses.  Some points need a little more elaboration here.

** Innovation

The VOUnits Recommendation still intends _not_ to be innovative, in the sense that it is intended to indicate the _intersection_ of the existing syntaxes for unit strings, as much as possible.  Thus we have aimed not to add features.  

In particular (despite some comments on the RFC page and off-line) we have not extended the syntax of decimal numbers.  In fact, the only place where such a number would appear in this specification is in the form of a numerical scaling factor before a unit (for example '0.1nm', indicating that the Angstrom is the _unit_, as opposed to _quantity_): we restrict such scaling factors to round powers of ten, and in any case expect these to be rather rare.

Similarly, the desire to aim for the intersection of the pre-existing grammars has more-or-less forced us towards '.' for multiplication and '**' for powers.  No, I don't think '**' looks pretty, either.

Since we have included a list of 'known units' which is mostly the _union_ of the known units in the various prior syntaxes,  and since we suggest that parsers should be able to cope with unknown units, we have felt free to be add a few units to the list of 'known units' for VOUnits.

** Binary prefixes

The document still insists that the SI prefixes always refer to powers of 10, and do not switch to indicating powers of 2^10 when applied to bits and bytes.  The document now provides more rationale and references, and includes a list of the binary prefixes.  There has been a fair amount of discussion and changing-of-minds, in all directions, on this point.  We hope that the result, in the document, is at least explicitly-motivated and self-consistent.

** A units data model

We're steering well clear of that!

There has been a certain amount of mild perplexity that, in all the years of the IVOA, there's been no attention to what surely should be the obvious low-hanging interoperability fruit of a consensus on unit strings.  This may be because everyone has thought the question either too obvious or too boring to give head-room to.  Alternatively, it may be that this discussion notoriously and promptly turns into a megathread on metrological arcana.

If this document, and the associated library, create a modicum of consistency, that would be a good thing.


** Structure of the document

Because of some reordering of the text, and expansion of tables, the document may look more different from the previous version, than it actually is.

Any and all comments most welcome.

----

Version 0.8 of the Unity library <https://bitbucket.org/nxg/unity/> is up-to-date with respect to these changes.

I suggest that any discussion happens on the Semantics list rather than the Interop one.

Best wishes,

Norman


-- 
Norman Gray  :  http://nxg.me.uk
SUPA School of Physics and Astronomy, University of Glasgow, UK