Version numbering (was: citing IVOA standards)

[Norman Gray]

Greetings,

On 2008 Jun 17, at 00:41, Séverin Gaudet wrote:

> "Standards numbering nomenclature" (http://www.ivoa.net/cgi-bin/twiki/bin/view/IVOA/TCGStandardsNumberingNomenclature 
> ).


The proposal here strikes me as being more intricate than is necessary.

The role of version numbers is (a) to distinguish versions from each  
other unambiguously, and (b) to indicate _some_ functional  
relationships between versions.  In the case of documents, these  
relationships are much simpler than they are in the case of software,  
basically limited to one document superseding its predecessor.

To me, the proposed scheme is confusing, because it tries to encode  
the document's status (as WD, RFC and so on) in the version number,  
and does so by implying that a document with version number 0.9 is  
somehow 90% of the way through its process (version numbers are not  
real numbers).

What certainly is _very_ confusing about the current versioning scheme  
<http://www.ivoa.net/Documents/Notes/DocStd/Procedures-20040425.html#WorkingDrafts 
 > is that the pre-REC and post-REC version numbers are not  
continuous: thus WD-1.0 is a different document from REC-1.0.

I can't find any cast-iron guidance on versioning at w3.org  
(surprisingly), but in the W3 WD I'm currently most familiar with (the  
SKOS one), they appear to be using date as their only pre-REC version  
number.

So, I propose:

   * All Recommendations have a version number n.m
   * All pre-Recommendation documents, including internal drafts,  
formal working drafts and proposed recommendations, have a version  
number yyyymmdd, with the status indicated by a prefix.

Thus for a given document, we will have DRAFT-20080101, WD-20080201,  
PR-20080301 and either REC-1.0 or plain '1.0'.  That scheme is  
monotonic (so satisfies properties (a) and (b)), and makes the  
document's status explicit rather than implicit.

Since this thread started with Igor's remarks about ADS bibcodes, I  
suggest that RECs have 'n.m.' as their PPPP and an explicitly noted  
four-character code for their VVVV; I don't expect ADS would be  
interested in giving codes to any pre-REC documents.

Best wishes,

Norman


-- 
Norman Gray  :  http://nxg.me.uk
Physics and Astronomy, University of Leicester