Guidelines and Procedures note and version numbers

Martin Hill mchill at dial.pipex.com
Thu Apr 1 05:40:28 PST 2004


Marco C. Leoni wrote:

> - About version numbers
> 
> First of all, the standard process is about document specification, 
> whereas software implementation is something different.

If the full standard was described in the document, I would agree; the 
schemas/wsdls/etc are implementation issues.  However we are using the 
schemas/wsdls to describe the standard, so I see them as being part of 
the specification.  In other words, if I read the SkyNode document, I 
see the 'sorts of things' that an astrogrid datacenter must supply, but 
it is not in sufficient detail for me to create a datacenter that 
conforms closely enough to the standard that it will work with the VO. 
For that I need the WSDL and schemas.

> Having said that, let me recall a discussion Tony, Ray and myself had 
> some time ago:
> 
> "[...] I would prefer:    http://www.ivoa.net/xml/VOResource/v0.9
> Since the status as working draft is irrelevant for the namespace - it *IS*
> the namespace we are using for that schema, it is not a working draft, 
> it is actual.
> [...]
> We cannot guarantee a 1:1 relationship between schema and document so it is
> pointless to tie the schema namespace to any document naming scheme."  
> [Tony]
> 
> This should explain why we do not want to have the same numbering for 
> documents and xml files.

While Tony's comment is technically true (we can't guarantee it) would 
it not make sense as a working practice?  On the other hand, I 
understand that having to copy half a dozen files for a small change in 
one of them would be a pain...

> By the way, having ten possibilities to go from one level to another 
> should be sufficient, simply requires some attention in producing a new 
> *document*.

? Ten's not nearly enough!  SkyNode *was* passing v0.8 (somehow it's 
slipped back to 0.7.1?) and there's only one or two prototypes of it so 
far.  Can we expect only two more changes before it's ready for general 
release v1?  Why restrict ourselves to that if we have another 
industry-common versioning pattern?

Now I really am going to take some time off.... honest I will be quiet 
now...

MC




More information about the stdproc mailing list