Votes? [was: Vocabularies issues]

[Frederic Hessman]

>> 1. Format of non-distributed "Master-vocabulary":
>> 	_X_ nothing mandated (i.e. it's the business of the publishers and  
>> not ours to decide how publishers do their hidden work)
>> 	___ Turtle
>> 	___ text
>
> I tend to agree with this, but I anticipate some disagreement here  
> once the WD is examined outside this list.

Huh?  Damn communists!   If we can't agree on this point........!

>> 2. Format of the published vocabularies:
>> 	_X_ XML or Turtle (let Darwin decide, assuming that the unproven  
>> assertion that XML alone isn't good enough is true)
>> 	___ XML
>
> I'll make sure to include detailed rationale in the document.  I  
> don't think this is an assertion that can be proved in advance one  
> way or the other.  The discussion is only about the weight of  
> anticipated problems and benefits of either option.

Aha - so you're going to let me change my vote, now that you're not  
quite so certain? :-)

>> 3.  Versioning
>> 	___ explicit in namespace (e.g. http://myvocab.org/myvocab_v1.1#mytoken)
>> 	___ hidden, like in Dublin Core (e.g. http://myvocab.org/myvocab/#mytoken)
>> 	_X_ both, a new option: the hidden version is the latest, but  
>> really points to explicitly labeled version, so that old versions  
>> remain accessible.
> [...]
>> My vote on issue 3 is cheating, since Norman didn't list the third  
>> option:  hidden versioning is convenient for most users as long as  
>> there is a mechanism for determining what version one has read in  
>> the past (to see if anything has changed)
>
> If I'm correctly understanding what your 'both' option is, then I  
> think that's what Dublic Core does.  If you dereference the  
> versionless DC terms namespace <http://purl.org/dc/terms/> then what  
> you get is a redirection to (versioned) <http://dublincore.org/2008/01/14/dcterms.rdf 
> >.  This was option 1 in <http://www.ivoa.net/forum/semantics/0801/0660.htm 
> >.  This is what you mean, yes?

Well, the question is, whether a user can find out what version s/he's  
being fed.   Maybe I don't care which version it is, but maybe I do  
(indeed, I probably will).

> In this option, clients do any required reasoning with a versioned  
> namespace, and the published/publicised URL is unversioned, and  
> really only used for discovery.  This sounds plausible to me.  Is  
> this an actual Best Practice, does anyone know?  Does it have  
> downsides?

or is it too complex to find out?    Catching a redirection can't  
possibly be the way of determining the version of a vocabulary.  If  
there's no Dublin-core-style way of documenting versioning, then we  
need to stipulate explicit versioning (not very pretty) or mandate the  
inclusion of an official versioning element like

	<ivoa:vocabularyVersion>1.1</ivoa:vocabularyVersion>

even if it isn't standard (it's not our fault that our non- 
astronomical colleagues are careless).  This way, http://myvocab.org/myvocab#mytoken 
  looks like it's long-term, but a glance at the internal  
documentation can be made to double-check if necessary.   If no  
version is given, then one must assume that the vocabulary is stable  
and trustworthy (e.g. constellation name vocabulary) or undocumented  
and thus unstable and untrustworthy (depending upon the context).

Let's then call this option #4 - versioning in content, not in  
namespace.


Rick

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.ivoa.net/pipermail/semantics/attachments/20080204/ea177a70/attachment-0001.html>