Error column in VOResource -> metadata standardization

[Martin Hill]

Pierre Didelon wrote:

> 
> 
> Tony Linde wrote:
> 
>> I completely agree, Doug. We should standardize on what we can agree as a
>> common standard - via the DM effort. 
> 
> Which will be materialised in a precise data structure in a specify format
> (XML specific or whatever else) with precise specificaly define 
> ("semantic"?) content.
> 
>> But any extensions should follow some
>> standard extension mechanism 
> 
> Why not, in this case, using a generic data structure (FITS header, 
> VOTable header,
> or whatever else at group convenance), to allow generic look up tool,
> visualisation and search, and all possible generic behaviors/handling 
> allowed
> by this structure.

I would (obviously!) vote for using straight XML to allow generic look 
ups and searches (eg using XPaths).  This starts us at level 3 below, 
but I think we can improve on that (see below).  For example, we can use 
VODataService as a start for table-based data services.  This means we 
can be 'free' to develop the structure, rather than having to keep it 
within constraints defined by an irrelevent form (ie FITS, VOTable).

> 
>> so that, as you say, they can at least be seen
>> by users or included and passed on by applications.
> 
> generic structure would certainly allow this.
> 
> So you can handle information/metadata like a three level cake;
> - the first one having precisely defined structure AND content (semantic?)
> which covers the 'most' common and important part, fully searcheable
> and directly operationnal for processing and analysis
> - the second one having a defined structure but a free content,
> less searchable or harvestable, but nevertheless searchable...
> You can always imagining and code a search looking for a certain value of a
> Fits keyword list.
> - and even a third one with a free stucture and content; i.e. 
> proprietary format
> in the sens of data producer specific format, this one non-searcheable 
> and perhaps
> even not vizualisable, but which can be propagate to user, and if he knows
> how to handle it, it could be usefull for some of them.
> 
> You can even imagine a standard process to promote some information from
> level 2 to level 1; create a new piece of info (structure and content)
> at level 1 and process (in a generic way?) level 2 structure to feed newly
> created level 1 piece of info.

Yes I like this.  However I think we can define a set of standard tags 
(such as <UCD>, <Unit>, <Keyword>) that can help do more informed 
searches even at level 3, and should make it easier to migrate from 3 to 1.

> 
> Is this silly, already in the reflexion domain or completly out of subject?
> I must admit that I din't follow the discussions on this list very 
> accuratly,
> only keeping an eye on it, so forgive me any non-appropriate intervening.
> Regards,
> Pierre

I gather it is possible to add extensions to VOResource but I don't know 
how it works...

This would certainly give a mechanism for semi-independently working on 
particular services within the VO, and perhaps even a way that some 
organisations can add information (eg error to VODataService which 
started this off!) without requiring all to do so?

Cheers,

MC

-- 
Martin Hill
www.mchill.net
07901 55 24 66