utypes: a proposal

[bonnarel at alinda.u-strasbg.fr]

Hi Igor,
That's a little bit older. François Ochsenbein and me we made
various presentations on this in Villafranca (fall 2005) and
Victoria (spring 2006) meetings.
    We were more or less convinced by Doug and others that we had
to detach utypes from xml serialisations. The fact that the
emerging ssa was also using the dot syntax was considered as a strong
argument in favor of changing our mind.

    By the way if you really want to infer an xpath from the utype (which I
think is actually vreally useful) it's always possible.

    The real difference would be if we would allow xpath
syntax with brackets to specify some attributes or elements values
(syntax which I also proposed some time ago, as Aurelien from
ESA VO did yesterday) in order to specify generic classes into
specific ones  (characterizationAxis towards "characterizationAxis with ucd
equals phys.polarization.Stokes")
   The latter problem is also related to the UFI paradigm, which I don't
think can have a string-like solution. The More I think to it, the more 
I think
combination queries using utypes in the "SELECT", "FROM" and "WHERE" 
statements
may probably achieve most of what we want to do under this paradigm...

Cheers
François


Quoting Igor Chilingarian <Igor.Chilingarian at obspm.fr>:

> Hi Norman,
>
> On Thu, 30 Oct 2008, Norman Gray wrote:
>
>> How about this:
>>
>>   For each literal value defined in a data model, define as its utype
>>   that XPath which would retrieve the literal values from
>>   the 'natural' XML serialisation of the data model.
>>
>> That's all -- beginning and end of proposal.  The following 
>> illustrates how this would appear.
>
> Excellent proposal! I wanted to post exactly the same message on the DM list,
> but you were the first! Francois tried to propose this syntax almost 
> two years
> ago for the CharDM but it was rejected by the sub-group, so it never appeared
> on the general DM list. As far as I remember, at that time the argument was
> "it is a way too complex".
>
> I don't understand why not to follow this idea. Presently, UTYPEs look like
> "almost-XPath" but not exactly. The differences are: (1) "." instead of "/";
> (2) no "[]" with additional constraints; (3) no reference to a 
> namespace (this
> seems to change with the latest Francois' proposal).
>
> Since the UTYPEs may be used outside an XML context, the only issue is how to
> deal with namespaces, I mean how to resolve them if there is no explicit link
> to a corresponding XML schema.
>
> With best regards,
>  						Igor
>