definition of registry

[Keith Noddle]

Hi Patrick,

On Wed, 2003-04-30 at 18:52, Patrick Dowler wrote:
> On April 30, 2003 10:34, Robert Hanisch wrote:
> >     A registry is used to identify and locate resources satisfying
> > user-specified criteria, and to direct more detailed information
> > requests to
> > the relevant services.
> 
> On April 30, 2003 10:02, Ray Plante wrote:
> > A few items to think about with regard to approaching queries to
> > registries.
> >
> > 1.  Overlap with VOQL WG:
> 
> These two points together indicate something that - frankly - scares me about
> the current thinking in the registry WG: there is way too much detail at the
> registry level that really belongs in the service(s) themselves. 
I'm not sure I follow. If I understand correctly, Ray is suggesting the
Query Languages align where possible, not the data (or metadata) they
are querying.

> 
> As Bob said, "direct more detailed information requests to the relevant 
> services". The goal is the registry is to find those relevant services too 
> which is is worth sending the real query. That doesn't men that every one of 
> those services will actually have results to provide. It means they might
> have results. 

> If you think about services in a very simplistic key=value sense (just for
> elucidation of my point - clearly insufficient) then the service is able to 
> answer questions like "how many key1=X do you have?" while the registry
> should do no more than "how many services know about key1?" The registry 
> should have no concept of what the values are in any one service.
> 
No disagreement here. The most we can expect of a registry is that it
says a resource DOES NOT meet the requirement. To do that with any
useful differentiation implies the registry must have a reasonably
complete description of the resources, otherwise queries end up getting
sent to all and sundry and the purpose of the registry is defeated. The
goal should be that the registry limits as far as possible unnecessary
querying of (use of?) resources whilst itself maintaining the minimum 
metadata about those resources. Having said that, the registry is also a
publication mechanism and resources need to able to "advertise"
themselves through the registry.

Some of the AstroGrid people have been further refining the XML query
syntax and at the risk of making this another long email, below is the
(paraphrased) latest thinking.

Keith.

-----------------------------

Picking up from the previous proposal, if the 'and' and 'or' operators
are 'container' elements, the query becomes simpler, and more
human-readable.

i.e.

SELECT * FROM <registry> WHERE 
   (
      TYPE="white dwarf star" AND 
      (WAVELENGTH="optical" OR WAVELENGTH="uv") AND 
      (KEYWORD="BPM 16274" OR 
         KEYWORD="GD 50" OR 
         KEYWORD="HST photometric standards"
      )
   )

becomes:

   SELECT type, coverage, format FROM <registry> WHERE 
      (
      TYPE="white dwarf star"
      AND
         (
         WAVELENGTH="optical"
         OR
         WAVELENGTH="uv"
         )
      AND 
         (
         KEYWORD="BPM 16274"
         OR
         KEYWORD="GD 50"
         OR
         KEYWORD="HST photometric standards"
         )
      )

(note the addition of fields in the SELECT statement)

If the default operation is AND, this leads to an example query which looks like:

   <query>
      <fields>
         <field path="type"/>
         <field path="coverage"/>
         <field path="format//*"/>
      </fields>
      <criteria>
         <field path="type" oper="EQUALS">white dwarf star</field>
         <field path="instrument/type" oper="EQUALS">patrol</field>
         <or>
            <field path="wavelength" oper="EQUALS">optical</field>
            <field path="wavelength" oper="EQUALS">uv</field>
         </or>
         <or>
            <field path="keyword" oper="EQUALS">BPM 16274</field>
            <field path="keyword" oper="EQUALS">GD 50</field>
            <field path="keyword" oper="EQUALS">HST photometric standards</field>
         </or>
      </criteria>
   </query>

-- 

Keith Noddle			Phone:  +44 (0)116 223 1894
AstroGrid Technical Lead	Fax:    +44 (0)116 252 3311
Dept of Physics & Astronomy	Mobile: +44 (0)7721 926 461
University of Leicester		Email:  ktn at star.le.ac.uk
Leicester, UK   LE1 7RH		Web:    http://www.astrogrid.org