Minutes MWG 2003-09-25

[Ray Plante]

Hi Tony,

One thing to keep in mind about ADQL is that while it has a data 
model--that is, it has the same basic components--as SQL (XQuery has 
similar components), it is rendered in XML with its contraints parsed into 
a tree.  A main objection to using either SQL or XQuery is that 
implementers would invariably find themselves having to parse the queries 
to transform them into a form that can be used by the local database.  
XML, of course, is easier to parse and transform.  

This has numerous advantages, including:
  *  you hide the peculiarities of the local database (e.g. not all 
       SQL support is the same)
  *  you don't lock anyone into using either an SQL or XML database.

Mapping XML into a relational model is always a challenge, of course, and 
we in the NVO have been exploring various techniques.  In my current 
favorite scheme, column names are simple XPaths.

On Thu, 2 Oct 2003, Tony Linde wrote:
> Given schema 0.8.2/3, what would a SQL-like query look like?

With many details glossed over...Where one puts a column name an XPath to 
the Resource metadata element would go.  

While I don't think we can say for sure that ADQL is the right answer at 
this time, I don't see a compelling reason yet to say that registries 
should be treated differently from other databases.  (We still 
need to compare coverage regions and such in both cases.)  
Nevertheless, prototyping different search languages is a good idea at 
this time.  We may discover the fundemental differences or different 
requirements.  Or we may discover requirements that need to be placed on 
ADQL.  

cheers,
Ray