A few questions

[Paul Harrison]

Ray Plante wrote:

>On Thu, 8 Jul 2004, Wil O'Mullane wrote:
>
>  
>
>>>This is a detail we have not addressed.  Our options are: 
>>>
>>>  1. we do things the proper XML way:  the ADQL query itself defines the 
>>>     prefixes it will use.  
>>>
>>>  2. we mandate the use of predefined prefixes:
>>>
>>>        vr:interface/vs:resultType='text/xml+votable'
>>>      
>>>
>>1. Will only work for ADQL/x - remeber we also have ADQL/s
>>    
>>
>
>This is a good point.  
>
>The motivation for using simple XPaths in place of column names is that it 
>accommadates both XML and relational databases to a reasonable degree.  To 
>a relational database, these paths are just strings that map to column 
>names (or they may be the actual column names).  It should *not* be 
>necessary for the RDB registry to parse the XPath.  I think this is a good 
>argument to throw out 1).  
>
>With regard to [], I think I've demonstrated that we really do not need to 
>allow them.  Do people agree?
>
>  
>
I think that if we have the registry interfaces defined in XML then we 
have to do option 2 as a minimum - XML standards are very clear about 
namespaces, and tools are free to chose whatever prefix they want - it 
is the namespace that is important.  If you have a pure xml registry 
then you will have to prepend the appropriate namespace definitions to 
the query anyway. If someone defines a registry extension with their own 
namespace, then if you are using a pure xml approach it will just work 
with no further programming effort in your query code - The sql people 
are going to have to write new interpreter code for the extension.

I think that there is a danger of making so many compromises to try to 
accommodate both the sql and xml approaches to the registry that we end 
up with something that it pretty ugly for both. Maybe now is the time to 
really decide if we want to define the registy in terms of sql or xml 
schema and stick to the implications of that choice. SQL and XML have 
different natural query semantics because of the fundamentally different 
data topologies - if the registry is defined in XML terms then people 
will naturally think of doing queries that take advantage of the xml 
structure. This revision has flattened the structure to make it more 
amenable to mapping to SQL tables, but as registry extensions are 
allowed (and expressed in XML schema) this will be a constant battle in 
the future which will impede progress and interoperability.

-- 
Dr. Paul Harrison, Astrogrid Developer
MERLIN/VLBI National Facility, University of Manchester,
Jodrell Bank Observatory, Macclesfield, Cheshire SK11 9DL, U.K.
tel +44 (0)1477 572681 (direct), 571321 (switch) - 07904025192 (mobile).