perfsonar development work

[Roman <>]

Martin Swany wrote:
> Hi Roman, All,
>
> > > In relation to the discussions today in Berlin, it was decided that 
> > > this would be more appropriate (note the lack of the 'type' 
> > > attribute in the parameter element):
> > >
> > >
> > > <nmwg:metadata id="meta1">
> > >   <netutil:subject id="subj1">
> > >       <nmwgt:interface>
> > >           <!-- ... some interface elements -->
> > >       </nmwgt:interface>
> > >    </netutil:subject>
> > >    
> > > <nmwg:eventType>http://ggf.org/ns/nmwg/characteristic/utilization/2.0</nmwg:eventType> 
> > >
> > >    <nmwg:parameters>
> > >        <nmwg:parameter 
> > > name="keyword">http://geant2.net</nmwg:parameter>
> >
> > Just one comment: type attribute could be convenient for filtering 
> > (useful for categorizing keywords).
>
> I think that we have generally adhered to the convention that parameters
> are name/value pairs.  I believe that we should think very carefully 
> about
> changing this.
>
> Based on Jeff's arguments, I really don't think that we need this to make
> the keywords approach work.  If we need additional functionality, then
> we should extend the "value" part of the name/value pair of parameter.
> We could put a type in there without changing the parameter model
> in a way that I think is confusing.

I agree, adding new attribute(s) to the parameter element is risky. The 
pair name/value should rather be rigid. But in some situation more 
information for keywords would be very useful.

Roman




> best,
> martin