perfsonar development work

["Jeff W. Boote" <>]

On May 28, 2008, at 12:55 PM, Jason Zurawski wrote:

> All;
>
>
> > > Even if this is the case, I don't know how eventTypes (at least if  
> > > we are not talking of creating new ones) can actually define this.  
> > > Suppose we have a SQL-MA which serves OWAMP data (as obtained from  
> > > CL-MP). It would register the "http://ggf.org/ns/nmwg/tools/owamp/2.0 
> > > " eventType. But this is the same eventType that a CL-MP which  
> > > permits OWAMP measurements would register.
> > This is interesting and new to me.
> >
> > Jeff, could you comment ?
>
>
> Services that deal with particular kinds of data need to advertise  
> it.  In your original example the SQL-MA and the CL-MP would both  
> register the OWAMP eventType.  The SQL-MA may advertise other  
> eventTypes as well (as would the CL-MP including ping and iperf,  
> etc.).  From the standpoint of discovery it is just important to  
> categorize things properly.
>
> Other metadata (e.g. the serviceType etc.) would help the interested  
> client/service/user decide on other issues such as if they are  
> interested in an MA or an MP.

I don't think I could say it any clearer...

One additional thought... One thing that I think I have mentioned in  
the past that I think could help this effort, is if service types were  
specified using a URN structure as well. (Just like message types...  
event types... etc...) Having this information specified in a  
structured way makes it more powerful because similar things are  
easier to relate to each other. (Specifically - the service type would  
help indicate to the client the 'protocols' this service supports.

jeff