perfsonar development work

[Martin Swany <>]

Hi all,

> I don't think the situation is any different than an MA actually, if  
> you do not have the topology elements (in your case they do not  
> exist) you can't register them.  You do have access to what  
> eventType(s) are available though, and the service itself may have  
> ACLs that allow it to make measurements in certain domains or IP  
> ranges (or it could be wide open) so these items may be substituted  
> for a particular 'interface' or 'endPoint'.  Your example above  
> looks acceptable.

I'm sorry to join late, and this may have been pointed out.  One
difference between and MA and MP (and one reason why this
discussion was neglected for a bit) is that an MA can automatically
summarize what is storing (which can potentially be anything!)
whereas an MP needs to indicate specific eventTypes that it
is capable of generating.  Earlier in this thread someone said
that the SQL-MA would register a bunch of eventTypes, but in
reality, it can potentially store a wide range of things, and really
can only tell what it *is* storing (and this can be derived from its
metadata.)

As Jason said, global discovery depends on having consistency
of registered information.

Also, as Jeff said, we believe that the service types should be
structured (URIs) as well.

best,
martin