[glue-wg] Planning for OGF 28
Laurence.Field at cern.ch
Wed Mar 10 15:18:14 CST 2010
One of the main purposes of GLUE is to advertise information to clients.
It is not aiming to model the internal workings of a services and only
information that a client needs should be exposed. A client would not
care about the garbage collection service so it would not be published.
A client may care about a database service and the quality level would
be a property of that service. The fact that the quality level was
implemented by a hidden service is irrelevant for the client.
Andre Merzky wrote:
> Quoting [stephen.burke at stfc.ac.uk] (Mar 10 2010):
>>> Service, Endpoint and AccessPolicy and I haven't noticed anything more
>>> for them.
>> Actually I just noticed one thing for Service which is slightly
>> ambiguous. We have:
>> QualityLevel QualityLevel_t 1 The maturity of the Service in terms of
>> the quality of the underlying software components; the value corresponds
>> to the highest QualityLevel among the available Endpoints.
>> However you can potentially have a Service with no Endpoints -
>> presumably you could still define its QualityLevel, but formally that
>> description wouldn't apply.
> Sorry if I jump into the discussion sideways - I am not really part
> of the glue group, just lurking here.
> Anyway: yes, you certainly can have services w/o endpoints, or at
> least which have no need to expose an endpoint. For example, we
> have a garbage collection service which at specific intervals scans
> our data bases and purges expired entries. That service is never
> contacted from client side, and nobody really cares where it runs
> (apart from the deployment point of view of course).
> And yes, that service could very well specify quality levels, such
> as run frequency.
> Not sure if that is relevant to the discussion, just wanted to throw
> that in ;-)
> Best, Andre.
More information about the glue-wg