[dais-wg] DAIS Minutes from 21/06/05 Telcon.
Mario Antonioletti
mario at epcc.ed.ac.uk
Wed Jun 22 07:50:47 CDT 2005
DAIS Telcon - 21/06/05
======================
Chair: Norman Paton
Notes: Mario Antonioletti
Attendees:
Norman Paton, University of Manchester
Mario Antonioletti, EPCC
Susan Malaika, IBM
Simon Laws, IBM
Thomas Soddemann, RZG
Amy Krause, EPCC
Agenda:
+---
Susan: no longer have type 1 and 2 resources - now call them
internally and externally managed resources.
Simon: so the current position is:
Sue wrote some coherent notes.
Deal with externally managed or service managed data resources.
Externally managed data resources are as before - lifetime not
managed by DAIS, opposite is the case for the service managed
resources.
There are a set of core interfaces - these applies to both
kind of data resources, they define some properties and
some messages as well as messages for accessing the data
resources. The messages contain the name of the resources
being accessed.
The realisation define some properties and some messages and
some additional properties/messages for the given
realisation. They rely on passing the resource names.
There are also some optional things:
The data conduit - say that this is a WS-Resource which
represents a data resource and it's purpose is to describe
the lifetime of the underlying data resources ... it's
a facade over the data resource.
The other thing that has been pulled out separately is an
optional catalogue - a list of names, and also possibly EPRs
and other things.
When you build the service you can include the catalogue, core
interfaces and realisation interfaces and may include conduit
interfaces ... in the case where you try and do lifetime - you
either manage to do it or you get an error.
That's it ...
Norman: my concern is that when we stand up and suggest this as the
model we may get criticised for not using WSRF.
Simon: I say we are ... there is an optional data conduit that can be
used to control the lifetime.
There is a 1-1 relationship between a conduit and a resource
....
Susan: the conduit provides a WSRF representation ... so you can have
a WSRF view ...
Norman: so can have many conduits associated with a service which is
why you have separated it from the list of names ... the
conduit is optional so if you have no conduit then those
messages must have the names in the body regardless of whether
it is an externally/service managed resource. So if you don't
have a conduit you have no lifetime management ... if you have
a conduit then the messages are sent to the EPR so that there
are up to 2 names for every resource - the conduit name and
the underlying resource ... an externally managed resource may
pass the name of the underlying resource ...
...
Simon: not happy about unplugging the relationship between the conduit
and the data resources there is no way to plug it back
together again ....
Norman: have 2 naming schemes and fault when you send a lifetime
message if there is no conduit. In the case where you create a
service managed resource what does the service return?
Simon: a name and an EPR - the address. The EPR could point to a
service or it could reference the conduit. It's just an
EPR. The name is a synthetic name but it could have a greater
significance.
Norman: the scheme is trying to define the messages in a way separate
from the conduit but allows these messages to be blended in
with the conduit ...
Simon: the conduit allows more generic resources ... the core and
realisation can define the messages without having to worry
about whether you are addressing the conduit or the resource...
...
Norman: two possible issues that might arise with this model:
- the WS-I people may think that they have a dependency on
WSRF, e.g. have to use WSRF for lifetime and there is an
association between a conduit and WSRF. They are forced to
use WSRF for lifetime.
Susan: there is a destroy...
Simon: so this is something we need to decide - do we include a
destroy? That allows you to manage the lifetime but not the
soft state ....
Norman: having a destroy that takes a name cuts slightly against the
conduit. If you stick with the names in the body then you are
a message short of achieving your goal ...
- The WSRF people might think that the use of WSRF is somewhat
indirect .... because you have two kinds of names when you
could have got by with one - an EPR.
Simon: it doesn't stop an implementation from doing that ...
Norman: you can ignore the names ... but you have properties
that pass the name forwards and back that are somewhat
redundant...
Susan: can think of them as descriptions...
Norman: I don't think that helps. ...
Neither of these sound really problematical ....
So who is going to present this next week?
Simon: is the model plausible and can we describe it?
Sue has written some words ... which
includes some MUSTs and MAYs...
Norman: where are we at with the associated spec and WSDL?
Simon: WSDL is done, subject to Amy reviewing the XML WSDL. The core
document has not caught up yet.
Susan: so do you want to plug in the text?
Simon: would be good if someone could start doing this ...
Susan I will start.
Susan: now the WSDL is done - need to go through the other specs and
compare with what is on the specs.
...
Susan: did you see the DMTF message I sent out? Essentially there are
3 DMTF groups we are dealing with - one to get the namespace,
another to generate the XML and a third to review the model.
Have been dealing with all three groups - if anyone else would
like to join in then that would be good. The calls are every
other Monday - would be nice to have someone else dropped
in. Can only have one GGF rep per group. Fairly easy for
academics to join.
Would be nice to push the namespace thing ... someone to
encourage them ...
Norman: I'm willing to consider one of those things ... have to decide
where I can make the best contribution.
...
Need to decide about GGF presentations.
- Thomas can do the object realisation, 20 mins to
present and 10 mins for discussion.
Also XML, what do we want to present and what do we want
feedback in.
If the object one gets a slot in the DAIS sessions ... need
to get people that are willing to contribute. Currently have:
Thomas, William Sanchez and some observations from Objectivity
... possibly some interest from Manchester.
Thomas: yep, want to get some more interest...
...
Norman: Socialisation is going to take place after submission ...
regarding implementation - we know that we are getting an XML
implementation from Manchester under DAIT, a relational
implementation from Hursley possibly under DAIT and we are
going to get an implementation from Oracle - not sure if for
both the relational and XML realisations.
Need to speak to Dave about the implementation from Oracle.
Mario: at one point Ohio were keen to do an XML implementation....
Norman: yes, they have been involved so we need to ask them ....
So, we need to put in a plan to do the interoperability.
Simon: we also need to define what we mean by interoperability...
Norman: say an IBM client should be able to access an Oracle
implementation and vice versa....
So what do we need to say about the relational and xml specs?
Not that much.
So we spend most of the first session on the model and the
mapping.
Someone needs to present this .... it will be me either Susan
or me.
Susan: let me see how I get on with the documents and I can get back
to you ...
Norman: so we will have to be in touch ...
Need to prepare slides for XML/Relational ....
Susan/Mario can do the relational ...
Amy can do the XML spec ...
....
So need a spec of the WS-DAI spec soon
and then try to do the other specs.
+--
+-----------------------------------------------------------------------+
|Mario Antonioletti:EPCC,JCMB,The King's Buildings,Edinburgh EH9 3JZ. |
|Tel:0131 650 5141|mario at epcc.ed.ac.uk|http://www.epcc.ed.ac.uk/~mario/ |
+-----------------------------------------------------------------------+
More information about the dais-wg
mailing list