[dais-wg] Telcon Minutes from 17/05/05.
Savas Parastatidis
Savas.Parastatidis at newcastle.ac.uk
Sat May 21 19:05:31 CDT 2005
Dear all,
I am really surprised that the discussion has moved once again to
whether the interfaces should be designed using an object-oriented
approach.
>From the minutes below, I see that there are still some serious
misunderstandings of the position I have been expressing...
- I have never suggested that there is no state encapsulated or used by
services. Have a look at the amazon.com Web Services. They provide
access to lots and lots of data without requiring explicit modelling of
state. They use explicit identifiers to access 'datasets' stored by the
service (e.g. product id in order to get information about a product).
- The use of .NET or Java or other technologies has nothing to do with
service design and Web Services technologies.
- The use of a particular technology to implement the way in which state
is stored and managed by a service implementation has nothing to do with
interoperability.
- Also, the 'context' approach is a mechanism to achieve stateful
interactions or to logically correlate messages in a distributed
activity. There are other mechanisms as well (e.g. the use of session
IDs).
- The use of WS-I Basic Profile specs only does not mean that state
cannot exist behind the service boundaries. Look at the SkyServer,
Amazon Web Services, etc.
I hope these help!
Best regards,
--
Savas Parastatidis
http://savas.parastatidis.name
> -----Original Message-----
> From: owner-dais-wg at ggf.org [mailto:owner-dais-wg at ggf.org] On Behalf
Of
> Mario Antonioletti
> Sent: Thursday, May 19, 2005 9:41 AM
> To: dais-wg at ggf.org
> Subject: [dais-wg] Telcon Minutes from 17/05/05.
>
>
> DAIS Telcon - 17/05/05
> ======================
>
> Chair: Dave Pearson
> Notes: Mario Antonioletti
>
> Attendees:
> ----------
>
> Mario Antonioletti, EPCC
> Dave Pearson, Oracle
> Thomas Soddemann, RZG
> Susan Malaika, IBM
> Allen Luniewski, IBM
> Simon Laws, IBM
>
> Agenda:
> -------
>
> o Session Planning for GGF14
> o Object Data Access Realisation
>
> Actions:
> --------
>
> [Norman] Create a list of options for update of disconnected data
sets.
>
> [Simon] Write up the top level IO operation to go in the core spec.
>
> [Mario/Simon/Amy] Look at DAIS issue 1200 and either decide what needs
> to done to resolve this and respond accordingly on
> Grid Forge or flag it as an issue that needs to be
> discussed at some future telcon.
>
> [Mario] Continue working on the DAIS contribution to the data
> architecture.
>
> [Susan] Look into how the XML attributes and elements for the resource
> properties relate to the DMTF CIM model.
>
> -----
>
> Thomas: posted a document on Grid Forge.
>
>
http://forge.gridforum.org/projects/dais-wg/document/WS-DAIO/en/
> (have to sign in to see this and possibly be a DAIS member).
>
> It's a start. Finding it difficult to find time. Need help to
> complete this for GGF14.
>
> Pearson: what specific help? The specific issues in your mail?
>
> Thomas: have ideas to bring into the document but would like other
> input. For instance on how to reference objects by web
> services - do not want to stick to one technology, e.g. WSRF.
> Later on may be able to match to WSRF.
>
> Susan: there is a naming effort going on with OGSA - can use reference
> parameters. May want to look at that.
>
> Thomas: that would be useful.
> There is some kind of collection and that takes a naming
> schema and then we are done for the first draft.
>
> Susan: how do you name your objects?
>
> Thomas: the reference by itself is not a big deal but it's what it has
> behind it - an object has a state and a WS is usually
> stateless.
>
> Susan: but that's why WSRF is there ...
>
> Thomas: want to be a little bit more abstract to allow the use of
> other technologies .... could possibly use COBRA could be an
> implementation. Also, have .Net. These would offer a state.
>
> Dave: are you saying that a WS-I framework is not very appropriate for
> you?
>
> Thomas: yep.
>
> Dave: Savas would say that there are ways of exposing state using
> contexts.
>
> Thomas: I should think about that.
>
> Dave: Savas has always proposed that there are ways of doing this. The
> mapping's framework offers some ideas.
>
> Susan: the business of passing names and state is part of the content
of
> the mappings document.
>
> Thomas: you leave it to the implementor to address the statefulness
> ... then you end up with implementations that are not
> interoperable.
>
> Susan: the mappings document was just there to explore the space
>
> Thomas: if you know the thing is stateful then you have to be more
> careful.
>
> Susan: but this is true for relational and XML databases.
>
> Dave: hopefully we can avoid exposing that state ...
>
> Mario: do you have a use case in mind?
>
> Thomas: if you get a reference to an object you are directly
> interacting with that object. In a relational model
> you are interacting with a result set, which may be
> a copy but may not be directly linked to the database.
>
> If you get the object out of the database - there is
> only one object regardless of the number of cases.
> You also have a number of links.
>
> Dave: the result sets here is the output from a query. Whereas
> in an Object database the output is some data. Thomas
> is talking about manipulating the object itself.
> If you have an object that has several objects behind
> it then that is different from a view.
>
> What state do you wish to expose Thomas? I can't find this in
> your document.
>
> Thomas: talking about inheritance - gives different views on the same
> object. This has to be dealt with. A simple object, like an
> XML bean, then you have attributes that can be Set and
> Get ... as an implementor you have to decide whether you
> expose a single object or expose multiple copies of objects.
>
> *Simon joins*
>
> Dave: a problem that Thomas is how you expose the state of an object
> ...
>
> Simon: what do the interfaces look like? access to objects and
> collections of objects?
>
> Thomas: yes ... there are ops like select and update where you get
> objects, there is also find where you get data, etc.
>
> Simon: it's like the XML realisations where you have documents and
> collections and the response is a set of XML fragments - we
> might have to expand those ideas to deal with your objects
> ... is there any work of XML to object mappings? ....
>
> Thomas: you can represent the state in XML.
>
> Simon: we could adopt the mapping to get the state of an object ...
> is there any way of remembering the function of an object?
>
> Thomas: it depends on the implementation of the object database -
> there are different kinds of data associated with an object.
>
> Simon: there's still information about the type and its version
> but not the functions that the object might support.
>
> Thomas:in in principle you have the methods and the version - for
> instance objects need to be serializable in Java.
>
> Simon: sounds that something could be done like in the XML....
>
> Thomas: have not thought about it that way. Originally started from
> the relational realisation and derived the object
> realisation. Could look at the XML spec .. sounds like a good
> idea.
>
> Dave: in Oracle have something called toplink that might be of
interest,
> Also Ajax may be of interest - useful documents in the IBM web
> site.
>
> Thomas: Ajax is a web technology where java script is executed at the
> client side and loads XML from the server - is that it?
>
> Dave: don't know if it might be useful...
>
> Thomas: no, not if it's what I described as that is just a publishing
> technology....
>
> [want to give a third party (implementor, e.g.) the
> possibility to expose toplink and other proprietary O/R mapping
> technologies employing another namespace]
>
> Mario: so you are just looking for ideas and asking for feedback...
>
> Thomas: yes
>
> Dave: Norman is the best person to deal with.
>
> Mario: we could try to read the document and feedback to Thomas in the
> meantime.
>
> ...
>
> Thomas: will have a look the XML spec to see if I can get ideas from
> that. If anyone else has additional ideas will be grateful for
> those. Also, would be grateful to get help to evolve the spec
> for GGF14.
>
> Susan: the document deadline is the 27th May.
>
> Thomas: will try to get the document into a better shape by then.
>
> Susan: should be ok to put in a preliminary draft.
>
> Dave: yes, in particular if you have a list of known of issues
>
> Susan: and any points you want to pursue and expand on ...
>
> Thomas: there are comments that are not going to be the final
document.
>
> ...
>
> Dave: should get in touch with the CERN database group ... will email
> you with some names. They may have done some prototyping...
>
> Thomas: we use objectivity ...
>
> ...
>
> Dave: It might also be worth sending Norman an additional mail.
>
> Planning for GGF:
>
> Mario, Thomas, Dave will be there.
> Simon not planning to go at this stage.
> Expect Norman to go.
> Susan not sure yet.
> Amy is going.
> Allen is doubtful as to whether he will be able to attend.
>
> Dave: has any progress been made.
>
> Simon: I don't think I should be the road block.
>
> Mario: have not had time. Have not done much.
>
> Susan: can we get an extension?
>
> Dave: can send an email to Steve Crumb to see if we can
> get an extension. Probably the best I can do.
>
> Susan: do Mario/Simon have half a day this week to discuss things.
>
> Simon: I can do some time this week. Have been looking at the
> WSDL.
>
> Mario: this week is fine, I'm off next week.
>
> Simon: I'm unavailable next week. Have had effort promised from Paul
> and Malcolm to get review time if we get something ready.
>
> Mario and Simon will get together to try and do something.
>
> Dave: hope to do two sessions at GGF to cover the different specs.
> Half session for objects and one a half sessions to present the
> other material.
>
> Susan: ok.
>
> *Discussion regarding what we should not clash with DAIS*
>
> Simon: if we can come out of this week with the WSDL then that allows
> progress to be made.
>
> +---
>
>
+-----------------------------------------------------------------------
+
> |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