DMI Telcon - 12 September
=========================

Attendees:
	   Steve Newhouse, Microsoft
	   Mario Antonioletti, EPCC
	   Michel Drescher, Fujitsu
	   Allen Luniewski, IBM

Agenda:

1.       IPR Notice
2.       Previous Minutes and Action Review (no actions from last
         meeting)
3.       Agenda Bashing
4.       Review comments on the spec made by Allen, Steven and Michel
5.       WSDL and XML Schema Review and Implementation
6.       Socialising the Spec
7.       AOB

+--

3. Review of comments

Allen sent an email with comments on the spec (draft 35). This
elicited further comment from Steven and Allen.

Most of these were good catches and are being addressed (new version
will go up soon after telcon).

Only noting salient points:

Issue between the created and the scheduled state of the DTI. Steven
advocates that a change of state can take place if the user invokes
the start operation or if the scheduled start time is reached. If a
DTI is created without a scheduled state then it will only migrate to
a scheduled state if it is done manually (start operation is
invoked). This needs to be clarified in the spec document. The text in
the document is wrong/unclear which is causing the confusion.

Issue as to whether Activate and Start should be swapped around
in terms of their semantics.

For section 5.1.1 need to add text to say that the DTI state will
migrate only if moved manually if the start operation is used.

The start not before time would equate to the scheduled time of the
DTI.

There is an issue as to how you represent a "start as soon as
possible" or "start NOW" time for the service - needs to be addressed
off-line.

Need to be very clear about the start and created time and what the
semantics are. It is not very clear in the document just now.

Question about the DEPRs. Defining a PortType that extracts the
contents of the metadata section (source and sink have to be able to
generate this metadata). How much do we specify this. One option is to
leave as is now and see if we get any comments about this when we go
to public comment.

The URL attribute should probably be renamed. It should be a URI, the
name, as it stands, is misleading.  This data type for the XML Schema
reflects this already.

The spec will be updated by Michel and some of the points uncovered by
Michel will be discussed at a future telcon (Michel will send out an
email to the list with this).

5. WSDL and XML Schema Review and Implementation

Michel produced the WSDL and XML Schema currently included in the
spec.

Steven took the WSDL/XML Schema produced by Michel and run it through
Microsoft tooling. No issues thrown up. Michel used eclipse to
generate the XML Schema/WSDL. No surprise then that Mario had no 
problems with eclipse.

The HPC Profile looking at supporting a restriction on the JSDL data
staging elements. Will require re-engineering of the Microsoft HPCP C#
web service implementation - need to check that transfers have been
successful - this capability could be implemented using part of DMI.
Steven will be thinking about this. Michel will be looking at trying
out a DMI implementation later. 

Can try to look for other volunteers to try and do implementations.
Any volunteers out there?

When Michel was doing the schema for the the start, stop, activate,
etc operations noted that response elements had to be defined in the
WSDL. Currently have define empty response messages - without these
you can't throw a fault. only the status element currently has
content. Probably need to work these back into the spec - they need to
be synchronous. Michel will have a look. Others should think about
this too.

6. Socialising the Specs

Michel has been sending version the spec to people. No comments thus
far.

Steven will talk to Andrew Grimshaw about the spec (both are currently
attending the UK AHM).

7. AOB

Michel has some thoughts of the WSRF rendering but will talk about this
later.

Michel will do the changes to the spec discussed during the telcon.


One for Ravi: as DMI chair we need your feedback on the current
version of the spec.

Need to discuss planning for OGF - produce a slide set.

Michel noted things to discuss might be the position of DMI within the
OGF landscape - how it relates and is distinct from gridFTP, ByteIO,
etc. Possibly have some hypothetical client pseudo code to show easy
it would be to use.