Minutes of the OGSA-D WG telcon, 4th May 2005 0) Actions Arising - Dave to try to arrange data transfer discussion. - Allen to mail Steve about adding Allen and Dave as GridForge admins. - Mario to further update the access section. - Dave to approach transaction RG re GGF session? 1) Early discussion * Roll call Dave Berry, NeSC (Note taker) Mario Antonioletti, EPCC Mark Morgan, University of Virginia Fred Maciel, Hitachi Susan Malaika, IBM Minutes approval Agenda bashing - added context map to Technical discussion 2) Action report - Fred to send notes on how to use Majordomo to Allen and Dave. [Done. Fred also noted that some GGF mailing lists tend to cause bounces with too many hops. Allen noted there were some initial problems but they seem to have cleared up.] - Dave to try to arrange data transfer discussion. [Ongoing]. - Allen to write up naming use cases for passing to OGSA WG (after review in our WG). [Initial draft sent to Dave, Cecile and Ann]. - Ann and Allen to clean up the replication draft and mail it around. [Done] 3) WG status We were announced in the GGF general mailing. This has provoked a steady but slow stream of subscriptions to the mailing list. We have a web page on GridForge at http://forge.gridforum.org/projects/ogsa-d-wg. It is currently empty so we need to add title page, charter, members, documents, trackers, etc. Allen to mail Steve about adding Allen and Dave as admins. 4) F2F planning Meeting May 26th. Dave sent out hotel info last week. 5) Technical content - ByteIO. Neil and Mark met yesterday. They went through Mark's draft use-case template. Mark will send out this plus the draft interface that he wrote some months ago, to the ByteIO mailing list. - Data Access. Mario has written a second draft that takes some parts from the Core DAIS spec and the ByteIO interface. In so doing he encountered various questions of scope that need to be resolved, e.g. relationships between resources and services, whether to support write access. Mark commented that in all his experience with ByteIO-type interfaces in Grids, transactions had never been an issue. Scope: Mario noted that in his message, "data service represents" should be "data service talks to". Allen asked why it should matter whether a service should reveal how many services it talks to. Mario gave a couple of scenarios (e.g. provenance), which Allen thought could be dealt with by policy requirements. Mario was concerned that policy agreements might not be enough. Dave noted that the OGSA model of resources applies. Mario's message asked about service-to-service access mechanisms. Dave said that data transfer issues should be left to the data transfer part of the architecture. Mario's message asked about write operations. In wS-DAI, update is supported by passing data with the operation, but not by uploading data from a third-party. Dave ask how this applies to the data access interface as opposed to the architecture as a whole. Mario explained that the data access interface might need to know whether they are operating within a transactional context. Allen noted two cases here: one where the implementation of a federated service uses transactions internally, and the other is where the external operation operates within a transactional context. Dave questioned whether these could be truly seperated. Susan explained that a clean separation was not possible and that the context identifier must be passed through. The transaction context may be an explicit parameter or could be a callback, or by using the transaction identifier in the soap header. Susan suggested that for now we should just state how transactions might be incorporated. Susan suggested that we talk to the transaction group. Dave suggested pencilling in a session with them at the next GGF. Mario to do an update of this section. - Context Map. Mario has produced a context map of the architecture using CMap and asked for feedback. 6) Wrap up AOB DONM: next week, on storage.