DAIS Telcon 13/07/04 ==================== Chair : Norman Paton Minutes: Mario Antonioletti Present: Mario Antonioletti, EPCC Norman Paton, University of Manchester Simon Laws, IBM Vijay Dialany, University of Southampton Susan Malaika, IBM Agenda: 1. Minutes/Actions. 2. Discussion of Briefing Document on "socialising" the specifications (Dave Pearson). 3. Update on issue logging for specifications (Mario Antonioletti). 4. Update on process for finalising mapping scenarios document (Simon Laws). 5. Update on process for addressing mapping issues for specifications (Simon Laws). 6. Discussion on possible meeting on 26th July, in association with WSRF meeting (Norman Paton). 7. DAIS/File Access Group Interactions (Brian Collins) 8. AOB New Actions: [Mario] Mail list stating that process for resolving trackers outlined in previous mailing is the one to be used in future. Actions from previous meeting: [Dave] Post a straw man executive summary and set of questions before the next call for people to comment on. Done. [Mario] Post a summary of existing DAIS issues to the list and let folks know that the Grid Forge trackers are being used for this. Done. [Mario] Propose a process for closing issues that are posted on Grid Forge. Check out what other GGF groups are doing in this space. To be done by the next call. Done. [Norman] Email the people involved in the Object Data Access Realisation to see if this effort can be picked up again. On-going. ---- 1. Minutes/Actions. 2. Discussion of Briefing Document on "socialising" the specifications (Dave Pearson). Dave not present so this item was not discussed. 3. Update on issue logging for specifications (Mario Antonioletti). Mario: Put out a mail with a list of existing trackers and artifacts. Suggested a process for resolving artifacts (issues). Norman suggest that Mario should mail the list stating that this is going to be the process for resolving issues. 4. Update on process for finalising mapping scenarios document (Simon Laws). There is a call on the 27th of July. Will try to close down the mappings document. No particular progress but moving forward. Sastry, one of the authors, will be leaving Oracle so that's an incentive for this to be resolved soon. 5. Update on process for addressing mapping issues for specifications (Simon Laws). Norman: discussion of the singleton resource pattern going on in the WSRF mailing list after Susan posted a mailing there. Simon: ... there is a disagreement about the model to be used to identify resources. *Susan joins* Norman asks Susan to comment Susan: it's still under discussion. Simon: ... singleton resource pattern would allow a web services and WSRF based framework to be used ... Susan: you are still supposed to have reference properties ... Simon: you don't need reference properties if you have only one resource. According to WSRF people you still have an EPR ... the model works in both situations but that does not answer the skeptics that believe that resources should not be modelled. Norman: ... RESTful is a distraction as you encode the resource in the URL ... ... Simon: encoding the resource properties in the URL is seen as a bad thing as that is protocol specific ... Susan: in DAIS we name the database, the table, it's part of the resource - all these things have established names and access mechanisms ... you think that someone will be against that ... who would argue against it? Simon: if things are already identified what is the objective of WSRF? We had this conversation a long time ago that it's all about identity ... we had an OGSA-DAI Technical Review Board where Malcolm Atkinson pointed us to the OGSA document that has something about naming which is very useful ... it splits the naming problem down into three things without being specific to the mapping - the three things are a human readable name, an abstract name - unique for all time and an address ... you can map between those three things ... this is a useful framework to use ... also they gave examples in the document ... someone has thought what this means in WSRF and in other things ... we should think about that ourselves. Norman: section 3.9.4.1 on page 54 or the OGSA document ... this seems to be a very specific use of a WSRF EPR ... you can pass something that is not physical at all ... Susan: yes Simon: that EPR with no renewable reference part identifies a particular resource. Norman: Could see how I could go to a service ... pass an EPR, the service could unpack that ... it could be location transparent but the person that uses that EPR cannot tell... Simon: some will think that this is a great way of building distributed databases or you could use it to hide stuff ... some people talk about bank account ids and other about disks which are at opposite end of the spectrum ... we can use the OGSA framework. Norman: has the renewable resource document been published? Simon: no. Norman: trying to find a mode of operation that makes most people make comfortable or be able to factor it in if people require it ... Susan: will do some more probing .. Norman: look forward to this ... Susan: what should we ask next? Simon: what happens if I already have an id? Norman: could use the scenarios ... have two scenarios ... where the database is the data resource or where the result is treated as a data resource ... we could make that community look at our scenarios ... when the database is our resource you have 3 options: a service fronts one database in which case you don't want to be bothered with a WSRF construct if you only have one resource, the second possibility is that you have a naming scheme that is understood by the underlying management system that understands this, e.g. Oracle has a global naming scheme for databases, and you could use that ... and the final possibility is that you do have WSRF. The last two examples have multiple resources ... we would like to be able to use option 1 - our desire is to allow options 1 and 3 to co-exist in the spec ... or there is no WSRF issue for option 2 or it's ruled out of bounds ... it could become a 2 step process .. look up that name in a registry and get an EPR and then address the service through the EPR ... could get a level of indirection ... could be a good thing ... Susan: ruling out option 2 rules out a lot of web services ... Simon: number 2 is not just a database thing ... Susan: sure but we want them to look at the scenarios ... Simon: WSRF people will say in option 1 you have an EPR, in option 2 you map it to an EPR and option 3 is an EPR. Norman: but there are different views on this ... the question is though how do we make 1 a sub-option for 3 ... Simon: but WS-Addressing was not written for WSRF, WSRF was written to use WS-Addressing ... if you are using a URL then clearly it's not an EPR but you could also use that as an EPR then the consumer can't tell if a WSRF EPR or a standard WS EPR... Norman: as a client this is a good thing but the service provider might get grumpy about naming its internal resources ... want to achieve the client not to know these things. Susan: how can it see if it can get reference properties? Simon: can look at the WSDL and see if the methods are supported or send a message and if faults ... ... Susan: talk a little about the result set one ... when you have a result set you do not have the luxury of a global name ... unless we invented one. Simon: there is a space in the metadata for it ... Susan: it's a matter of implementation ... you could use a temporary table. Norman: if you created a temporary table and imported the name the management of that table would be managed by the underlying resource whereas if it's managed by the service it's a first class citizen in the service space and a 2nd class name in the resource space Susan: have been talking about this today ... bit of resource that is tightly coupled to the service ... WSRF is ideal Norman: it makes it a first class citizen of the service ... 6. Discussion on possible meeting on 26th July, in association with WSRF meeting (Norman Paton). Norman: Meeting on the 26th supposed to tie with the WSRF face-to-face taking place in London. Susan: There are two Oracle people attending the WSRF meeting but they are not DAIS ... they are from the application server group but they should be ok about this ... the WSRF is all day on the 27th ... one of them is flying in for the Monday and the other is coming in on the Tuesday ... Norman: so there is a possibility on the Monday ... Susan: but I will not be able to attend in the afternoon ... Norman: could meet up with Simon and meet up and discuss the scenarios ... we should be able to have a useful discussion with someone from an AP Server ... could also possibly meet post the WSRF meeting but I will not be able to attend. Susan: what about a meeting after the WSRF days ... * general discussion trying to pin down a time to meet * 7. DAIS/File Access Group Interactions (Brian Collins) Could not do as neither Neil nor Brian were not on the call. 8. AOB Susan: CGS model stuff is proceeding nicely. Being vetted by the IBM SQL standards people. Norman: there is a question whether the spec team should try to meet at some point. May be able to suggest a suitable time next week. Mario: Status of charter? Norman: area chairs have been sent the updated charter but have not had a reply. ---