GGF 11 - 08/06/04 - DAIS Session 2: Specification Realisations ============================================================== Chair: Norman Paton Note Taker: Mario Antonioletti Agenda for this meeting: 1. Relational Realisation: Brian Collins 2. XML Realisation: Amrey Krause 3. Summary/Plans --- Norman introduces the session. 1. Relational Realisation: Brian Collins Presentation at: http://forge.gridforum.org/projects/dais-wg/document/GGF11-DAIS2-Relational/en/1 Have changed the name of the specifications to use the "WS-" prefix. Specific changes made since the GGF10 version of the spec are outlined in the slides. There is a small mistake in the spec submitted to GGF SQLResponseAccess should really be labelled a factory. This will be corrected. Can have alternative formats for the data coming back. The valid set of return data types is specified by the SQLResponseDescription. Have tried to be consistent with faults coming back both in the relational and xml realisations. Still not entirely consistent but we are getting better. A couple of scenarios are described in diagrams in the slides. Guy Rixon: in the behavioural properties ... can a custom form be used? If you could put in one of the interfaces are ADQL.... Dave Pearson: what would you expect the service to do ... Charlie Little: you could say that this is not supported .... Brian: we chose the three types - common standards across vendors ... Guy: but you may want to know the dialect ... the client may want to check this ... if the service does not understand then you want to know why it does not understand... Norman Paton: the values in the spec are indicative ... these are not necessarily the legal values ... Guy: you could have a version, e.g. 1.8. Norman: yes you could. Brian: if you put a statement that is illegal to that database then you will get an error at the database level which will then bubble up to the service level and be passed back to the service. Guy: do you have to put a flavour in the execute query that could be used? Brian: no ... we are interacting with CIM to see if general stuff can be added ... Dave: your motivation for asking this is that when you are discovering services you want to ensure that your client will be able to talk to the data resource? Guy: yes ... however this could be done through other processes ... Norman: in an early version of the spec you could specify the language constructs that could be used with that data resource but we thought that was the beginning of the slippery slope ... now we do this very simple thing ... if you go with the CIM model you will get versioning and the optional packages that the data resource supports ... also vendors publish versions of this model which could be used by the service. Charlie: a compatibility service could do this .... Norman: but that is difficult to do ... Guy: could it be possible to make this a small set of items? Brian: the fact that the is already in the plural suggest this... Charlie: what do you mean by data management? Brian: decided that this is out of scope ... there are three views to this: - The management of the service, e.g. starting services, etc this is covered by the DMTF's MOWS specs; - There is the management of the underlying resource - start database, stop database, basic DFDL operations some of which could be done using MUWS but to that properly you need to have appropriate informational properties and that is where CIM comes in - as we do not parse the statements that are passed into the services we do not stop you doing this so you can do that by passing SQL queries in but that is vendor specific; - the third area is the management of the service and the resource ... Norman: credential mapping would be in this third area ... there is a smallish number of such things ... Charlie: by Data management then you don't mean anything like replication? Norman: we don't ... not sure if that is acceptable ... we do not consider data replication to be in scope .. have not looked at properties that would cover data replication ... Charlie: asking because I am part of the data replication service ... people keep on asking about files .... Brian: there is some work on data replication going on in the infod working group. Dave: that is a use case ... the only replication that you could use would be triggered replication of a database ... we do not do not do replication at the architectural level ... Allen Luniewski: there is nothing to stop you using the framework to do replication. Norman: that is true ... at the service layer we do not say anything about data replication ... there is nothing to stop you to do that behind the scene.... Let me invite comment on a single thing ... we have adopted a single SQLExecute operation as we are not going to parse the input ... the downside of this is that this produces a rather generic result which is more difficult to unpack which produces a challenge for tooling ... in terms of the factory pattern a collection being returned has to wrap something more complicated ... how do people feel about this? The alternative is to have SQLQuery and SQLUpdate which give you more constrained results. ... Norman: the return is an SQLExecuteResponse which wraps the different return types ... Guy votes for the way that it has been changed to. Norman: happy to receive feedback at the meeting or through the list. ---- 2. XML Realisation: Amrey Krause Presentation at: http://forge.gridforum.org/projects/dais-wg/document/GGF11-DAIS2-XML/en/1 Amy presents the main changes from GGF10 - now have an XMLSquenceAccess which is similar to the relational spec. Other changes are outlined in the presentation slides. Operations have not changed other than the first letter being capitalised for the xml collection access operations. ... Some of the portTypes have been tided up a little bit. W3C currently working in XMLSerializationMethod ... and also working in XMLSerializationProperties ... Amy starts talking about issues associated with the current XML spec. Mentions the association of multiple XML Schemas with a collection. Norman: why can there be multiple schemas? That is the way that it is out there. Dave: does that correspond to views in tables? I don't think that that is the same thing ... ... Norman: the difficulty in this context is that the XML standards are not there ... Need use cases. Steve Langella: you could put everything in one schemas but if you had a schema for books and magazines then it might be best to associate a document with one of the schema ... if you had a schema that extended another schema then would you require to have both schemas there? Yes, that is similar to my next issue.... these things need to be resolved. Norman: if you don't have the ability to remove documents using XPath you could do that one document at a time .. the use of XPath is an optimisation Steve: I suggested using XPath to remove documents ... I believe that you can do already do this using XUpdate. XML management standards do not support management so we have provided our own interfaces ... when interfaces become available we may be able to use these but up until the point that that happens we need to provide interfaces. Norman: sounds to me like we need to support that kind of functionality ... if over a period that functionality become available then we could support that ... it's not as if we have a large number of operations. I agree ... I think we should leave it there for now. Guy Rixon: in DFDL have binary representations of files ... to which you'll be able to perform XPath queries ... was wondering whether you would be able to support that kind of functionality ... Is that not the kind of thing that you would put behind a file based interface ... we have not done that ... Norman: that is an implementation interface ... you could use these interfaces to wrap this kind of access ... However if you want to return binary ... we only return xml... Norman: but we have the capability to return multiple formats. It is a legitimate question to see whether the current spec could used to provide that type of access ... ---- 3. Summary/Plans Presentation at: http://forge.gridforum.org/projects/dais-wg/document/GGF11-DAIS2-Summary/en/1 Have 12 weeks until the next GGF12 document deadline. This includes July/August when people generally take holidays. Dave Pearson and I discussed activities that are required for DAIS. - Socialising - take the out specs out and expose them to get feedback and encourage involvement. We think that we are in the right position to do this now. There are still mapping issues but the core bits are more stable. Need to have an executive summary and questionnaire to bring out issues. Have targeted interaction and report back to GGF12. Dave has volunteered to move this forward. Dave: would be interesting to get a response from people in the IT industry, system integrators ... people that might like to adopt the standards and try to implement them ... looking for communities. Charlie Little: work for a software vendor ... what about object data bases or other access mechanisms that do not fit in these patterns ... Norman: we have sought to accommodate different realisations ... currently also inching towards an object realisation ... have a 15 min slot this afternoon to talk about this ... want to keep the number of realisations down at the moment ... looking for feedback. Charlie: trying to look at the difference between the specs compared to what was there before and trying to catch up ... important to keep in mind that these specs should not limit the types of data access ... do not want to restrict this ... try to keep that in mind... Brian Collins: we are also kicking off a files BOF which is a potential fourth realisation .... Charlie: is the metadata working group that has a coordination activity that needs to be coordinated with this? Steve Langella: there is no metadata working group ... we are trying to find a home .. Charlie: can use metadata as the driver for a more generic query system ... ... Currently have 5 documents in progress. Three of these have been updated for this GGF ... the OGSA data service document has not been updated for this GGF. The specs are thus slightly out of phase. Want to focus on the OGSA data services document for GGF12. Have a discussion on the OGSA data services document at a session later on today. Also have the mapping document that will be aired for the first time at this GGF. Look forward to this going through as an informational document for GGF12. Need to ensure that we have a nominated person from DAIS that is tracking/involved in the OGSA data architecture document. Contemplating having a DAIS f2f possibly coinciding with the OGSA Data Architecture document. Guy: when will the specifications become standards? There are explicit dependencies on other standards that require to be resolved before these drafts can go to specification status... the answer to this question largely depends on the mapping question. ----