DAIS Session 3: XML and Relational Realisations ----------------------------------------------- GGF10, 12th March Chair: Norman Paton Notes: Mario Antonioletti (mario@epcc.ed.ac.uk) +----------------------------------------------------------- Presentations available from: https://forge.gridforum.org/projects/dais-wg Documents->GGF Documents/GGF10/Presentations Make sure that the document filter is working correctly. Unattributed statements in general belong to the speaker at that point in time. +----------------------------------------------------------- Norman introduces the session and gives some background on the specifications. [See the Document_Structures presentation on Grid Forge]. In future try to minimise the dependency of DAIS specifications on other specs. Explore mappings up to and perhaps beyond GGF11. Recast the specs for GGF11 but may not include mappings. These will be included later. Potential Structure (suggested in the slide) Mapping of the specification to an infrastructure will be included at the end of the spec. Alec: we found the same need some time ago .... did you consider a notation that could be used for specifying your spec? Norman: we did not find one. Malcolm Atkinson: WSDL is not changing and as that is being used already then that should still be used ..... Norman: we will provide wsdl and xml schema ... quite a lot of the material will be independent of the mappings, we believe. The capabilities will also be independent of the mappings. A lot of that could be nailed precisely - the mapping will most probably affect the SOAP headers or at least that is the hope. Simon: We have not used wsdl from other infrastructures .... Norman: are there any comments from any other groups? We are not unique in being in this position. ... if people have comments or bright ideas then do let us know? Malcolm: specs are now being partitioned into differnt layers ... in a primer it would be useful to show users how to compile these into a useful framework ... need for a primer. Norman: I fully accept that ... will deal with that in a later session. ---- Susan Malaika present status of the relational specification [See the Relational_Realization_Specification_Update presentation on Grid Forge.] Set out the agenda for her presentation. Showed the document structure hierarchy that currently exists within DAIS. Relational specification describes the extensions from the base spec. The relational specification was not re-issued for this GGF. We did not make any changes - it still uses ogsi. Planning to make changes for GGF11. For the relational spec we are issuing SQL as opposed to any other kind of request. You can get the whole result set or you can have the result held for you. You can do updates. There has been discussion as to whether there should be DDL as well as DML. The base spec is tryi8ng to come with terms to describe the behaviour. In the relational spec we have to fit within the base as it evolves and make sure that material provided in the base spec can be applied to the relational realisation. There are some other things that we need to do more work on that are specific to relational databases. Logical constructs from a relational database (listed in the slide). There are other properties with physical schema and database system capabilities. Some of these may overlap with the terms. This needs to be clarified for GGF11. Description properties may be derived from other groups/specifications - CIM - ISO spec - JDBC Metadata we need to decide which pieces we pick up from where. Review of specification activities in the XML/Relational world. - DMTF CIM 2.9 (extend this for the Grid environment) - SQL 2003 has extensions of SQL to support XML - may include XPath, XQuery Norman: is CIM not paradigm indepenedent? You are right at the higher level but for the database characteristics is it not relational...? Norman: I don't think that there is anything that is relational specific ... You are right. It is very high level what's in there for database. There is no table/column at this point. - JDBC Webrowse JSR114 we are not java specific but there is a definition for annotating results in XML ... Malcolm: group has chosen to postpone their mappings ... you may care to use different mappings this is an inefficient one.... Susan: JDBC has evolved ... it has improved. Malcolm: but there is copying of data ... we should not bind to specifications that will introduce inefficiencies ... Susan: but this is only a particular example ... we do not necessarily have to use this ... SQL 2003 ... we need to be able to represent sql in xml. There is an XML representation of query-response (jsr114). All vendors have their own implementation - annotating a result set. In SQL 2003 can publish SQL result sets in xml - the result set will be tagged up. Also have a native xml type in databases ... when you retrieve it, it will be in xml. We tag result sets in DAIS in xml - it could be that this xml is generated at the relational database level. There is also some synergy taking place between the SQL and XQuery groups. What people are doing with xml in relational databases: - work with xml but it is stored in the database without the tags (shredded xml) Norman: the xml spec has been done assuming that we are working with a native xml database ... do we have to provide support for shredded databases? When you shred an XML all information about the original xml is lost. You have to know that the original document was xml. Shredding is not standardised. It's on the SQL standardisation list but has not been done yet. At present every vendor has got their own implementation. An example of a shredded piece of xml is shown (see slides). Dave Pearson: you can hide that ... you can have an xml view of the relational view... Discussion about the mapping relational<->xml and the bearing this has on DAIS and whether DAIS should do this. View is that that work should be part of the SQL standardisation effort and not DAIS. Norman: what impact does this have on us? We may want to have a little paper or some observations. SQL can have XML operations - pass things as strings so that should not matter to us. The results can already be in xml - again I do not think that this matters. We do not need to know that they are shredding or publishing. Norman: given an xml document and a document that is shredded into relational database - can you then run an XPath query on the shredded tables? No one to my knowledge is offering this capability over shredded tables. Went over some of the issues for the relational specification (these are in the slides). For GGF11: - resolve issues - achieve consistency with base spec - continue relationship with CGS and DMTF ---- Amy Krause gives an update on the XML realisation. [See the XML_Realization_Specification_Update on Grid Forge] Amy went over the structure of her presentation. Specification mostly the same as was in OGSI 1.0, follows the model of the base spec. Have mostly removed WS-Agreement, moved operations to different portTypes, added more inputs/outputs/faults to some operations. The document will have greater changes for GGF11. Review of the existing XML specifications that are relevant to the XML specification (see the slides). A question was made about the distinction between XQuery and XQueryX. There were some terms for the xml realisations (in slides) Asked for input for GGF11. Norman: what do you mean versioning support? Instance level versioning. Dave Pearson: the list of terms are slightly different from the ones that Susan had? I did not include the ones that Simon already had. Issues are similar to the other specs. The operations have not really changed over the last 2 or 3 GGFs. Steve and Shannon have done an implementation. Future work: - achieve consistency - identify metadata properties - identify and define terms. Question about the status of XUpdate. XUpdate not an official spec but XQuery will include update capabilities.