DAIS GGF 12 - Session 3 - 23/09/04 ---------------------------------- Chair : Norman Paton Note Taker: Mario Antonioletti 1. Socialising the Specification: Dave Pearson 2. Access to Data In Files - Relationship to DAIS Specifications: Brian Collins ---- Norman opens the session. Dave Pearson will talk about the socialising of the specs. Dave: See the slides at: http://forge.gridforum.org/projects/dais-wg/document/ggf12-dais-session3-socialising/en/1 DAIS specifications have been stable for some time in terms of functionality and metadata. Held back by external dependencies that are outside our control. Investigating mappings. The rationale for socialising is that the development has been done by a small core team it is now time to raise awareness. Try to get feedback from inside and outside of GGF. Want to try to get folks to implement the specifications. Format for socialising - target audience are: - Other groups within GGF. - Research communities and institutes like CERN - Vendors, already have involvement from Oracle and IBM ... want to try to get buy in from other vendors as well. - Commercial organisation - others mentioned in the slides ... Have produced a briefing note, references to the latest spec and a questionnaire. Want to have this done by the end of the year. Want to go through the questions very briefly so people can feedback. - Want to know where the approach taken by DAIS is the most appropriate. Should it be done through a single spec? Brian Collins: specs have been evolving - the current versions still have known bugs ... will these be stated in the covering letter? I say that "we believe" that they are complete... Malcolm Atkinson: with socialisation there is a goal of getting commitment from companies that might implement the standard so it might be worthwhile if they would implement them as well as using them. That's on the next page. It would be useful to get some commitment? Malcolm: yes, for the standard to stand up you must get people like Oracle and IBM to implement. Equally there is a risk if you go to MS, ... and they say that they will not implement them. Malcolm: there was a nice diagram presented at a previous session about how standards develop, there is an experimentation phase followed by an adoption phase ... if they don't experiment then they are unlikely to adopt. It would be nice if they publicly stated that they are going to adopt. Greg Riccardi: I have a use case ... if I have a file I can identify through a service interface ... and I want to replace the file system with a database ... a natural evolution ... I want to be able to have a fixed reference to that file... need a data recipe ... document based access would help this ... it's real value added for people that want to store non-volatile data... when you put a picture up and you want to drill down to get that data ... It's an interesting use case. Norman: there's two aspects to that: the description of the recipe and the provision of the API. DAIS at the moment provides an API for passing through existing language. Given a description language DAIS could provide an interface. To me, it sounds like a separate document... Malcolm: it does link in quite nicely with 3rd party delivery and data connection language. I think this is an important aspect. It is important to get some standards out but this should be the next thing. Would be happy to be involved. Greg: in the socialisation ... you need value added ... look for places where the standard model does not work ... you want these static references ... OK, the next set of questions, deal with the fact that given the set of functionalities we provide do these fully expose the capabilities of the underlying resource. Also, these deal with the mapping problem. A difficult problem but we seem to have made some progress last night with WSRF but there are other things like WS-I. Greg: we did not reach any consensus as to whether there should be more than one mapping. Norman: that is right. Things have not stabilised yet. There is going to be a WSRF document that is going to be available soon that is going to be useful. Some of the previous issues raised - the singleton resource pattern raised at GGF11 - has produced some interaction and is having an effect ... did not reach a decision yesterday .... one is better than many ... Malcolm: one is going to be unrealistic ... you are never going to get agreement and full adoption from all ... one should not bind to a particular underlying technology ... if we try to enumerate all our dependencies then we will never have any standards... there are many standards that are evolving out there.... standards should be written in a neutral way ... worried that an announcement of this as a complete standard as there will always be some dependencies... We shall wait for the WSRF TC embodiment document and see what people want and try to come up with a decision after that. ... at the moment we are told that we should map to WSRF but at the moment this does not seem to be stable enough. Malcolm: I think standards do not do a detailed mapping that should be an implementation decision. Norman: but you need WSDL and if you are going to be using WSRF there must be a reference to WSRF in the WSDL ... there are other specs like transactions we do not do say anything about... Susan Malaika: the embodiment document will provide ways of mapping... Savas Parastatidis: in order for 2 implementations to be inter-operable they have to have the same binding ... Malcolm: there are so many groups that are saying that we can't move until we have a definition for X ... there are a huge number of such dependencies out there ... we are in a deadlock situation and to get out of this some of these standards need to be pushed out. Savas: that there is the solution of using the current stable tools and when the other iterations are finalised then you push the new spec ... Malcolm: but that will not take into account the fact that it requires to inter-operate with other specs ... Well, will be soliciting feedback from Newcastle/NeSC in the questionnaire. Final questions solicit whether these functionalities meet there are any requirements. Are there any missing questions? Malcolm: what about the non-functional requirements? If we could get people to enumerate the non-functional requirements ... these would then inform the implementations. We then need to be looking for inter-operable implementations .... these would then form the requirements. Service level requirements. Malcolm: how do we get people whose commitment would be valuable if it was made publicly made ... the CERNs, commercial vendors, etc. Have been drawing up a list of commercial contacts in companies that we could feed the questionnaire. Need such views over individual views. --- Norman: One of the things mentioned by Malcolm is to push the draft specs to full specs. We thought it would be useful to get one of the vendors to come up and speak about the current vendors are doing in the WS space. Savas will be speaking about SQL Server 2005 beta 2. Slides available from: http://forge.gridforum.org/projects/dais-wg/document/ggf12-dais-session3-savas/en/1 Could not find any presentations from Microsoft. Do not work Microsoft so this may not be an accurate view. In SQL Server 2005 every stored procedure can be exposed as a web service. Run create end point and then expose it as a web method. Can expose as http, or as an http GET call, e.g. a RESTful service. Yesterday someone mentioned that every stored procedure becomes an end point in this. That is not the case with SQL Server 2005. They have a bunch of methods that can be accessed as web services .. have a bunch of methods such as CancelJob ... you pass the id of the job to cancel that job as a parameter to the operation. Security is managed through http security. No WS-Security as MS has only released an experimental version of WS-Security. Do not expect to see this until this is merged with Indigo ... so it's only in the next version of SQL Server that you will see this. There is no need for IIS. Can run code inside in the database. Can use .Net code. CLS is hosted in the server. ... Wrote a small stored procedure using C#. Can embed this in the database and it's then exposed as a web service. Norman: any questions or clarifications from Savas? Now Susan will be providing an IBM perspective. Susan: Slides not yet available. Going to speak from the point of view of a product person. DB2 offers WS to access DB2 constructs. Can run stored procedures, can drive SQL queries and can drive XML queries and can shred XML. Can take WSDL and produce a SQL function - can view as a client proxy. Will return the result as an XML document. Can call a web service through a nickname wrapper ... DAIS does not do this at all. There are two kinds of way of ?: - administrator states what bits of SQL are exposed as WSs. This is described in an XML file. - Can do this using a dynamic query as in DAIS. All this works with websphere or tomcat. Lots of details in the slides... Do not conform to WSRF yet ... can use sqlxml to return the results ... allows you to generate the results as XML. Can send in a dynamic query. Have only done one shots. If IBM was asked to implement DAIS ... have already implemented what we think people need ... tried to wait DAIS but we had to move on. Went ahead and did it. Would have to put a case for implementing DAIS when we already have things that do similar things. It would come to me for putting a case for implementing DAIS. Not sure if Dave's document said why it would be a good things ... Peter Ziu: would you describe available the functionality as being available to DB2 or any database ... For the inputs coming in ... your first database has to be DB2 ... DB2 federates many database. Can use this stuff with other database. Peter: Is there an underlying standard ... like ODBC, JDBC to integrate database vendors ... We have JDBC to do this ... the kind of options would also be available to JDBC. JDBC is being used behind the scenes. We do not call this an API. Thomas Soddemann: if I change a JDBC driver could I use the mySQL driver? Most vendors optimise their drivers ... a lot of performance work goes into this. ... Mario: would you also have to pursue the endorsement? yes. Dave Pearson: Slides available from: http://forge.gridforum.org/projects/dais-wg/document/ggf12-dais-session3-pearson/en/1 Talks about Oracle. Goes though the Oracle web service feature. Wanted to integrate everything within our application framework. Built on J2EE. In terms of capabilities, there was a demand from customers who used PL/SQL who wanted to make these available as WS end points. A lot of this is about integrating it into our development environment. What we though our customers wanted were: - development simplicity, ... Different delivery capabilities ... to email, pdas,... Oracle WS vs DAIS ... slide list the differences. Savas: what is a complex web service? Something that deals with orchestration with Oracle work flow. Thomas: do we have to buy the process execution manager to execute BPEL? Yes. We bought that from Collaxa. --- Norman invites the speakers to come to the front and take questions from the audience. Malcolm: Can we embed our implementations inside the DBMS? Norman: that is can you write a stored procedure that implements a DAIS procedure and embed this inside your DBMS, say sqlExecute that implements the DAIS interface. Savas: yes, I think that could be done ... Dave: I see a problem, why would we want to do it as it would bypass a lot of the implementation that we have and also what do you do with metadata? Savas: why do you want to do this? Do you not want to expose schema outside your organisation? Dave: database schema are under the control of the DBA which are not exposed. Savas: they do what big companies want ... Dave: we have another product that allows you to construct queries through a query tool without having to know the database schema. Susan: for DB2 we would have to put forward a delta cost as we have already done our own implementation. Would have a staged delivery approach. Greg: can you implement DAIS ... could the existing implementation be used to implement DAIS ... as a third party product. Susan: there is nothing to stop you. Greg: but it is implemented in a non-vendor specific things ... you want to take advantage of internals ... Savas: Once you want to product specific things you cannot assume that there is a common way of things ... that is why you are using WS. Susan: that is a point of WS ... Greg: but that's a particular WS strategy ... Dave: all vendors want to differentiate their products ... if you look at 3rd party access products ... they provide the same functionality across multiple products ... maybe you could go to them ... ??: to me WS if I was cast with doing something quickly would look at the integration that's done by the vendors and expose that ... for the most part should be able to use a remote API ... there already is a lot of work inter-operability ... Susan: that is what DAIS does, it is remoting JDBC without having connections ... the benefit of not having an API is that you don't require any specific things ... do not want clients to have specific drivers but there will not be any database specific vendor optimisation ... there is no way to do DB optimisation... Greg: the fundamental thing that DAIS is the asynchronous part ... Dave: in order for Oracle to see value is that by doing an implementation it would increase its sales and largely I don't think that it would take that view ... currently have different sets of requirements - large number of users working concurrently... etc there's no proof DAIS can do this ... most of the bases are already covered and unless there is a lot of customer demand for asynchronous capabilities and large data sets then there is not much of a push to implement DAIS ... Susan: customers can come with a bunch of standards that you want to implement ... customers do not want to be tied to one vendor ... Dave: I do not disagree with that ... Susan: customers are becoming aware of GGF standards ... Savas: companies are now OGSA advertising compliant tools already ... ---