Thursday, October 21, 2004

Factors Driving a Web Services Infrastructure Yikes - August 23 and no postings.  From a great start last year to spotty over the last 6 months.  Overload of stuff to do in Web services land which has has some possibility of respite now that Jon Maron has joined me as a second PM for core Web services here at Oracle.  We cover the core Web service runtime at Oracle (J2EE 1.4 JAX-RPC, SOAP, WSDL, WS-Security, reliable messaging, addressing, policy etc) and Web services management infrastructure used to manage it - all both of which are used across the Oracle platform. In trying to describe what Jon was getting himself into I noted six categories that drive Web services at Oracle - numbered but not ordered (there are more but I thought these would be of interest to folks): 1. Performance.  When you are building a standards based infrastructure how the heck to you differentiate and make customers interested in your offering.  One area is performance - out-of-the-box you want to be faster than the other guys - this is why Oracle invests so much in things like SpecJAppServer and, not so visibly because of a lack of benchmark, we have the same focus for Web services performance. 2. Manageability.  Next area is the so-called 'ilities' - reliability, scalability, managability, availability.  This is home territory for Oracle and always has been.  Managability has been a big focus for OracleAS 10.1.3 which I have noted several times in this blog over the last year and will continue to be. Web services management is an area in much flux and the platform vendor offerings in this area seem surprisingly weak in my humble opinion.  An article in Oracle Magazine I have in December will give a taste of what we have done here.   Like with development as I note below, this area is ripe for innovation and real value that platform vendors like Oracle can bring to the table.  It is also hugely competitive as small vendors see a place where they can scoot under the radar of the platform players. 3. Interoperability.  One more 'ility'  but unique to Web services.  Not being able to talk to the other guy in Web services basically makes doing Web services pretty close to useless. Why do all that nice SOAizing if you can talk across heterogeneous systems?  A big area of focus for without which the whole Web services story is a house of cards.  WS-I provides the prescriptive approach to interoperability but fundamental to being successful in this part of the story is actually working methodically through our competitive vendor stacks and finding out all the idiosyncracies they have done and determining whether or not the investment to interoperate with them is worth the development effort, regardless of WS-I compliance. 4. Development.  JAX-RPC is but the tip of the iceberg for Web services.  Where the heck is the IntelliJ of Web services?  While we and others do not a bad job of making publishing (top down/bottom up) and consuming Web services not bad and even provide a few tricks - annotations, nice visualizations of services - this space is completely underserved.   There is an boatload of things that a real Web services developer needs - WSDL editing, namespace management, schema management, load testing, unit testing, modeling notations, interface definition tooling, pattern based development ... like management, an area ripe for the taking. 5. Platform integration.  We are the infrastructure providers that many Oracle groups use - JDev builds a design time, Enterprise Manager uses our Mbeans to manage the services, E-Business Suite publishes services, BPEL PM consumes Oracle/any Web service as well as being  a service itself, Identity Management provides the WS-Security hook points for us, Database Web services, Portal renders Web services, Warehouse Builder publishes ETL scripts as services ... Oracle brings a platform to the table of which huge amounts are already service enabled but amongst which there are always more pieces that bring value to the customer table. 6. Those crazy standards.   Some are no brainers - J2EE 1.4.   BPEL.  Others are more interesting - WS-Reliability or WS-ReliableMessaging?  How a little more estoteric: Attachments - choices being: the now dead DIME, the more commonly used MIME, the WS-I sanctioned SwA or the newly emerging MTOM.   The machinations around WS-Addressing and WS-Policy cause interesting heartache to customers who want product now with a certain feature set versus one that is based on open standards and is interoperable.  An interesting maze to navigate which quite often is sadly indecipherable to customers who simply want to ge the job done but due to vendor politicking, posturing and strategy sometimes get left behind. Now that Jon is blogging, hopefully between the two of us you will see more coverage in more detail as Oracle's Web service platform continues to come to market and learn a bit more of how we are tackling these issues.  comment [] 12:11:07 AM