A panel with Paul Hill, Tony Chang, and Bruce Vincent.
Paul Hill kicks off the discussion with the obligatory “What is SOA” slide.
the first step is getting multiple groups to share a definition of what is meant by service.
MIT’s EAG guide has multiple definitions of service.
SOA is a set of design principles that decomposes common functionality into discrete services that can be used by a variety of systems.
It is not a technology.
Service Models – Burton talks about Infrastructure Service Model (authn, authz, auditing, logging, session management, data persistence, transactions) – enterprises try to expose these for their developers; and Business process management, that models the services on business functions instead of underlying technology, e.g. courses, purchase, add/drop, assessment, etc.
Bruce shows a diagram that Stanford has been working on to show services around collaboration tools.
MIT has done some work on categorization of services, but as they started to dive down into the technology to deliver web services, people have come forward with new services that they either need or can provide – e.g. the student services folks came forward and identified a need for geocoding (zip code to lat-long, and also internationally). They had never anticipated that need (and it’s interesting to note that the effort was catalyzed by the roadmap documentation effort).
Tony’s talking about how the UW is undertaking some pilot projects to build community around SOA in order to understand how we might engage as a university around those efforts.
Ken from NYU says that they’re working on an ITIL service catalog approach to listing the services that they offer, rather than taking on an SOA approach at present. Kitty notes that the admin systems group at Michigan is working with units to define data services around admin systems.
Lots of folks are working on building service catalogs – Chicago’s new one is at http://findit.uchicago.edu, Washington’s is at http://www.washington.edu/cac/planning/, Michigan’s old one is at http://www.itd.umich.edu/services/index.ph
The panel is now talking about obstacles to SOA – designing for reusability takes additional time and money – developers don’t know how to think about opening up their interfaces, don’t know how to plan for different toolsets than the ones they’re using. There’s a group at MIT called the IT SPARC group – they engage with projects that don’t have enough funding to take an enterprise approach to report the underfunding of projects that are too narrowly focused – they plea for one-time funding allocations to broaden scope. It’s been partially successful to date, but it’s early in the group’s existence.
Tony’s talking about the UW pilot SOA project, and how the approach is to try to understand how to build a framework that will encourage technology community around shared web services.
Bruce brings up the issue of maintaining backwards compatibility in services.
Paul points out that instrumentation of services is critical in terms of knowing who’s actually using services.
Brendan is talking about the fact that there are also data problems that may be exposed in the services – data definitions can change the meaning of data without notice to the users of services. This can be somewhat mitigated by strong data dictionary practices, but most universities haven’t been really strong in that area. As Tom Barton notes, this is not a new problem.
Paul says that many of the development units on campus haven’t yet made the transition to be able to consume web services, and are more comfortable dealing with traditional interface mechanisms like c libraries, jar files, etc.