Servicemodel Ad Hoc Meeting held Fredericton, NB, Canada, June 26,27, 2001
Updated: July 9, 2001 10:57 - Page Maintained by Allan Doyle

Present - June 26
Doug Nebert
Rob Atkinson
Pedro Gonçalves
Lou Reich
Allan Doyle
Yonsook Enloe
Brian McLeod
Terry Fisher
John Evans
Ananth Rao
[George Percivall - were you there???]

Minutes

This was largely a planning meeting to see who could commit to doing anything over the summer. The outlook seems to be favorable. The following tasks were touched on. More details below.

• Object Modeling and Typer Hierarchies - Interested parties: Rob, Serge, Doug, Lou
• WSDL instantiations and hands-on testing - Galdos, CubeWerx, Pedro, John
• Catalog Request/Operations - Compusult, Ananth, Lou
• Catalog Responses - Jeff, Compusult, Ananth, Lou
• Metadata/Capabilities (migrate to XML Schema, etc.)

Some details about these that may or may not have been discussed:

Object Modeling & Type Hierarchies - This is where we need to think abstractly and document the results in UML. Things like whether the WFS and Catalog share a common base class or whether a Web View Service and a Web Map Service have anything in common. I think the biggest items here are:

• Is there a "Web Object Service" base class?
• Can we tease out the commonalities among the W*S and Catalog services?
• What definitions of Service, Interface, Operation, etc. should we pick and stick with?
• What is the difference between a Taxonomy and a Typer Hierarchy (leading question - see June 27 below)

WSDL - We need people to try converting existing services to this as a test.

Catalog Request/Operations - We need to continue/complete the job of documenting the stateless catalog, either as a subclass of a common base that is shared by the WFS or as it currently stands so people can use it. (Maybe the latter is finished and I've lost track of where to look)

Catalog Responses - We're not quite finished with the Full, Summary, etc. debate. We have to finish this...

Metadata/Capabilities - We have to migrate capabilities to XML Schema, see how it fits the WSDL model. We need service level, interface level, and "content" level metadata. I think we've been too hung up in trying to build Version 2 or 3 quality metadata descriptions and perhaps we should revisit this from the ground up.

Other issues that came up in the meeting - we really need to come to grips with data vs. service and we need to have one or two people plug themselves in to the WSDL and UDDI communities.

Present - June 27
Lou Reich
Doug Nebert
Allan Doyle
Rob Atkinson

We had an impromtu lunch meeting and talked primarily about the data vs. services and object modeling parts.

Taxonomy vs. Type Hierarchy

There was some feeling that we should work with 19119's service taxonomy as a way to classify services. However we also realized that a taxonomy should not be confused with a type hierarchy in the object modeling sense. Maybe a taxonomy is really the semantic view, and a type hierarchy is the schema/syntactic view of the same things.

Semantics vs. Syntactic/Schema

On the data side, an example of this would be "Paths I can ride my bicycle on" as a semantic class, and TIGER/Line as a syntactic schema that can contain these kinds of things.

We also talked about a possible duality between Data Containers (e.g. a Location Organizer Folder) and Service Containers (e.g. a Service Organizer Folder [Who coined that term? Was it George? I think he came up with this at the ISO 19119 editing session in Nashua])

Service Chains

One thought about Opaque Service Chains - Right now we essentially have opaque chains behind the WMS interface. What's happening (e.g. with the SLD) is that we're finding ourselves wanting to provide control inputs to the "hidden" parts of the chain (e.g. to a possible underlying WFS or RDBMS). We have to ask ourselves whether this leads to an undesirable "pollution" of the exposed interfaces. Perhaps it makes more sense to come to grips with service chains in a more explicit way sooner rather than later.

Registries

A goal of developing service registries/catalogs is to not requrie a priori knowledge of server locations. E.g. the keynote Digital Earth demos at DE2001 were based on already knowing which WMS instances were going to be used. True, the WMS instances were queried for their capabilities, but these were just used to construct long pick lists.